AMD's Future Chips & SoC's: News, Info & Rumours.

Page 70 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

goldstone77

Distinguished
Aug 22, 2012
2,245
14
19,965


No, the point is that the 2700X is actually cheaper total cost of ownership for anyone who doesn't have spare 8700K CPU cooler buying from the store you posted in your country. But here in America and I assume most other countries MSRP on the 8700K is still higher and doesn't come with a CPU cooler!
 


That is a strawman though. If you go by MSRP alone, Intel IS more expensive. There's no room for argument there. If you can find stores with nice discounts, then good for you, but you can't use that for extrapolation of a neutral point. Like you say, it applies to you and you alone. I'm not going to spend hundreds of pounds, dollars or euros to go to that store and get a couple euros worth of discount. <removed by moderator>

The value proposition of the 2700X is really good for new builders that intend to run their CPUs stock. For better or for worse, Intel forces you to buy a cooling solution with the K, but bundles one with the non-Ks. The K's have higher clock speeds (nominal), so there's people buying them with the intention to run them stock, so you're also leaving that crowd out. You could make a nice table with the different scenarios and apply a generic value statement for each, but I do think that AMD has done a superb job with the positioning of the 2700X and learning from the 1800X launch.

Cheers!

EDIT: Typos.

<Moderator warning: Keep the personal attacks out of it or there will be ejections>
 

jdwii

Splendid


Agreed that is actually my favorite chip from Intel at the moment for gaming it's really all you need even more so budget gamers.

For everyone else the 2700X really is close to the 8700K i mean enthuisists will probably pick Intel since the 8700K can be OC higher and yes its still a little bit better in games and lets not kid ourselfs these are enthusiasts CPU's the power consumption being higher isn't the end of the world. Most of us have overkill power supply's and very good cooling as said before yes it sucks but at the same time its not like its bulldozer. :D
 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790


Yes, it is exactly the same core than Zen. He also proved the memory controller is exactly the same in Raven Ridge. The 3% IPC in Cinebench (<2% IPC average) is a result of higher clocks and some BIOS/AGESA tricks on X470 mobos.
 

Gon Freecss

Reputable
Apr 28, 2015
448
0
4,810
Wait, doesn't the <2% IPC increase come from cache and memory latency improvements? That's why it's >7% faster in WinRAR or something clock for clock.

Also, I don't know why people rave about Zen+ bringing IPC improvements when it's barely an improvement. Skylake-SP/X brings a bigger IPC improvement. So much for my "AMD does more than Intel IPC wise in one year than two years".
 


Sigh... Keep in mind Zen v1.5 is just that: "1.5". They only tweaked the process a bit and memory related gremlins. A 3% """IPC""" increase is really good in fact without touching the underlying uArch nor changing transistors for the CPU logic itself. They managed to squeeze a lot in very little time.

On the Intel side, they added 2 full cores to Kaby Lake and called it Covfefe Lake. You can call it plan B, a rush-job or whatever, but it also had non-IPC benefits everyone is happy about. At the end of the day, IPC is not all in a CPU. Looks like some people is slowly understanding it after so many years.

And as a final note... It would appear GloFo managed to keep the manufacturing costs down for 12nm or something along those lines, because the price drop was really good. I know AMD would have not been able to get away with pricing it at 500 like they did with the 1800X, but they didn't price it marginally higher either using the bundled coolers as excuse. There's something interesting to read between the lines there.

Cheers!
 

Gon Freecss

Reputable
Apr 28, 2015
448
0
4,810
It's shown to be <2% on average though. That's lower than the IPC increases Skylake-X brought. Why are people claiming Intel did not increase IPC in the two year span since Skylake?

They're not charging $500 because they used that to target the 8 core HEDT at the time, the 6900K. In applications, they're more or less on par with the 8700K, that's why they priced it around that CPU.
 


Because Skylake-X is HEDT? The predecessor was BW-E, right?

And the 6900K was 1K, but then the 7820X happened at 600. Even if the overall MoBo+CPU+RAM+Cooling was more expensive, I'd say AMD played it safe with the 1800X launch, but still got as close as possible. I'd imagine if Intel wouldn't have launched the i7 8700K, AMD would have cannibalized the i7 7700K and whatever 4-core successor to it and Intel was able to predict that far at least. I don't think they predicted losing all the low bracket of the HEDT platform though...

Cheers!

EDIT: Typo.
 

Gon Freecss

Reputable
Apr 28, 2015
448
0
4,810
Skylake-X brought IPC improvements over Skylake. That is an IPC increase Intel has done recently.

