Discussion: AMD Ryzen

[8350rocks]

Maybe a 3.8-4.2ghz with clise to 6900 consumption. That would be a thing a beauty

From what I gather, a 125W TDP part would have a bclk at 4.0 GHz and turbo of 4.4 based on F3 stepping.

 

[os2wiz]

With 3.6 GHZ Ryzen with its 4.0 turbo on it

Are those clocks confirmed? Last info I have is that the 3.6/4.0GHz Ryzen sample wasn't a 95W chip, it exceed the 100W mark.

Show me that information that you have that the Ryzen ES at CES was using above 100 watts. I have read nothing to substantiate what you are saying. Where is your source?????

 

[os2wiz]

The way i see it is the product of IPC is throughput and Canard PC showed enough of that. And the 1188 leak which was unanimously accept by many including Juan who quickly branded it Ivybridge without factoring in clockspeed, i was able to show you all that the product of 1188 can equate to a 5960X which has similar throughput at similar clocks.

On CPC's own review they said a 5960 had the same productivity score at the same clock speed, i then showed a 6600k being marginally better than the 5960X in a number of games yet it is the same percentage as reflected on CPC's benchmarks.

While i wish sarwars theory pans out and it beats Broadwell, i just don't see it, Ryzen could match Broadwell with clockspeed advantage though.

Ivy/sandy performance would be disastrous to market,. 5 year old and 6 generations behind type performance on 14nm and almost 4x the on die resources would essentially lock AMD into a slow death spiral, it is something Juan wants judging by his comments on other sites. Those that say sandy/ivy is good enough clearly haven't vseen how archiac the performance is. I replaced a 2400 with a 4460 in a SFF theatre pc,. It was around 30 seconds faster than the 2400 in the blender demo. Those with 2500/2600's are holding off because a similar level system on a new architecture is about $300 more than it was in 2011, so it is more an affordability thing, so the maximum is overclock,. But you need a 4.8ghz sandy to sort of equal a 6700/6600 at the price of electrical and heat damage notably to the power supply and motherboard.

So essentially you are saying Ryzen IPC is at Haswell levels and with its higher clock speed it equals Broadwell. That makes sense to me. With 3.6 GHZ Ryzen with its 4.0 turbo on it should slightly exceed the I7 6900k in most tests and real world activity.

That is what I tried conveying yesterday using all the numbers I had to work off and they seem to add up. Not claiming it is right but I did feel that Haswell was more or less the higher floor. It is also a good performance base as the distance from Haswell to Kaby lake is not that far.

I think your postulation seems to be correct. Haswell is not bad at all and with the 3.6 GHZ mark for base frequency it is a really relevant cpu for todays needs both work and play. To me the turbo being 4.0 or 3.9 is irrelevant and insignificant. If 3.9 GHZ brings it no more than 95 watts, great, then that should be its release. Now if AMD decides to have a binned 3.8 GHZ chip with 4.1 or 4.2 turbo with 120 or 125 watts TDP that would be nice for some, but not me. I prefer to overclock with my water loop. I am not seeking 4.5 GHZ, 4.1 or 4.2 base clock with no turbo would be fine for me.
By the way I have not seen an official explanation of turbo under Ryzen some have speculated here it is different than on previous AMD chips, it essentially stays at turbo speed as long as the cooling is adequate,though I have heard no confirmation of this from AMD. I assume that their OC technology does allow the user to attempt an overclock independent of the automatic OC. But I may be wrong on that.

 

[juanrga]

f3 and firmware may need tweaks to iron out wasted power before the QS due in February

With 3.6 GHZ Ryzen with its 4.0 turbo on it

Are those clocks confirmed? Last info I have is that the 3.6/4.0GHz Ryzen sample wasn't a 95W chip, it exceed the 100W mark.

F3 stepping has 3.6 bclk, turbo will be 3.9 or 4.0.

The chip I have mentioned is a F4.

 

[juanrga]

Maybe a 3.8-4.2ghz with clise to 6900 consumption. That would be a thing a beauty

From what I gather, a 125W TDP part would have a bclk at 4.0 GHz and turbo of 4.4 based on F3 stepping.

