Amd Ryzen Threadripper & X399 MegaThread! FAQ & Resources

[goldstone77]

i9 consuming less power than the TR 1950X. Even OC to 4.6GHz it is consuming less power than 3.9GHz TR. Still the i9 is accused of consuming huge amounts of power and putting houses on fire. LOL

At stock the 1950X with MCM package and 6 more cores at stock consumes 20 watts more! Overclocked 45 watts more! Bwahahahaha! This shows the 7900X is getting stomped and you think it's a win? You have to take off those beer googles you are wearing! They are distorting your ability to use logic and reason!

 

[-Fran-]

I know where RyZen is diffused. It doesn't change the fact that traditionally AMD chips have been diffused in Germany (from classic Athlons to recentest A10s) and that Germany has been traditionally pro-AMD, which explains why WCCFTECH had to select sales in that country to make another sensationalist article.

Overall, AMD marketshare has increased by only 4% since Q4 2016, and the relatively weak sales of RyZen didn't help AMD to compensate for losses reported this year. Regarding gaming niche, AMD is loosing marketshare. AMD lose another 1.37% in August

http://store.steampowered.com/hwsurvey/processormfg/

http://store.steampowered.com/hwsurvey/processormfg/

You're losing view of the whole picture, Juan.

I don't expect TR to be a records sale thing nor turn the tide; same for Ryzen TBH. Intel fanbois are plenty everywhere, including Germany FYI. For heavy duty tasks AMD has reached parity with Intel, but it's not *better* in all metrics, so you will always argue "but this! and this other important thing! not better, hence worse!", and it's an annoying circle-jerk you're creating.

What the numbers we're giving indicate a simple thing: AMD has a competitive CPU. If you really want to understand that, you have to go back and see how Bulldozer and Piledriver numbers were in the first half of the year for each. This is nothing like that time. And Steam, obviously, won't account for things that escape the gaming market. But more importantly, they account for laptops and desktop PCs. I'm still awaiting some AMD laptops, specially in the cheaper arena with their APUs, so start displacing some of Intel's dominance there. If you really want to play the "market share" figures, wait until then for a fun time seeing how the graphs *will* fluctuate... Well, if the APUs don't suck, obviously. I'm cautiously optimistic for them though.

Cheers!

 

[genz]

Those disputes between AMD and Intel were solved when both companies signed an agreement that included cross-licensing of patents.

You have said this numerous times, exactly the same. Your wording is evasive. There is no exchange of technologies within most X-lic agreements. It is for separately developed technology and bears no resemblance to an agreement like AMD licensing them Radeon IGPU tech, which would be specific, limited and heavily technologically walled off so advantages couldn't be transplanted to Iris Pro, esp VCA. I cannot waste more time on this so I will direct you to the example we used and show you how you were so knowledgeable you got caught in a rather obvious trap.

The MMX cross licensing agreement did not give AMD the right to use Intel's MMX technology in their processors or any others... it existed purely because AMD developed a way to add Matrix Math instructions within their arch then realised that Intel has trademarked the initials to Matrix Math Extensions so they could neither say they had support for MMX nor Matrix Math Extensions. It's illegal to use abbreviations in trademarks for this reason. AMD sued and the 'cross-license agreement' was a compromise to allow Intel to continue using a technically illegal abbreviation as a trademark without further challenge from AMD, whilst letting AMD say it supported MMX.... sharing the name between them and still forcing others to have to litigate to claim MMX support without infringing on a trademark. It was purely marketing/law/biz based and both AMD and Intel had already separately completed development. They didn't share a bit of tech at any stage so you're full of proverbial.

When I mean "access to patents", evidently I mean that AMD can use technology based in Intel patents and vice verse, without being sued for patent violation. That is what the cross-licensing agreement granted to both companies: access to other's technology. I will repeat once again that the agreement is not limited to x86, but also includes other stuff, including graphics. Sure that you did heard rumors about a hybrid chip with Intel CPU and Radeon GPU. Guess what? Three different sources confirmed me that the prototype is real.

I have 3 prototype Apple computers running on Developer platforms. I can disclose this to you because I have a license, but if I kept my mouth shut I could have anything in prototype here I wanted without breaking any laws and hearsay would be hearsay.

Who has said you that the change from ring to mesh is a response to AMD? Intel has been using mesh in the Phi line, and we know that both ordinary Xeon line and Phi line are converging. We have known for a looong time that SKL Xeons would get AVX-512 support. I have known it since 2014 and it is evident that the SKL Xeon design was planned many years ago even before "Zen" was a public codename for a future muarch.

New to me, but I could never see the point of Phi on a PCI-E bus. I don't VM. For everyone else it's like trying to sell a 50 x P2... MMX cards on a single x16. Now that's somewhere that could do with a talk about latency!

The average overclock for the 7900X is 4549MHz. This means some people is achieving lower overclock and others are achieving higher overclock; this is the proper definition of average. Der8auer chip got 6% higher overclock than the average.

At 4.6Ghz Techspot quoted the Fully Integrated Voltage Regulator on the CPU as reporting 400w power draw. Hold that thought.

Der8auer praised the i9 chip, I quoted him just above and then you now change the argument to motherboards. Yes, he criticized some mobos for having an incorrect desing. He demonstrated that some incorrectly desinged mobos have a fancy (useless) plate above the VRM region that works as a heat insulator and overheats the VRM, generating a problem. He demonstrated that solving the problem is so easy as retiring that fancy plate from the mobo and ensuring proper cooling goes to the VRM region..

BIOS updates brought power draw down... lower performance too. Power efficiency came from somewhere. Further fuel to the flames of 'SkyX is a hack' and the BIOS and the motherboard issues were likely down to the fact that the 7900X wasn't announced with the original lineup... because the original lineup still wasn't competitive.

Buying a VRM fan was what he said to do. People haven't needed to do that since Pentium 4 EE... except when LN2 cooling. Bad sign, and not a mistake I see being made by so many vendors at once, only on lower end products, and without similar issues on other competing platforms with similar TDPs, that they issued boards for months before.

The interest in the X299 platform is reflected in the high number of mobos available and chip sales. Your own amazon.com best selling chip says that the i9-7900X is just behind the TR 1950X in sales, and selling much better than the TR 1920X. The i9-7900X is #34 whereas the TR 1920x is #56.

I am surprised than the SKL-X chips are selling so well despite so many biased reviews and people paid for posting negative comments on forums and stores. Check the nest link. This guy didn't write a review of the Intel chip, he wrote a rant to try to increase sales of AMD chips

Then you surprise me. As someone who so keenly talked up the AMD thread, it would seem obvious to me that you of all people should know that Intel has a huge market share lead, a 10 year performance lead and millions of faithful customers. It would take more than one product range to sway them from buying blue when they have for a decade. Must I mention the Pentium 4 again because we are both aware it sold quite well.

 

[goldstone77]

Spoiler

Spoiler

Spoiler

Sorry I'm late to the party, and this is now an old review that came out on August 10th. I still think it's worth posting so here you go!
[video="https://www.youtube.com/watch?v=G9JR_v-4BaQ&ab_channel=HardwareUnboxed"][/video]
AMD Ryzen Threadripper 1950X & 1920X Review
By Steven Walton on August 10, 2017
"After much speculation and an official unveiling last week, it's finally time to see if Threadripper can bring competition to the high-end desktop segment while delivering the value and efficiency we've come to expect from other Ryzen processors.
Before we jump into the benchmarks, here's how we've set up the comparison. For this review, we'll be testing the new Threadripper 1950X and 1920X along with the Ryzen 7 1700 and 1800X, as well as Intel's competing Core i7-7800X, 7820X and Core i9-7900X.
Compared to the Ryzen 7 1800X, which until now was leading AMD's pack, the Threadripper 1920X comes at a 75% price premium and offers 50% more cores. You also get quad-channel memory support and significantly more PCI Express lanes."
"While we're curious to see how Threadripper fares against AMD's more affordable lineup, I'm more interested in learning how they compare against Intel's Skylake-X parts, so you can expect an emphasis on those results.
The Skylake-X CPUs were installed on an Asrock Fatal1ty X299 Professional Gaming i9 motherboard, while the Threadripper CPUs were tested on the Asrock X399 Taichi motherboard. Both platforms were fitted with G.Skill's TridentZ DDR4-3200 CL14 64GB quad-channel memory kit. Big thanks go to G.Skill for providing all that memory."
"The X299 platform was installed on the Praxis Wetbench and was cooled using Thermaltake's Pacific RL360 custom liquid cooled kit. Meanwhile Threadripper was tested on a sheet of foam on my desk and was cooled by the Thermaltake Floe Riing RGB 360 AIO liquid cooler."