Don't think they predicted they'd have to release an 8 core, but they did plan on releasing a 6 core even before Ryzen was released (and they did in fact release it over 6 months after Ryzen came, pushed Coffee Lake early).
 


But they are completely different uArch families... You're technically correct, but that comparison is weird because I'm pretty sure there are 2 very different divisions working for each and follow different design ideas for each. I guess there is a gray area for that, but I'm not sure how to consider them as "related".

And Intel had to be prepared for whatever Zen was going to be; at least there's agreement on that. AMD would try to do the same, I guess.

Cheers!
 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790


Yes, the IPC gain are a consequence of cache and memory latency improvements, which are a consequence of higher clocks and BIOS/AGESA tricks.
 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790


There is no Zen v1.5. It is the same core than at launch.

That marketing 3% IPC gain is when comparing 2993MHz Zen+ vs 2666MHz Zen, as Anandtech did. Several sites tested both chips with same memory and the overall IPC change was nearly zero, with some positive gains in some benches being compensated by a loss of IPC in others.

Also CoffeeLake is not only Kabylake with two extra cores. There are other changes including a faster memory controller that increases the IPC a bit in memory bound workloads.



So since everything what Glofo promised about 12LP turned to be a huge piece of propaganda, now we invent the 'theory' that Glofo focused on making the node cheaper right?
 


Looks like you haven't read many reviews about KL and CL then. Go and read about their performance at the same speed. I have absolutely no idea where you get your information from, but it is certainly wrong.

Intel and AMD benefit from faster RAM anyway. You've said so yourself many times. Why can't improving the IMC capabilities be considered part of the CPU design improvements? Are you being term-pedantic here or just dismissing the improvements that Intel also touts as "generational leaps"? Also, what was the performance increase at the same clocks between Kaby Lake and Skylake again? 1%-2%?



Yeah. AMD doesn't sell under-cost like Intel (or at least, it doesn't seem to), so it has to be GloFo making the manufacturing cheaper, right? :)

EDIT: https://twitter.com/IanCutress/status/987458764470210560

It seems that the results are verifiable and consistent in AT's case. Maybe Asus hit some sort of magic sauce with their latest BIOS revision?

Cheers!
 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790


Total IPC is a function of IPC(core), IPC(LLC), and IPC(main memory). Faster memory means the core has faster access to information, which means the core has to wait less cycles to process the information, which means IPC increases. To measure it, one has to use memory-bound workloads. Most reviews didn't and they used beta and early BIOS versions, which masked the results.



I just mentioned how faster IMC was part of CoffeeLake improvements. So?



Maybe first RyZen version was overpriced and now AMD is reducing margins. This speculation sounds much more pausible than the speculation "Glofo failed on everything that said about 12LP because focused on making a cheap node".



That is his way of saying he doesn't care about his flawed results. Why would he care? He will get his paycheck anyways :sarcastic:

0YvNJMW.jpg

 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790


Ok, I did read more reviews and AT lose his ranking as worst biased review. TechRadar got the gold medal with claims that 2700x beats 8700k in any single-core benchmark. Basically TechRadard pretends that Zen+ IPC about 15% higher than Intel cores. An example

7WCtXMZPBaFWBmd4zPuGCG-650-80.jpg


Not even AT has published such nonsensical CB15 scores.
 

jdwii

Splendid


Pretty sure Pinnacle Ridge supports higher speed memory making IPC higher. Again the 2700X for 329$ seems competitive.
 


Sigh...



More sigh...



Yeah, you don't seem to be reading between the lines. You're just taking the information at face value.



That is a very unfair interpretation of Ian's work and effort. If you have a beef against the numbers, try and demonstrate why they're incorrect. I mean, you should be really familiar dealing with "unpopular opinions", right? Sympathize a little bit here.

EDIT: Fixed quote.
 

Gon Freecss

Reputable
Apr 28, 2015
448
0
4,810
It's not unfair or whatever. He screwed up. He should own up to his mistakes.

Also, AMD does indeed violate its TDP ratings. The 2700X dissipates 145W at stock, when it's marketed to dissipate 105W. And AMD says it needs a cooler capable of dissipating 105W for it to be at its optimal temperature of 61.8 degrees C. Guess what? With the stock cooler (Wraith Prism), which is rated for a lot higher than 105W, it doesn't keep it below 62, at stock mind you.

Temps1.png


See the dishonesty here? But no, you gotta appeal to authority!
 