I get about 140W based in F4 steeping.

 

[sarinaide]

At 4-4.4 that would be a mighty strong octo core even at 130-140w id buy it as a gaming server.

Get a decent x370 chipset hit it with my 2 256GB M.2 (2280) for some Raid boot drive love then hook up my 4x3TB drives.

Just need to save up for a 1080 or another 470

 

[cdrkf]

At 4-4.4 that would be a mighty strong octo core even at 130-140w id buy it as a gaming server.

Get a decent x370 chipset hit it with my 2 256GB M.2 (2280) for some Raid boot drive love then hook up my 4x3TB drives.

Just need to save up for a 1080 or another 470

If you are waiting for Zen you may as well wait to see what Vega brings to the table (what we've seen so far puts it between the 1080 and Titan XP / 1080ti whenever that gets released).

 

[salgado18]

Maybe a 3.8-4.2ghz with clise to 6900 consumption. That would be a thing a beauty

From what I gather, a 125W TDP part would have a bclk at 4.0 GHz and turbo of 4.4 based on F3 stepping.

I get about 140W based in F4 steeping.

What's the chance that AMD goes nuts again (9590-like) and releases a 140W 4c/8t chip? How fast would that one be?

 

[James Mason]

Maybe a 3.8-4.2ghz with clise to 6900 consumption. That would be a thing a beauty

From what I gather, a 125W TDP part would have a bclk at 4.0 GHz and turbo of 4.4 based on F3 stepping.

I get about 140W based in F4 steeping.

What's the chance that AMD goes nuts again (9590-like) and releases a 140W 4c/8t chip? How fast would that one be?

Let's hope they don't because that series of chips is the biggest POS out on the market. They should have stopped producing and selling those long ago.

 

[salgado18]

What's the chance that AMD goes nuts again (9590-like) and releases a 140W 4c/8t chip? How fast would that one be?

Let's hope they don't because that series of chips is the biggest POS out on the market. They should have stopped producing and selling those long ago.

They are derived from a bad series of chips. Good stuff can't be made with bad materials.

But Zen is looking good, and AMD doesn't seem afraid to push the envelope, like Intel is. The 4c i7s could have very powerful editions, like an i7 7750X, with base 4.4 and turbo 4.8, if they were willing to push TDP up to 120W.

I think AMD could, even with lower IPC, grab some wanted attention, by pushing TDP to the limits on mainstream CPUs. Of course not 220W again, that's stupid, but 140W for enthusiast 4 core CPUs would be awesome.

 

[ComputerSecurityGuy]

Honestly, from a sheer economic perspective, they might do it. Dies that couldn't meet server voltage and power efficiency could retail to the gaming/enthusiast market that cares less about power consumption. I'm not a huge fab expert, but I think such chips would exist.

 

[Saga Lout]

Keep it clean and on track please, folks. No personal attacks and only evidenced facts, not opinion or speculation.

Thanks.

 

[sarinaide]

if you have Haswell + IPC there is no reason to push beyond 130W, in fact I would say that 105-125W would be the maximimu threshold I would like to see it pushed, after that is just getting to much.

 

[Rogue Leader]

Just FYI there is no rule against speculation or opinions specific to Ryzen, thats what this thread is for, discussion of such things. The rule is against attacking people personally for their opinions or speculation (or facts for that matter). As well taking said opinions and speculation down a rabbit hole away from the topic of the thread. Please do not submit alerts because you don't like other peoples opinions, unless said opinion personally attacks or insults you or someone else.

If you do not understand this very simple concept please PM me. If you do not feel you can adhere to it I suggest leaving the conversation. If you do not adhere to said rule we will exit you from the conversation, permanently.