Overclocking Results
"As expected the Threadripper CPUs were able to hit 4 GHz but unfortunately like Ryzen 7 weren’t able to go beyond that point. Extreme voltages might see you hit 4.1 - 4.2 GHz like what we’ve seen with some Ryzen CPUs but for 24/7 usage 4 GHz looks to be the limit."

"Anyway at 4 GHz the 1950X spat out an incredible score of 3408 pts in Cinebench though due to the way XFR works this overclock actually slightly reduced the single thread performance. You could probably fine tune the overclock to avoid this but with time against us we didn’t play around with the overclocking too much."

"Overclocked the Blender render time for the 1950X was cut down by an impressive 11% while the 1920X time was reduced by 6%. This means even overclocked the Core i9-7900X wasn’t much faster than the 1920X and wasn’t able to match the stock 1950X in this application."

"Interestingly this time when testing with Corona the 1920X saw a massive 17% performance improvement when overclocked while the 1950X was just 7% faster. This meant both overclocked Threadripper parts were much faster than the overclocked 7900X."

"Back to the kind of margins seen in Blender we find that while the 1950X was 9% faster overclocked the 1920X was just 5% faster. This meant when testing with Premiere Pro CC the 1920X did trail the Core i9-7900X while the 1950X was quite a lot faster."

"Finally we have the overclocking power consumption figures and here Threadripper looks quite good, especially in relation to the Core i7-7900X. Whereas the Intel CPU pushed total system consumption to 388 watts, the 1950X hit 358 watts, an 8% reduction for what was often between 20 - 30% more performance."

UMA vs. NUMA
"As mentioned on the last page, you switch between memory access modes on Threadripper processors. 'Distributes' mode uses 'Uniform Memory Access' is enabled by default so this is what we've been testing with so far. However, you can switch to 'local' mode for 'Non-Uniform Memory Access' and this allows each of the Zeppelin dies to prioritize which cores access certain parts of the system memory. This basically prioritizes the nearest cores to improve overall latency for gaming applications that tend to place a premium on fast memory access. So let's see how this impacts gaming performance as well as a few productivity workloads..."

"Switching to NUMA has improved the Battlefield 1 performance and now Threadripper is delivering similar frame rates to the Ryzen 7 1800X. In fact, the 1950X creeps ahead ever so slightly and now is that much slower than the Core i7-7820X and Core i9-7900X."

"F1 2016 also sees decent gains when switching to NUMA though this time isn't able to match the Ryzen 7 1800X and this means it's also well down on the Intel CPUs."

"Performance in Civilization VI was already great and no extra performance was had when switching to NUMA."

"Looking at application performance, we see a slight performance decline with POVRay, but nothing serious."

"Blender also saw a slight decline in performance with the 1950X being about 4% slower using NUMA."

"Last up we have the Corona test, in which both NUMA and UMA offered the same performance. There's not much to see here, so let's move on to some consumption figures."

"In a nutshell, for the same price you get more CPU performance, lower power consumption and improved operating temperatures with Threadripper, so why would you buy the Core i9-7900X over the 1950X?
Intel recently announced the specs for its upcoming 12, 14, 16 and 18-core Skylake-X parts. At a guess, I would say consumers will need to spend at least $1,400 on the i9-7940X to match AMD's 1950X and I seriously doubt the 40% price premium will be worth it.
The X299 platform's lack of ECC memory support is another issue. Whereas Threadripper supports ECC, the Skylake-X chips don't and that means anyone serious about their workstation won't even consider Intel's high-end desktop platform. Unless Intel is willing to budge on pricing I can't see why anyone would invest in X299.

At the beginning of 2017, who would have thought we'd see AMD dethrone Intel at the very top of the high-end desktop CPU segment. It certainly wasn't us, but we're glad we can look forward to some competition. It's an exciting time to be a PC enthusiast."
"Pros: The Threadripper 1920X offers more performance, uses less power and runs cooler than the Core i9-7900X. It also supports ECC memory (Intel's X299 platform doesn't). Compared to the Ryzen 7 1800X, the 1920X touts quad-channel memory support and more PCIe lanes.
Cons: The $1,000 Threadripper 1950X is competitively priced but nonetheless comes at a hefty premium. Threadripper falls a tad behind when it comes to gaming."

I have three objections to that review.

(i) They tested different versions of the essentially the same kind of workload. They tested rendering with Corona, CineBench, Blender, POVRay... This is useless to get a overall picture of the performance of a chip. For instance TR 1950X is faster than i9-7900X in Blender, but the 1950X is slower in WinRAR. Moreover, they didn't check any rendering application where the TR 1950X is slower. The i9-7900X is faster in Mental Ray or V-Ray for instance.

(ii) They used overclock RAM on the TR chips, which automatically overclocks the IF interconnect. but they didn't overclock the mesh interconnect on the SKL-X chips.

(iii) Pricing is incorrect. They give the price of the i9 as $1150, but even AMD reports the i9 as a $999 chip.

Therefore the performance/price graphics they gave are twice biased. One because they only measure rendering performance on benchmarks favoring AMD, and another because increased pricing of Intel chips.

<i> I don't agree with your suggestions for inclusion or exclusion and neither did the reviewer, because these benchmarks shows real world performance for professionals who would and are buying these chips. Notice price drop on the notoriously not popular 7900X since the release of ThreadRipper.
<ii> Your complaint only shows the weakness of the mesh, and doesn't show real world applications for majority of the population that would just pop a memory dimm in have it work at it's registered speed which is the case here.
<iii> The review is almost a month old now. Amazon the 7900X is only $962.99. This price demonstrates the unpopularity of the entire skylake-x line in general. No professional wants to buy this chip, because of the numerous issues with this entire product line that caused outrage across the popular YouTubers! ThreadRipper still remains the better value!

(i) My remark was about real world performance. A complete review would have tested the chips on rendering, encoding, compiling, scientific computations, finance,... Instead the HU review was mostly rendering, rendering, rendering, rendering, rendering,... and games. Moreover even in the subset of rendering they didn't use any rendering application where the i9-7900X is faster than the TR 1950X like V-ray.
(ii) The mesh interconnect is fine. My complain is about how biased are some recent reviews like HU that compare overclock to stock. To be unbiased, we would compare stock to stock or overclock to overclock.
(iii) The i9-7900X did launch at $999. This is the price reported even in AMD slides. The HU review inflated the prices of Intel chips and their performance/price graphs are biased.

Amazon selling it now with a 3% discount, just disproves your claim about unpopularity. From what I can see the chip and others in the SKL-X are selling well, despite the huge bias shown in the professional press. I have checked several owner threads and people is very happy with their SKL-X chips. Even der8auer wrote something about this bias from the media:

From my point of view Skylake-X is a pretty impressive CPU. We now have have a 10-core that's a lot cheaper than the Broadwell-E before and it clocks a lot higher. Even with the stock Intel paste I was able to reach 4.8 GHz on the CPU (i9-7900X) using Corsair 280 (mm) AIO. So there is still some headroom, I guess if you use a custom new water cooling you might be able to hit 4.9 GHz on a very very good chip without delidding. So after this test I delidded the CPU and replaced the stock TIM with liquid metal and this helped me push the CPU with an AIO to 5 GHz.

So we had Broadwell-E before who could run like 4.3-4.4 GHz, it cost 1700€. Now we have Skylake-X which is a lot cheaper and we CAN push it to 5 GHz, so what's all this negative press about? I don't really understand it. So from my point of view this is a very impressive CPU, so we have very high single-thread performance on the 10-core and also high MT performance, which we didn't have before in a Broadwell-E (probably comparing to mainstream).

Der8auer used a Pre-tested Skylake-X i9 7900X using AIO! Most people can only get 4.4GHz-4.5GHz with custom liquid coolers! You must have forgot about De8auers "The X299 VRM Disaster" where he talks about X299 systems pulling so much power they could catch fire! Every YouTuber, Hardware Unboxed, and Tom's Hardware have all gave dismal reviews of the 7900X! You are not watching the entire video and using his statements without all the facts. I've went over this in great detail already on the X299 thread! Also, thermal throttling and VRM consuming over 400 watts of power added to the scandal! You really need to stop over hyping the 7900X, because it's notoriously unpopular, and your statements to the contrary defy history. Unless you are going to put up a real argument with verifiable numbers and links to back up your claims don't bother replying!