Gon Freecss

Reputable
Apr 28, 2015
448
0
4,810
Okay, but AMD states you need a cooler that dissipates 105W for the temperature to be 61.8 degrees C. As noted above, that's not the case whatsoever, even with a cooler that can dissipate much more than 105W of heat.
 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790


We already know that AMD lied again about TDPs. From the Stilt review:

Personally, I think that AMD should have rated these CPUs for 140W TDP instead of the 105W rating they ended up with.

Since 105W TDP rated Pinnacle Ridge CPUs are allowed to sustain >= 141.75W of power draw, and more importantly because at stock they do consume significantly more than the rated 105W even in real world multithreaded workloads, their advertised power rating in my opinion is not entirely fair and might end up misleading the consumers. The measured sustained power consumption for a stock 2700X was 127.64W (132W peak) during X264 encoding and 142.52W (146.5W peak) during Prime95 28.10. In comparison, a stock i9-7960X CPU with its power limit reduced from the default 165 / 206W to 140 / 175W (PL1, PL2) sustained 139.82W power draw and had a peak draw of 168W in the very same X264 workload.

The first ones to mention that 105W wasn't the real TDP were the guys at Canard, when published the first review of 2700X some time ago. The Canard review got 140W for the 2700X and 103W for the 2600X.

 
I won't keep on debating what or what isn't TDP, but you guys really really need to understand a bit more about "typical workloads". You, Juan, have called some benchmarks "power bugs" when they just execute specific code that really stresses the CPUs in a linear way. What is your criteria for that? What is the criteria for AMD? For Intel? Why not call every single "stress" test "power bugs"? I haven't seen a single game benchmark in all reviews that show AMD pulling that much power at any given time nor average. Have some examples?

In regards to the cooling solution. Yeah, AMD bundles cooling solutions now with all the new lineup, but for stock operation it is sufficient as demonstrated by reviews. What's the problem there? Have you seen spec sheets stating the CPUs cannot go beyond 60°C? 80°C? 100°C?

I mean, you guys are trying so hard to dismiss the 2700X improvements and I don't even understand why... Like, seriously, WHY?

EDIT: In regards to IPC:
Ryzen-2000%20IPC_575px.png
 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790
Reviewers know what is the definition of TDP and how it is measured. Reviews found that advertised TDPs aren't the real TDPs. Professional overclockers as The Stilt also claim that real TDP of the 2700X is 140W. We can spend another year negating this fact, just as some negated the real TDPs of 1000-series aren't the advertised,...

So giving real facts for 2700X is dismissing improvements? Mentioning that Anandtech and TechRadard reviews are pure nonsense is dismissing improvements? ROFL

AMD claims "~3%" IPC gains for 2000-series in CB. Check the tomshardware review, check any other good review

aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9FL1AvNzY1MTY5L29yaWdpbmFsL2ltYWdlMDA4LnBuZw==

15241006322z6uo19egw_4_4.png


Both show 2700X has 2--3% higher IPC than 1800X, which falls within the expected improvement for Pinnacle Ridge.

But Anandtech got 22% IPC increase in CB. A nonsensical 22%!!! So mentioning that Anandtech review is a piece of nonsense is now dismissing the 2700X improvements? ROFL
 
Yeah, exactly 1 graph is an outlier and Ian is looking into why that happened (including the games), but the rest falls right in line with all other tests. But hey, why focus on the other 10 graphs, right?

Toms didn't activate the MCE option in their MoBos either (for some reason), so their threaded tests ARE skewed:
But as we often find with MCE, AMD's Precision Boost Overdrive is enabled by default in many BIOSes. After extensive experimentation, we can conclude that the option doesn't deliver an appreciable performance gain in its current form. Thus, we ran our tests with Precision Boost Overdrive disabled.

In regards to that HardOCP actually explain this (ironically) in layman terms pretty nicely:
Precision Boost 2 on the Ryzen 2 is a totally different animal, and as you will see in our benchmarks, it really shows. And the better cooling you have, the better Precision Boost 2 works. PB2 is much more dynamic, not that the original Precision Boost was "dynamic" at all. Basically once you broke a 2 core load with PB, the CPU downclocked to its base speed. PB2 waves bye-bye to that and this is a testament to the new 12nm process.

Why aren't you talking about that?

EDIT: Typo.
EDIT2: Although, in all fairness, it should not matter for locked speeds? I would imagine? Mr der8auer mentioned he used a different approach to OC with the 2700X, so it could be related to how you lock the speeds? There's a lot of questions that I still have, to be honest.