 

[jdwii]

Hi guys i figured i'd share a video from paulshardware about his guesses on prices of Ryzen his is even lower than mine in the 100-200$ market but he does agree with many here about Ryzen being over 600$ for their 8 core part.

https://www.youtube.com/watch?v=nGbC6XLCneU&t=0s

I just think the platform is not as high-end as X99 it’s more between intel's mainstream and X99(I’m not saying it’s not good I actually think it’s more practical for us consumers and gamers). That's why i'm also sticking to my 500$ max price for a 8 core Ryzen with probably a cheaper one offered for 380-400$ would be a nice price to hit Intel's 6 core

Also nicely said Rogue Leader i like reading here on this forum lets stay nice

 

[sarinaide]

I actually think there is enough out there to suggest the initial leak months ago on pricing may actually be correct. As expressed above the cost on a X370, B350, B320/A300, X/A300 is essentially equated to Socket 1150 and not an Intel LGA2011 extreme platform. AMD is looking to leverage unrefuted value on a mainstream platform that will not only take back market share but ensure future client base as performance is good at a much lower price.

Can AMD succeed at this? yes they can the rX480 was a high end GPU sold at $200 and it grossed the highest net return on any GPU, AMD also sold out 100% of it's stocked silicon for the RX480.

SR7 8/16:

Black Addition 3.6-4.0Ghz - $650 shipped with a AMD Wraith cooler.
Standard 3.4-3.8Ghz - $500

SR5 6/12:

Standard: 3.6-4.0Ghz - $350

SR3:

Black Edition 4C/8T 3.8-4.2Ghz - $250 shipped with AMD Wraith cooler
Standard 4C/4T 4-4.4Ghz - $180

There may be a situation where AMD may want to sell higher base clock + turbo and move TDP up, which is fine as long as they leave the standard 8/16 and 6/12 SKU's with lower TDP for those not interested in pure performance. Alternatively they could sell a 4/8 SKU as a 4970 with out the box raw performance and leave the others to more conservative clock speed / TDP's

 

[Crumpet 1]

I would expect the black edition to ship with the currently unknown "AMD S.3.0 radiator" that we saw advertised on one of the CES motherboards.

 

[juanrga]

Maybe a 3.8-4.2ghz with clise to 6900 consumption. That would be a thing a beauty

From what I gather, a 125W TDP part would have a bclk at 4.0 GHz and turbo of 4.4 based on F3 stepping.

I get about 140W based in F4 steeping.

What's the chance that AMD goes nuts again (9590-like) and releases a 140W 4c/8t chip? How fast would that one be?

First, AMD and Intel don't measure TDP the same. Current 95W Zen samples consumed power similar to Intel 140W chips both in the public demos and in the CPCHarware preview. When asked about TDP, AMD's answer was "similar TDP to Broadwell-E".

Current 8C/16T Zen samples are rated at 95W, but the 4C/8T Zen samples are rated at 65W.

I see impossible a stock 140W 4C/8T Zen chip. The reason is heat density: heat per mm². You can fabricate 300W or 450W chips always and when they are large enough to maintain the heat density under control. The Piledriver FX-9000 series has 1.76x more heat density than 8000 series and the die is large. A hypothetical 140W 4C/8T Zen chip would have 2.15x more heat density than the stock 65W model, which already has higher heat density than the 95W 8C/16T Zen model.

I can imagine a 125W 'enthusiast edition' 8C/16T Zen model bundled with liquid cooling being released somewhat in 2018. The extra wattage would provide 5--10% higher clocks than standard 95W model.

 

[juanrga]

Hi guys i figured i'd share a video from paulshardware about his guesses on prices of Ryzen his is even lower than mine in the 100-200$ market but he does agree with many here about Ryzen being over 600$ for their 8 core part.

https://www.youtube.com/watch?v=nGbC6XLCneU&t=0s

I just think the platform is not as high-end as X99 it’s more between intel's mainstream and X99(I’m not saying it’s not good I actually think it’s more practical for us consumers and gamers). That's why i'm also sticking to my 500$ max price for a 8 core Ryzen with probably a cheaper one offered for 380-400$ would be a nice price to hit Intel's 6 core

Also nicely said Rogue Leader i like reading here on this forum lets stay nice

I think $500 is a good price estimation. However, I am getting rumors that Zen could be beyond $700.

The last rumors are

1. Ryzen is ready. The last delay is due to bug on mobos: a bug on the BIOS.

2. RyZen official presentation in late February in a special conference by AMD.

3. Ryzen available in first half of March. Only 8C/16T model will be available at launch.

4. The 4C/8T model available one month latter.

5. "Prices are not as cheap as you might think".