[video="https://www.youtube.com/watch?v=f7BqAjC4ZCc&ab_channel=der8auer"][/video]

There was very little interest in the X299 platform

Edit: Adding to context in rank of best selling
ThreadRipper 1950X 25/100 7800X 27/100 7900X 31/100 7820X 34/100 1900X 40/100 1920X 51/100
https://www.amazon.com/Best-Sellers-Computers-Accessories-Computer-CPU-Processors/zgbs/pc/229189#2
ThreadRipper 1950X 62/100 1920X 65/100 7820X 74/100 7900X 88/100 amazon UK best sellers list
https://www.amazon.co.uk/Best-Sellers-Computers-Accessories-Processors/zgbs/computers/430515031#5

The average overclock for the 7900X is 4549MHz. This means some people is achieving lower overclock and others are achieving higher overclock; this is the proper definition of average. Der8auer chip got 6% higher overclock than the average.

Der8auer praised the i9 chip, I quoted him just above and then you now change the argument to motherboards. Yes, he criticized some mobos for having an incorrect desing. He demonstrated that some incorrectly desinged mobos have a fancy (useless) plate above the VRM region that works as a heat insulator and overheats the VRM, generating a problem. He demonstrated that solving the problem is so easy as retiring that fancy plate from the mobo and ensuring proper cooling goes to the VRM region. Some of his problems were also related to the PSU that he chose

https://www.techpowerup.com/234744/intel-x299-platform-called-a-vrm-disaster-by-overclocker-der8auer

Tomshardware also wrote something about motherboards. If I am not misguided Tomshardware wrote that some motherboards violate Intel specifications. Thus it is just a question of choosing the correct mobo now and/or awaiting for the manufactures with issues to release revisions of their early desings. Also initial reviews used launch BIOS with problems that affected performance and power consumption. Some sites retested with the new BIOS and measured lower power consumption.

As demonstrated in a fomer post, the Hardware Unboxed comparison of TR vs i9 was incomplete and biased. And their review of the 7800X was even worse. Probably this was the worse review ever published! First, they tested an engineering sample instead a retail chip. Second, they used a mobo incompatible with the chip and even managed to burn the chip. Third, they crippled the performance of the chip by using GPU-bound and frame-limited settings.

There are some very wrong reviews of SKL-X and many biased SKL-X vs TR reviews: overclocked AMD chips compared to Intel chips on stock settings. Cherry picking benchmarks to favor throughput. Measuring performance on non-AVX workloads but measuring power consumption in AVX workloads, generating a bias in the measurement of efficiency,... This is a huge scandal!

The interest in the X299 platform is reflected in the high number of mobos available and chip sales. Your own amazon.com best selling chip says that the i9-7900X is just behind the TR 1950X in sales, and selling much better than the TR 1920X. The i9-7900X is #34 whereas the TR 1920x is #56.

I am surprised than the SKL-X chips are selling so well despite so many biased reviews and people paid for posting negative comments on forums and stores. Check the nest link. This guy didn't write a review of the Intel chip, he wrote a rant to try to increase sales of AMD chips

https://www.amazon.com/gp/product/B0745486S1/ref=s9_acsd_simh_hd_bw_bxcb_c_x_2_w?pf_rd_m=ATVPDKIKX0DER&pf_rd_s=merchandised-search-11&pf_rd_r=52F601VWB4WCG4A9G553&pf_rd_t=101&pf_rd_p=1381e980-855e-537b-94df-b4f4d642d410&pf_rd_i=229189

I agree with der8auer: "so what's all this negative press about? I don't really understand it."

There is no limit to what you will do to try and skew some hypothetical win! People are not making negative reviews, because they are getting paid! History tells us that is a dirty underhanded Intel tactic that is well documented!

Edit: Since when does Intel offer discounts on chips because that are selling good??? Bwahahaha you make me laugh! They increase prices on chips that sell good! They are lowering the price to try and be competitive with the superior selling 1950X!!!!!!!!

 

[juanrga]

i9 consuming less power than the TR 1950X. Even OC to 4.6GHz it is consuming less power than 3.9GHz TR. Still the i9 is accused of consuming huge amounts of power and putting houses on fire. LOL

At stock the 1950X with MCM package and 6 more cores at stock consumes 20 watts more! Overclocked 45 watts more! Bwahahahaha! This shows the 7900X is getting stomped and you think it's a win? You have to take off those beer googles you are wearing! They are distorting your ability to use logic and reason!

Nothing of that invalidates the fact that people that complained because the i9 was consuming ~400W, is now silent when TR consumes more power. The double standard is evident.

Spoiler

Not all cores are the same. AMD uses simpler narrow cores that consume less power each. Not a mystery here, simply a consequence of the laws of physics. That is the reason why Atoms cores are simpler than Xeon cores.

Not all process nodes are the same. 14LPP prioritizes efficiency over performance. That is why the 1950X hits a wall at 3.9GHz, whereas the i9 can hit above 4.5GHz.

Not all overclocks are the same. A chip doesn't consume the same power at 3.9GHz than at 4.6GHz. Note that increasing clocks from 3.4GHz to 3.9GHz in the 1950X has increased the power consumption by more than 100W. Increasing clocks from 3.3GHz to 4.6GHz in the 7900x has increased the power consumption only by 76W.

Pushing the 1950X to 4.6GHz would increase power consumption to about 700W. The i9 does ~500W at same clocks.

 

[goldstone77]

Spoiler

i9 consuming less power than the TR 1950X. Even OC to 4.6GHz it is consuming less power than 3.9GHz TR. Still the i9 is accused of consuming huge amounts of power and putting houses on fire. LOL

At stock the 1950X with MCM package and 6 more cores at stock consumes 20 watts more! Overclocked 45 watts more! Bwahahahaha! This shows the 7900X is getting stomped and you think it's a win? You have to take off those beer googles you are wearing! They are distorting your ability to use logic and reason!

Nothing of that invalidates the fact that people that complained because the i9 was consuming ~400W, is now silent when TR consumes more power. The double standard is evident.

Spoiler

Not all cores are the same. AMD uses simpler narrow cores that consume less power each. Not a mystery here, simply a consequence of the laws of physics. That is the reason why Atoms cores are simpler than Xeon cores.

Not all process nodes are the same. 14LPP prioritizes efficiency over performance. That is why the 1950X hits a wall at 3.9GHz, whereas the i9 can hit above 4.5GHz.

Not all overclocks are the same. A chip doesn't consume the same power at 3.9GHz than at 4.6GHz. Note that increasing clocks from 3.4GHz to 3.9GHz in the 1950X has increased the power consumption by more than 100W. Increasing clocks from 3.3GHz to 4.6GHz in the 7900x has increased the power consumption only by 76W.

Pushing the 1950X to 4.6GHz would increase power consumption to about 700W. The i9 does ~500W at same clocks.

You are just trying to be an Intel apologist! The fact is that for 20 more watts you get 6 more cores with ThreadRipper! That's an amazing performance metric in itself! It's doesn't take a genius too see you're just griefing over Intel!

 

[juanrga]

I know where RyZen is diffused. It doesn't change the fact that traditionally AMD chips have been diffused in Germany (from classic Athlons to recentest A10s) and that Germany has been traditionally pro-AMD, which explains why WCCFTECH had to select sales in that country to make another sensationalist article.

Overall, AMD marketshare has increased by only 4% since Q4 2016, and the relatively weak sales of RyZen didn't help AMD to compensate for losses reported this year. Regarding gaming niche, AMD is loosing marketshare. AMD lose another 1.37% in August

http://store.steampowered.com/hwsurvey/processormfg/

http://store.steampowered.com/hwsurvey/processormfg/

You're losing view of the whole picture, Juan.

I don't expect TR to be a records sale thing nor turn the tide; same for Ryzen TBH. Intel fanbois are plenty everywhere, including Germany FYI. For heavy duty tasks AMD has reached parity with Intel, but it's not *better* in all metrics, so you will always argue "but this! and this other important thing! not better, hence worse!", and it's an annoying circle-jerk you're creating.

What the numbers we're giving indicate a simple thing: AMD has a competitive CPU. If you really want to understand that, you have to go back and see how Bulldozer and Piledriver numbers were in the first half of the year for each. This is nothing like that time. And Steam, obviously, won't account for things that escape the gaming market. But more importantly, they account for laptops and desktop PCs. I'm still awaiting some AMD laptops, specially in the cheaper arena with their APUs, so start displacing some of Intel's dominance there. If you really want to play the "market share" figures, wait until then for a fun time seeing how the graphs *will* fluctuate... Well, if the APUs don't suck, obviously. I'm cautiously optimistic for them though.

No one here has negated that Zen is more competitive than Bulldozer or Piledriver, because it is more competitive. The point was other...

The point is that we have been reading enough hype and bold claims! Before launch the hype train pretended Zen was going to be the new K8 and take over. It didn't happen. Immediately after launch, reports about RyZen selling as hot cakes. It didn't happen and AMD soon had to reduce the pricing of the chips. The 1700X launch pricing was of $399. I can find it today for less than $300. It is a 25% discount... and it is not enough. I expect further discounts in next months.

A bit latter reports about RyZen giving to AMD the greatest marketshare increase in the history. It was fake. Even today Ryzen only brings 4% higher marketshare to AMD.

Now are the sales on a well-known pro-AMD country, where AMD have invested millions and millions.

Aren't three years enough of continuous hype and fake data?

 

[juanrga]

Who has said you that the change from ring to mesh is a response to AMD? Intel has been using mesh in the Phi line, and we know that both ordinary Xeon line and Phi line are converging. We have known for a looong time that SKL Xeons would get AVX-512 support. I have known it since 2014 and it is evident that the SKL Xeon design was planned many years ago even before "Zen" was a public codename for a future muarch.

New to me, but I could never see the point of Phi on a PCI-E bus

The Phi moved to standalone CPU with KNL. The PCie has been eliminated.

The PCIe card versions of the Phi are only for legacy users.

 

[juanrga]

There is no limit to what you will do to try and skew some hypothetical win! People are not making negative reviews, because they are getting paid! History tells us that is a dirty underhanded Intel tactic that is well documented!

Edit: Since when does Intel offer discounts on chips because that are selling good??? Bwahahaha you make me laugh! They increase prices on chips that sell good! They are lowering the price to try and be competitive with the superior selling 1950X!!!!!!!!

The launch price was $999. The current price is only a 4% discount.

The chips are being really discounted are those from AMD. For instance the 1700X did launch at $399, you can find it today by $100 less. This is a 25% discount. Expect further discounts in next months

 

[juanrga]

Spoiler

i9 consuming less power than the TR 1950X. Even OC to 4.6GHz it is consuming less power than 3.9GHz TR. Still the i9 is accused of consuming huge amounts of power and putting houses on fire. LOL

At stock the 1950X with MCM package and 6 more cores at stock consumes 20 watts more! Overclocked 45 watts more! Bwahahahaha! This shows the 7900X is getting stomped and you think it's a win? You have to take off those beer googles you are wearing! They are distorting your ability to use logic and reason!

Nothing of that invalidates the fact that people that complained because the i9 was consuming ~400W, is now silent when TR consumes more power. The double standard is evident.

Spoiler

Not all cores are the same. AMD uses simpler narrow cores that consume less power each. Not a mystery here, simply a consequence of the laws of physics. That is the reason why Atoms cores are simpler than Xeon cores.

Not all process nodes are the same. 14LPP prioritizes efficiency over performance. That is why the 1950X hits a wall at 3.9GHz, whereas the i9 can hit above 4.5GHz.

Not all overclocks are the same. A chip doesn't consume the same power at 3.9GHz than at 4.6GHz. Note that increasing clocks from 3.4GHz to 3.9GHz in the 1950X has increased the power consumption by more than 100W. Increasing clocks from 3.3GHz to 4.6GHz in the 7900x has increased the power consumption only by 76W.

Pushing the 1950X to 4.6GHz would increase power consumption to about 700W. The i9 does ~500W at same clocks.

You are just trying to be an Intel apologist! The fact is that for 20 more watts you get 6 more cores with ThreadRipper! That's an amazing performance metric in itself! It's doesn't take a genius too see you're just griefing over Intel!

I explained in the spoiler the flaws with that statement.

 

[goldstone77]

There is no limit to what you will do to try and skew some hypothetical win! People are not making negative reviews, because they are getting paid! History tells us that is a dirty underhanded Intel tactic that is well documented!

Edit: Since when does Intel offer discounts on chips because that are selling good??? Bwahahaha you make me laugh! They increase prices on chips that sell good! They are lowering the price to try and be competitive with the superior selling 1950X!!!!!!!!

The launch price was $999. The current price is only a 4% discount.

The chips are being really discounted are those from AMD. For instance the 1700X did launch at $399, you can find it today by $100 less. This is a 25% discount. Expect further discounts in next months

Off topic
Amd Ryzen Threadripper & X399 MegaThread! FAQ & Resources

 

[morenolee]

These are amazing benchmarks for the type of work I do (content creation)...amazing.

 

[goldstone77]

Spoiler

i9 consuming less power than the TR 1950X. Even OC to 4.6GHz it is consuming less power than 3.9GHz TR. Still the i9 is accused of consuming huge amounts of power and putting houses on fire. LOL

At stock the 1950X with MCM package and 6 more cores at stock consumes 20 watts more! Overclocked 45 watts more! Bwahahahaha! This shows the 7900X is getting stomped and you think it's a win? You have to take off those beer googles you are wearing! They are distorting your ability to use logic and reason!

Nothing of that invalidates the fact that people that complained because the i9 was consuming ~400W, is now silent when TR consumes more power. The double standard is evident.

Spoiler

Not all cores are the same. AMD uses simpler narrow cores that consume less power each. Not a mystery here, simply a consequence of the laws of physics. That is the reason why Atoms cores are simpler than Xeon cores.

Not all process nodes are the same. 14LPP prioritizes efficiency over performance. That is why the 1950X hits a wall at 3.9GHz, whereas the i9 can hit above 4.5GHz.

Not all overclocks are the same. A chip doesn't consume the same power at 3.9GHz than at 4.6GHz. Note that increasing clocks from 3.4GHz to 3.9GHz in the 1950X has increased the power consumption by more than 100W. Increasing clocks from 3.3GHz to 4.6GHz in the 7900x has increased the power consumption only by 76W.

Pushing the 1950X to 4.6GHz would increase power consumption to about 700W. The i9 does ~500W at same clocks.

You are just trying to be an Intel apologist! The fact is that for 20 more watts you get 6 more cores with ThreadRipper! That's an amazing performance metric in itself! It's doesn't take a genius too see you're just griefing over Intel!

I explained in the spoiler the flaws with that statement.

Your statements didn't change anything.

 

[aldaia]

Moreover even in the subset of rendering they didn't use any rendering application where the i9-7900X is faster than the TR 1950X like V-ray.

Here they do use V-ray. Looks like 1950X trounces the i9-7900X by 36%, even the inferior 1920X is faster than i9-7900X in "rendering application where the i9-7900X is faster than the TR 1950X like V-ray"

The review must be wrong! 😀😀😀

*** Stuff unrelated to V-ray removed ***

Bonus: V-Ray benchmark from other review, with the i9 being faster than the 1950x

The review you posted is reporting times of 200+ seconds for both 1950X and 7900X.

Oficcial V-ray submissions on 09//03/17 for:
1900X range from: 00:40.608 to 00:48.409
7900X range from: 00:52.392 to 01:03.037

The review I posted is inline with this data, yours is off by a factor of 4X. This is certainly not the official V-ray benchmark. I know someone that usually claims whole reviews are wrong because of small (<10%) discrepancies in a single benchmark.

 

[juanrga]

Spoiler

i9 consuming less power than the TR 1950X. Even OC to 4.6GHz it is consuming less power than 3.9GHz TR. Still the i9 is accused of consuming huge amounts of power and putting houses on fire. LOL

At stock the 1950X with MCM package and 6 more cores at stock consumes 20 watts more! Overclocked 45 watts more! Bwahahahaha! This shows the 7900X is getting stomped and you think it's a win? You have to take off those beer googles you are wearing! They are distorting your ability to use logic and reason!

Nothing of that invalidates the fact that people that complained because the i9 was consuming ~400W, is now silent when TR consumes more power. The double standard is evident.

Spoiler

Not all cores are the same. AMD uses simpler narrow cores that consume less power each. Not a mystery here, simply a consequence of the laws of physics. That is the reason why Atoms cores are simpler than Xeon cores.

Not all process nodes are the same. 14LPP prioritizes efficiency over performance. That is why the 1950X hits a wall at 3.9GHz, whereas the i9 can hit above 4.5GHz.

Not all overclocks are the same. A chip doesn't consume the same power at 3.9GHz than at 4.6GHz. Note that increasing clocks from 3.4GHz to 3.9GHz in the 1950X has increased the power consumption by more than 100W. Increasing clocks from 3.3GHz to 4.6GHz in the 7900x has increased the power consumption only by 76W.

Pushing the 1950X to 4.6GHz would increase power consumption to about 700W. The i9 does ~500W at same clocks.

You are just trying to be an Intel apologist! The fact is that for 20 more watts you get 6 more cores with ThreadRipper! That's an amazing performance metric in itself! It's doesn't take a genius too see you're just griefing over Intel!

I explained in the spoiler the flaws with that statement.

Your statements didn't change anything.

The thoughts hidden in the spoiler explain why statements like your "20 more watts you get 6 more cores" are misguided. I reproduce the thoughts here:

Not all cores are the same. AMD uses simpler narrow cores that consume less power each. Not a mystery here, simply a consequence of the laws of physics. That is the reason why Atoms cores are simpler than Xeon cores.

Not all process nodes are the same. 14LPP prioritizes efficiency over performance. That is why the 1950X hits a wall at 3.9GHz, whereas the i9 can hit above 4.5GHz.

Not all overclocks are the same. A chip doesn't consume the same power at 3.9GHz than at 4.6GHz. Note that increasing clocks from 3.4GHz to 3.9GHz in the 1950X has increased the power consumption by more than 100W. Increasing clocks from 3.3GHz to 4.6GHz in the 7900x has increased the power consumption only by 76W.

Pushing the 1950X to 4.6GHz would increase power consumption to about 700W. The i9 does ~500W at same clocks.

Of course, nothing of that change a bit the main fact that the people accusing the i9 of consuming huge amounts of power remained silent when TR consumed still more power.

 

[juanrga]

Moreover even in the subset of rendering they didn't use any rendering application where the i9-7900X is faster than the TR 1950X like V-ray.

Here they do use V-ray. Looks like 1950X trounces the i9-7900X by 36%, even the inferior 1920X is faster than i9-7900X in "rendering application where the i9-7900X is faster than the TR 1950X like V-ray"

The review must be wrong! 😀😀😀

*** Stuff unrelated to V-ray removed ***

Bonus: V-Ray benchmark from other review, with the i9 being faster than the 1950x

The review you posted is reporting times of 200+ seconds for both 1950X and 7900X.

Oficcial V-ray submissions on 09//03/17 for:
1900X range from: 00:40.608 to 00:48.409
7900X range from: 00:52.392 to 01:03.037

The review I posted is inline with this data, yours is off by a factor of 4X. This is certainly not the official V-ray benchmark. I know someone that usually claims whole reviews are wrong because of small (<10%) dicrepancies in a single benchmark.

No one said anything about anything 'official'.

Also one thing is the discrepancy between reviews that measure exactly the same and another thing are reviews that measure different scenes on a rendering engine. In the first case we have measurement errors. In the second case we have variation due to different workloads.

The scene used in the HFR review is more complex and realist and that is because it takes ~4x more time to render. Same happened with Blender. The toy "Zen scene" used by AMD marketing gave RyZen a 3% IPC lead over Broadwell. A more realistic scene gives to Broadwell the IPC lead.

It would be weird to pretend that Tomshardware review is "wrong", because the result with the Island scene doesn't agree with the Zeb scene used by AMD marketing. Therefore stop pretending that HFR review is wrong only because the result of the benchmark is not what you want.

My point was that HU review graphs posted above are biased because only reproduced cases where TR was faster, and didn't give a single benchmark where i9 is faster. But, we know HU reviews are biased. It was demonstrated.

But I have to thank you for mentioning the V-Ray 'official' benchmark. As you know they give a CPU list and a GPU list. The best value obtained by the TR 1950x is 00:40.608. The best value obtained by GPUs are in the 00:14.359 range. I.e. GPUs are ~2x faster. The TR chips shine on workloads that are better run on GPUs.

As I tried to explain more than once, the CPU must at least shine on latency workloads and the GPUs must at least shine on throughout workloads. AMD Zen design is oriented to throughput. Zen lags in latency-sensitive workloads, but works fine in workloads where a GPU works better.

 

[juanrga]

You are attempting to rewrite history with your comments! It's not going to work!

What rewrite? As stated it was a problem with some mobos and with the PSU he used:

In the video, der8auer elaborates to basically claim a completely lack of consistency among the quality of VRMs and their heatsinks in various manufacturers. In his first test, he takes a CPU that is known to do 5.0 GHz and on a Gigabyte Aorus branded mainboard found himself unable to even hit 4.6 GHz with dangerously high VRM temperatures. He goes on to blame the heatsinks on the VRMs, going so far to call the Gigabyte solution more of a "heat insulation" device than a cooler, as a simple small fan over the bare VRM array did many magnitudes better than a simple standard install with the stock VRM cooler attached. After an MSI-branded board did similar, it became clear this was not an isolated issue.

der8auer also went on to criticize the lack of voltage input in the form of many boards having only a "single 8-pin connector" which der8auer claims is not nearly enough. He claims a cable temperature of nearly 65 degrees Celsius on the 8-pin EPS cable which is obviously disconcerting, though TechPowerUp has been in discussions with renowned PSU-tester Jon Gerow (Jonnyguru) who feels the "all-in-one" cable design on the Super Flower PSU shown in the video may be partially to blame here for the heat level with that current draw. It's hard to tell which part is more at fault for that temperature and we will update that as we know more. Until then, here's is Jon Gerow's direct comment on the matter:

If you used the SuperFlower PSU in the video with the crystal connectors, that's part of your problem. Those "universal 9-pin connectors" have less conductors than most other modular PSUs because the same connector that's used for EPS12V, PCIe, etc. has to also support +5V and +3.3V for Molex and SATA and then there's an "LED pin" which, when grounded to a ground pin, turns on the interface's LED. A horribly bad design. This is why the wires would be so hot. I suggest checking the voltage at the PSU and then at the motherboard's EPS12V to see what the drop looks like under load. If the voltage is significantly lower than +12V, the board is going to have to pull more current than it normally would. I then suggest using that AX1500i you have on the shelf behind you and see if you end up with the same results since that modular cable for the EPS12V is four +12V pins and four grounds. -- jonny

There were also problems with some launch BIOS. All those problems are gone today.

 

[aldaia]

But I have to thank you for mentioning the V-Ray 'official' benchmark. As you know they give a CPU list and a GPU list. The best value obtained by the TR 1950x is 00:40.608. The best value obtained by GPUs are in the 00:14.359 range. I.e. GPUs are ~2x faster. The TR chips shine on workloads that are better run on GPUs.

As I tried to explain more than once, the CPU must at least shine on latency workloads and the GPUs must at least shine on throughout workloads. AMD Zen design is oriented to throughput. Zen lags in latency-sensitive workloads, but works fine in workloads where a GPU works better.

You are welcome. But it's me who should thank you.
The GPU time you quote required 4x Nvidia Quadro GP100 @ ~$8000 each. Comparing a $1000 CPU to $32000 worth of GPU's it's not fair at all.

Not to mention the difference in power draw and the fact that you still need a CPU anyway.

 

[juanrga]

But I have to thank you for mentioning the V-Ray 'official' benchmark. As you know they give a CPU list and a GPU list. The best value obtained by the TR 1950x is 00:40.608. The best value obtained by GPUs are in the 00:14.359 range. I.e. GPUs are ~2x faster. The TR chips shine on workloads that are better run on GPUs.

As I tried to explain more than once, the CPU must at least shine on latency workloads and the GPUs must at least shine on throughout workloads. AMD Zen design is oriented to throughput. Zen lags in latency-sensitive workloads, but works fine in workloads where a GPU works better.

You are welcome. But it's me who should thank you.
The GPU time you quote required 4x Nvidia Quadro GP100 @ ~$8000 each. Comparing a $1000 CPU to $32000 worth of GPU's it's not fair at all.

Not to mention the difference in power draw and the fact that you still need a CPU anyway.

My point about that CPUs would target CPU-like workloads, whereas GPUs would target GPU-like workloads. The fact that TR is being advertised for rendering tasks says it all.

Acquisition costs without considering total costs, including hour-work salaries, means nothing.

It is evident that 'official' benchmark is not using all the performance that bring the four Quadro cards, because a single 1080Ti does 16 seconds. Therefore pretending that one needs $32000 in GPUs to do the work of a $1000 is so ridiculous as pretending that 1080Ti is about so fast as four Quadro

Yes, one has to acquire a CPU with the GPU, and whereas the GPU does the rendering the CPU can do other work. This is the basis of the HSA approach by AMD, which recently has been forgotten by almost everyone.

 

[aldaia]

[
It is evident that 'official' benchmark is not using all the performance that bring the four Quadro cards, because a single 1080Ti does 16 seconds. Therefore pretending that one needs $32000 in GPUs to do the work of a $1000 is so ridiculous as pretending that 1080Ti is about so fast as four Quadro

You are wrong again, a single 1080 Ti completes V-ray in about 1 minute or more. To complete V-ray in 16 seconds you need 7x or 8x 1080Ti. You where clearly mislead by a typo.

1 00:16.091 Intel(R) Core(TM) i7-6900K CPU @ 3.20GHz x16, GeForce GTX 1080 Ti 11264MB x7
2 00:16.305 Intel(R) Core(TM) i7-6850K CPU @ 3.60GHz x12, GeForce GTX 1080 Ti 11264MB (missing x7 or x8)
3 00:16.333 Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz x32, GeForce GTX 1080 Ti 11264MB x8
4 00:16.601 Intel(R) Xeon(R) CPU E5-2630 v4 @ 2.20GHz x40, GeForce GTX 1080 Ti 11264MB x8

Results with single 1080 Ti:
153 00:58.298 Intel(R) Xeon(R) CPU E5-2683 v3 @ 2.00GHz x56, GeForce GTX 1080 Ti 11264MB
155 00:58.706 Intel(R) Core(TM) i7-5960X CPU @ 3.00GHz x16, GeForce GTX 1080 Ti 11264MB
156 00:59.206 AMD Ryzen 7 1700X Eight-Core Processor x16, GeForce GTX 1080 Ti 11264MB
157 01:00.021 AMD Ryzen 7 1700 Eight-Core Processor x16, GeForce GTX 1080 Ti 11264MB
158 01:00.315 Intel(R) Core(TM) i7-7700K CPU @ 4.20GHz x8, GeForce GTX 1080 Ti 11264MB
159 01:01.771 Intel(R) Core(TM) i7-5930K CPU @ 3.50GHz x12, GeForce GTX 1080 Ti 11264MB
160 01:02.236 Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz x8, GeForce GTX 1080 Ti 11264MB
...
209 01:18.122 Intel(R) Core(TM) i7-6950X CPU @ 3.00GHz x20, GeForce GTX 1080 Ti 11264MB
210 01:18.818 Intel(R) Xeon(R) CPU E5-2696 v3 @ 2.30GHz x36, GeForce GTX 1080 Ti 11264MB
211 01:18.984 Intel(R) Core(TM) i7-4770K CPU @ 3.50GHz x8, GeForce GTX 1080 Ti 11264MB
213 01:21.472 Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz x32, GeForce GTX 1080 Ti 11264MB

Have a good weekend.

 

[goldstone77]

Juanrga you claim many bias paid reviews to make Intel's Skylake-X look bad compared to AMD's ThreadRipper, but you are the only 1, and I mean only 1 saying this is happening! You are scouring the internet finding any scrap of contradiction to help your point. No matter how invalid to make a hypothetical win in your own mind! No one else is believes it! At some point rational people re-evaluate what they are saying in the presence of overwhelming evidence to the contrary! No matter how much you wish the sky is not blue it won't change the fact that for the overwhelming majority the sky is blue!

 

[8350rocks]

I have answered his question. I have confirmed that Intel has access to patents "outside the X86 realm" as part of a general cross-licensing agreement with AMD. "Cross-licensing" means that Intel can access AMD patents and AMD can access Intel patents. Intel and Nvidia signed a similar agreement.

Again, I direct you at the word 'blanket' and point at your evasive wording. Blanket means over then entire die, and the answer to that is no. There have been numerous lawsuits between AMD and Intel that show this. From the original MMX naming dispute to modern Quickpath tech being Intel exclusive despite likely being very easy to implement on Radeon IGPs. Modern x86 processors are mostly 'uncore' and AMD and Intel have little oversight into patents from either side in relation to uncore afaik. Sure AMD jut did the GPU deal, but that's unlikely to mean that they could just take what they want from Radeon and add it to Iris. 'Outside the x86 realm' means outside the actual major x86_64 cross licensing agreement AMD made with Intel that allowed them to implement EMT64 or vice versa with MMX etc etc. The answer to that AFAIK is no there is no all encompassing agreement.

Access means nothing: They are patents and can be accessed by anyone including me and you. Can you be more specific?

The reason why AMD CPU are more optimized for throughput than latency, doesn't have anything to do with "philosophy". Designing a latency-optimized muarch is very hard due the nonlinearities involved in the process, and AMD has to push in the opposite direction with Zen, due to lack of resources, as they did with Bulldozer. That is why the MC on RyZen is optimized for BW not latency; that is why Zen core is wider, with smaller IPC than Intel but higher SMT yields; that is why AMD brings moar cores (e.g. 32C vs 28C), and so on.

True, but so obviously subjectively written. Zen has higher yields because yields are done on a per-chip basis and there are 4+ in TR's MCM. 1 bad part on an i9 and it's the bin for you. 1 bad part on a TR and you're only binning 1 quarter of the chip, so binning for a 32c is about as difficult as an 8 core. To counter this a bit, Intel has had much more time and revisions with the node. Look at the RRP of TR4 and tell me that wasn't a great idea to reduce BOM, then realise that this strategy is also why 'moar cores' is better for them too. So you can turn more of them off between SKUs and use more half CCX units etc. TR's per core IPC is within 10% of i9's too.

Intel's 'very hard' 'optimised muarch' answer was only so difficult that it took 4 months for them to pull out the entire lineup from the TR announcement. You have to admit it's either a hack (rushed) or Intel were holding back on us and probably still are (they certainly had plenty of time to introduce their own massively multicore HEDTs before this).

As owner of a SkyX platform that's bad news in either direction.

Stop pretending that AMD is inventing the wheel. All companies have designed or are designing fast interconnects with tons of bandwidth for either homogeneous or heterogeneous chips. I can easily recall now designs from IBM, Nvidia, or Fujitsu.

EPYC/TR is not RyZen. As mentioned a couple times TR has a power throttling mechanism, from the SP3 inheritance, that reduces core clocks under the base clocks, when memory/IF are overclocked. The lack of performance that you mention is due to the cores loosing performance due to this automatic underclocking feature.

I'm not pretending AMD is reinventing the wheel, I am inferring that AMD is listening to it's inventors faster than Intel. i9s come with faster, lower latency core 2 core interconnects and a faster core itself, but lower lane counts of PCI and little in external (homo or hetero) new pipework otherwise UPI notwithstanding. This was it's response to AMD's IF and before that they had a ring interconnect which was already a fatter, lower latency pipe that simply couldn't scale to the cores needed because having the fastest pipe is not always having the lowest latency. The laser focus on latency lost to parallelism and throughput. Hold that thought.

AMD made a universally instructible SDN mesh interconnect that replaces Q/UPI/HyperTransport as well as core to core inter That's repurposing the fat multi socket pipe going to waste in i9 (but used for IGP in i7 downward) for HBM and GPGPU via NUMA. In addition, every CCX core is directly connected to the IMC via L2 then connected to IF, and IF is connected to the RAM and bridges dozens of times as a result... outside the cores. This actually gives the actual TR chip the capability to take and output far far more data without saturation than the i9, even if core 2 core isn't as great outside each MCM package... aka only relevant on processes that have over 16 threads worth of CPU demands, yet small enough to fit in L2+L3 without fetching from RAM. TR doesn't need L3 because it can access the RAM as fast many times and without having to copy to L3 first.

In short, isolate the effect of this and it looks like this: I might lose in a 20 core CPU-only render race with you, but if we both stack up on non-HBM GPUs then run the same render I will see a greater improvement than you due to PCI-E lanes. If we both stack up on HBM GPUs I will see a massive improvement due to deduplication of VRAM data and the fact that there is now a NUMA bus that can address that VRAM at super low latency compared to previous/current tech.

Latency is a key concept here. As stated in the AMD HSA specification, CPUs are a special kind of LCU: Latency Compute Unit.

Latency is the reason why AMD designed a new wider core, Zen, instead [of, I assume] reusing the narrow Jaguar cores for instance.

Latency is not IPC. Jaguar was technologically EOL long ago and was replaced because it was not competitive; there were dozens of reasons. You talk as if wider cores are not a throughput-prioritising decision: they are. In terms of die space, reducing the physical size of a wide pipe to fit it's predecessor (or making space for it elsewhere) and making a narrow pipe run faster are both very time consuming, as is parallelising with the intent to multiply throughput without regard to latency. I said before, latency improves throughput but not always the reverse so this is a non argument

Latency is the reason why L2 cache in Zen is private instead shared.

L2 is private in Zen because it's directly connected to the IMC. Fat L2 has always been the philosophy behind AMD's core cache distribution, whereas Intel liked faster, larger L3 than AMD and weaker L2. Intel switched to AMD's point of view just now because it makes sense when your ring is now a grid with an exponential amount of connections to bits of your SoC and your L3 needs access to all of them.

Latency is the reason why AMD targets higher clocks with turbo mode. It is also the reason why EPYC with higher-clocked RAM runs better due to reduction of both memory and IF latencies.

Latency is the reason why AMD locked the IF clock to the RAM clock. Otherwise, engineers would introduced intermediate buffers between clock domains and those buffers would increase latency.

Failing to acknowledge that this is only because AMD has a much much higher throughput IF solution consisting of multiple IMCs. Latency took a backseat compared to Intel's single IMC decision out of the core. Again, and I stress that I said this before, latency is important but not all-important. Certain bits are limited by other means. More specifically, having one part that is granularly of lower latency than another does not necessarily make for a lower latency total system, but having a high throughput part will ensure that it is *possible* to reach low latencies by virtue of optimal conditions. Often the latter will lead you to much lower latencies than the former, else we would not have multicore computers in the first place. The whole framework for threading and out-of order execution etc etc is something that would add significant internal latency to any computer compared to a single in-order arch design, but nothing compared to the advantages presented by the extra cores.

You mention "prediction and caches are the future, and all they do is improve throughput". This is wrong at two levels. First, they aren't the future, they are the present. Second they improve latency so much as throughput. Caches have much lower access times than main memory, which reduces latency when accessing data. Prediction algorithms ensure with some given probability that correct data/instructions are ready to enter the pipeline at the correct instant. When this doesn't happen, i.e. when the prediction fails, the cost is paid in the extra cycles needed to search the correct data/instructions, move the information to the pipeline and flush the incorrect data/istructions from the pipeline.

If you mean they are the present within present releases, sure they are, but that doesn't mean that the major performance benefits of parts in the next 5 years won't be down to improvements within prediction and caches.

Same happens with consoles. Latency is king. Contrary to a widespread myth GDDR5 doesn't have significantly worse latencies than DDR3. Moreover, the PS4 has cores clocked low (1.6GHz), which helps to reduces the latency gap between execution and memory pool.
Adding aGDDR5 memory subsystem optimized for throughput doesn't affect the same to jaguar cores @1.6GHz than to Skylake cores @4.5GHz. However, even with low core clocks, several game developers have had latency problems with the PS4. Precisely the first thing that Sony has made on the new Pro has been to increase the clocks on cores and memory, reducing latencies by up to 30%...

This all is well-known

Point number one is answered by point number two. PS4 GDDR CAS timing is in the high double to low triple digits, and I do not know the 'popular belief' stories but I know my and probably your GPU has much faster GDDR at much worse clock latencies. I am aware of GDDR's intrinsic strengths and that $ for $ you will find lower timings as a result of duplex design, but even with a 30% increase they lag severely behind similar clocked DDR3/4. I am also aware that the Xbox has DDR3 ram and a significant throughput deficit, yet a latency advantage further exacerbated by super low latency ESram... and has uglier games as a result and far more performance complaints. The example I gave was commentary on the example I gave, not the GDDR technology, but if you want commentary there, I will say that the large batch memory transfer optimisation is a core reason why it works better in GPUs but when L2+3 caches are big enough I'm sure it will be the same story for CPUs. Both AMD and Intel are moving toward parallelism and throughput over latency below L2 and I believe AMD is in the lead there. Everything below L2 is just a cache.

BIOS patches and microcode will remove the limitations on clocking IF I expect. Let's remember that TR shares more with EPYC than Ryzen and if Ryzen is anything to look at AMD is actually working on optimising IF on a daily basis... and still were late to market with 3200RAM support because of IF.

Those disputes between AMD and Intel were solved when both companies signed an agreement that included cross-licensing of patents.

When I mean "access to patents", evidently I mean that AMD can use technology based in Intel patents and vice verse, without being sued for patent violation. That is what the cross-licensing agreement granted to both companies: access to other's technology . I will repeat once again that the agreement is not limited to x86, but also includes other stuff, including graphics. Sure that you did heard rumors about a hybrid chip with Intel CPU and Radeon GPU. Guess what? Three different sources confirmed me that the prototype is real.

I have discussed "SMT yields"; this is a parameter of the microarchitecture. And you are confounding it with production yields, which is a parameter of the process node. I have discussed SMT yields as part of my explanation of why all the key elements in Zen point to a design optimized for throughput.

A minor note. ThreadRipper doesn't have four dies. There is only two.

Who has said you that the change from ring to mesh is a response to AMD? Intel has been using mesh in the Phi line, and we know that both ordinary Xeon line and Phi line are converging. We have known for a looong time that SKL Xeons would get AVX-512 support. I have known it since 2014 and it is evident that the SKL Xeon design was planned many years ago even before "Zen" was a public codename for a future muarch.

"This actually gives the actual TR chip the capability to take and output far far more data without saturation than the i9". Benchmarks tell the opposite

Yes, latency is not IPC. Never said both are the same, but IPC is a function of latency. Moreover CPUs are just LCUs: i.e. Latency Compute Units. That is the reason why latency is a key parameter for CPUs.

L2 is private in Zen because a shared L2 would increase latency and kill the IPC of the core. It is the same reason why the L2 is private in IBM and Intel high performance designs. The L2 being private doesn't have anything to do with the CCX approach neither with the IMC.

AMD has always used larger L2 and L1 because of latency issues with their designs. It doesn't have anything to do with the interconnect. IBM has been using larger L2 by similar reasons.

Intel changed now to a larger L2 and a non-inclusive L3 with SKL-SP. The non-inclusive L3 behaves similarly to the L4 in some Broadwell models. And the new L2 has a size of 1MB. Guess what other chip from Intel has 1MB L2 cache? The KNL Phi. It seems evident the change in cache hierarchy is motivated by the inclusion of AVX512 support and the convergence with the Phi line. Evidently those fundamental changes in the microarchitecture were made many many many years ago: before 2014.

GDDR5 has nearly identical latencies than DDR3: tRC = 40--48ns for both; tCCD = 2ns for GDDR5, 4ns for DDR3...

The reason why measuring latencies in GPUs gives significantly worse results is due to the memory controller and the cache structure. GPUs as TCUs (Throughput Compute Units) and the memory controller is optimized for throughput, not latency.

Do not confound the idea that x86 and x86-64 are cross licensed with the idea that AMD or Intel could use patents for various core technologies or memory technologies from their competitor.

The only thing included in the cross licensing deal is the ISA. Nothing else.

 

[goldstone77]

Spoiler

Spoiler

i9 consuming less power than the TR 1950X. Even OC to 4.6GHz it is consuming less power than 3.9GHz TR. Still the i9 is accused of consuming huge amounts of power and putting houses on fire. LOL

At stock the 1950X with MCM package and 6 more cores at stock consumes 20 watts more! Overclocked 45 watts more! Bwahahahaha! This shows the 7900X is getting stomped and you think it's a win? You have to take off those beer googles you are wearing! They are distorting your ability to use logic and reason!

Nothing of that invalidates the fact that people that complained because the i9 was consuming ~400W, is now silent when TR consumes more power. The double standard is evident.

Spoiler

Not all cores are the same. AMD uses simpler narrow cores that consume less power each. Not a mystery here, simply a consequence of the laws of physics. That is the reason why Atoms cores are simpler than Xeon cores.

Not all process nodes are the same. 14LPP prioritizes efficiency over performance. That is why the 1950X hits a wall at 3.9GHz, whereas the i9 can hit above 4.5GHz.

Not all overclocks are the same. A chip doesn't consume the same power at 3.9GHz than at 4.6GHz. Note that increasing clocks from 3.4GHz to 3.9GHz in the 1950X has increased the power consumption by more than 100W. Increasing clocks from 3.3GHz to 4.6GHz in the 7900x has increased the power consumption only by 76W.

Pushing the 1950X to 4.6GHz would increase power consumption to about 700W. The i9 does ~500W at same clocks.

You are just trying to be an Intel apologist! The fact is that for 20 more watts you get 6 more cores with ThreadRipper! That's an amazing performance metric in itself! It's doesn't take a genius too see you're just griefing over Intel!

I explained in the spoiler the flaws with that statement.

Your statements didn't change anything.

The thoughts hidden in the spoiler explain why statements like your "20 more watts you get 6 more cores" are misguided. I reproduce the thoughts here:

Not all cores are the same. AMD uses simpler narrow cores that consume less power each. Not a mystery here, simply a consequence of the laws of physics. That is the reason why Atoms cores are simpler than Xeon cores.

Not all process nodes are the same. 14LPP prioritizes efficiency over performance. That is why the 1950X hits a wall at 3.9GHz, whereas the i9 can hit above 4.5GHz.

Not all overclocks are the same. A chip doesn't consume the same power at 3.9GHz than at 4.6GHz. Note that increasing clocks from 3.4GHz to 3.9GHz in the 1950X has increased the power consumption by more than 100W. Increasing clocks from 3.3GHz to 4.6GHz in the 7900x has increased the power consumption only by 76W.

Pushing the 1950X to 4.6GHz would increase power consumption to about 700W. The i9 does ~500W at same clocks.

Of course, nothing of that change a bit the main fact that the people accusing the i9 of consuming huge amounts of power remained silent when TR consumed still more power.

Not all cores are the same. AMD uses simpler narrow cores that consume less power each. Not a mystery here, simply a consequence of the laws of physics.

True, and this only emphasizes my point about ThreadRipper's superior performance vs. power consumption compared to the 7900X based on design for professional content creators!

Not all process nodes are the same. 14LPP prioritizes efficiency over performance. That is why the 1950X hits a wall at 3.9GHz, whereas the i9 can hit above 4.5GHz.

You admit the process that ThreadRipper is built on offer's superior performance vs. power consumption compared to the 7900X for professional content creators!

Pushing the 1950X to 4.6GHz would increase power consumption to about 700W. The i9 does ~500W at same clocks.

This is a hypothetical statement that does not exist, which is the basis for your argument! I refer back to my statement What you said didn't change anything! You are just arguing to argue, and offer proof that you already know the truth! You are so bias even when you present the truth to yourself you can not find it!

Confirmation bias is the tendency people have to favor facts or arguments that confirm the beliefs and positions they already hold. The extreme form of this bias is referred to as “belief perseverance” when people hold onto their beliefs even after they've been proven false.

 

[goldstone77]

You are attempting to rewrite history with your comments! It's not going to work!

What rewrite? As stated it was a problem with some mobos and with the PSU he used:

Spoiler

In the video, der8auer elaborates to basically claim a completely lack of consistency among the quality of VRMs and their heatsinks in various manufacturers. In his first test, he takes a CPU that is known to do 5.0 GHz and on a Gigabyte Aorus branded mainboard found himself unable to even hit 4.6 GHz with dangerously high VRM temperatures. He goes on to blame the heatsinks on the VRMs, going so far to call the Gigabyte solution more of a "heat insulation" device than a cooler, as a simple small fan over the bare VRM array did many magnitudes better than a simple standard install with the stock VRM cooler attached. After an MSI-branded board did similar, it became clear this was not an isolated issue.

der8auer also went on to criticize the lack of voltage input in the form of many boards having only a "single 8-pin connector" which der8auer claims is not nearly enough. He claims a cable temperature of nearly 65 degrees Celsius on the 8-pin EPS cable which is obviously disconcerting, though TechPowerUp has been in discussions with renowned PSU-tester Jon Gerow (Jonnyguru) who feels the "all-in-one" cable design on the Super Flower PSU shown in the video may be partially to blame here for the heat level with that current draw. It's hard to tell which part is more at fault for that temperature and we will update that as we know more. Until then, here's is Jon Gerow's direct comment on the matter:

If you used the SuperFlower PSU in the video with the crystal connectors, that's part of your problem. Those "universal 9-pin connectors" have less conductors than most other modular PSUs because the same connector that's used for EPS12V, PCIe, etc. has to also support +5V and +3.3V for Molex and SATA and then there's an "LED pin" which, when grounded to a ground pin, turns on the interface's LED. A horribly bad design. This is why the wires would be so hot. I suggest checking the voltage at the PSU and then at the motherboard's EPS12V to see what the drop looks like under load. If the voltage is significantly lower than +12V, the board is going to have to pull more current than it normally would. I then suggest using that AX1500i you have on the shelf behind you and see if you end up with the same results since that modular cable for the EPS12V is four +12V pins and four grounds. -- jonny

There were also problems with some launch BIOS. All those problems are gone today.

You are deflecting from history by trying to focus on one part of it that added to the embarrassment! The History is that the entire Skylake-X product line made no sense, and was an obvious cash grab by Intel! All popular YouTube reviews expressed this! You use a "German"(In your mind a pro AMD country) overclocking enthusiast statements to try to change the narrative about Skylake-X! The numbers speak for themselves! The difference in views, likes, and dislikes between the most favorable YouTube commentary about Skylake-X vs. the most detrimental YouTube commentary! Linus 2,487,668 views 117K likes 1k dislikes vs. der8auer 39,781 views likes 651 dislikes 116. This is the history that you are dismissing or failing to recognize! And it's amusing to me that the reference that you use is a German in a country where you think is predominantly pro AMD, because for the first time in 15 year AMD out sold Intel! Bwahahaha! The ridiculousness of your contradictory statements continues to be a great source of amusement for me! I suggest you watch the Linus video!

Linus
I have been conspicuously silent. About Intel’s new high end desktop X299 platform, and the CPU’s to go along with it. And this was on purpose, but probably not for the reasons that you think. I’ve spent the last few days walking the show floor, chatting with industry folks, and generally trying to wrap my head around what the actual hell it is Intel is trying to do with this launch.

[video="https://www.youtube.com/watch?v=TWFzWRoVNnE&feature=youtu.be&ab_channel=LinusTechTips"][/video]
I have some things to say - Core i9 & X299
2,487,668 views 117K likes 1K dislikes

der8auer
https://www.youtube.com/watch?v=kpoies2JcmI&feature=youtu.be&ab_channel=der8auer
39,781 views likes 651 dislikes 116

 

[juanrga]

Juanrga you claim many bias paid reviews to make Intel's Skylake-X look bad compared to AMD's ThreadRipper

My claims was "many biased reviews" I didn't wrote paid. And I demonstrated the source of the bias affecting SKL-X:

- Using engineering samples instead retails chips.
- Using mobos incompatible with the chip and burning the chip.
- Using launch BIOS with problems and not retesting after with improved BIOS.
- Cherry picking benchmarks to warrant that the chips don't win any benchmark.
- Measuring power on 512bit workload, but then measuring performance in 128bit workload, ruining the efficiency.
- Crippling the CPU performance with GPU-bound and frame-limiting settings.
- Comparing SKL-X chip on stock settings to overclocked RyZen and TR chips, but labeling the graphs all chips as stock.
- Inflating prices of SKL-X chips on performance/price graphs.

And that is accompanied by bad press and fake news. For instance the fake news about AMD getting 10% increase in marketshare, or the articles about the SKylake bug that not only forgot to mention that the bug doesn't affect to the SKL-X line, but the same media didn't publish anything about thr RyZen bug.

As der8auer wrote after reviewing the i9-7900X: "so what's all this negative press about? I don't really understand it."