 

[sarwar_r87]

RX 480 is a midrange card. Higher midrange at best.

If there is a 150 dollar difference for 200 Mhz with a wrath cooler that all self respecting enthusiasts will not use, they won't sell much of SR7 BE

 

[juanrga]

About the claim Zen IPC is 55% higher than Excavator...

I believed that the 55% IPC claim that is being mentioned in several forums (including this) was a measurement made on final silicon. Posts like next one did me believe that

http://www.sweclockers.com/forum/post/16600771

It is not. Desdrenboy has pointed me that the origin of the 55% figure is a post by The Stilt on AT forums. What he said was

If Zen delivers 40% more IPC than Excavator, then it matches Sandy / Ivy Bridge. That hasn't changed and won't ever change.

However if it infact matches Broadwell-E's IPC, then the actual increase over Excavator is ~55% and not 40% like AMD stated.

https://forums.anandtech.com/threads/zen-es-benchmark-from-french-hardware-magazine.2495505/page-23#post-38656781

Therefore the 55% is a purely guessed number assuming that Zen has same IPC than Broadwell. In the same thread I found an interesting post where the author explains his hypothesis that Zen has higher SMT yields than Intel

https://forums.anandtech.com/threads/zen-es-benchmark-from-french-hardware-magazine.2495505/page-25#post-38657740

A bit more than The Stilt's original prediction, and still sounds low to some, but personally I still predict Ryzen will have greater SMT Yeild (%) than Broadwell, which means of course that throughput can be closer to Haswell/BW than it's ST performance is.

As for why I think this will be the case:- Distributed Int Schedulers, and a higher number of statically partitioned structures are the main reasons. Intel's more dynamic approach here should result in higher ST performance, but IMO is a mistake in regards to throughput perf/watt. if not perf/mm , It makes sense to me that AMD chose to avoid such a large unified scheduler.. This is after all an architecture designed to be as balanced as possible, But given it's 2017 now (Is here in Australia!) likely biased towards throughput wherever it didn't impact ST significantly,

His views agree with mines. I expect

IPC+SMT ~ Haswell
IPC < Haswell

 

[cdrkf]

Hi guys i figured i'd share a video from paulshardware about his guesses on prices of Ryzen his is even lower than mine in the 100-200$ market but he does agree with many here about Ryzen being over 600$ for their 8 core part.

https://www.youtube.com/watch?v=nGbC6XLCneU&t=0s

I just think the platform is not as high-end as X99 it’s more between intel's mainstream and X99(I’m not saying it’s not good I actually think it’s more practical for us consumers and gamers). That's why i'm also sticking to my 500$ max price for a 8 core Ryzen with probably a cheaper one offered for 380-400$ would be a nice price to hit Intel's 6 core

Also nicely said Rogue Leader i like reading here on this forum lets stay nice

I think $500 is a good price estimation. However, I am getting rumors that Zen could be beyond $700.

The last rumors are

1. Ryzen is ready. The last delay is due to bug on mobos: a bug on the BIOS.

2. RyZen official presentation in late February in a special conference by AMD.

3. Ryzen available in first half of March. Only 8C/16T model will be available at launch.

4. The 4C/8T model available one month latter.

5. "Prices are not as cheap as you might think".

It wouldn't surprise me if they overprice the 8c/16t model initially (at least the top sku). AMD have a habit of doing that with 'enthusiast' parts. That said, I guess it also shows they are fairly confident about the overall performance profile of the chip- and as with all AMD's top gear the prices will drop to where they should be once the initial wave of super enthusiasts dry up.

 

[sarwar_r87]

IPC+SMT ~ Haswell
IPC < Haswell

good to see not you comparing IPC < haswell and not ~ SB.

compared to Bristol ridge A10-9800, the numbers suggests that SR is 81% faster throughput per core. that means a 178 CB R15-ST score at 3.9Ghz.... (Intel (Kaby Lake) Core i7 7600@3.6Ghz has 180 or 9% slower than KL assuming same level of CPU resource utilization on ST).
https://s23.postimg.org/il35d95kb/Screenshot_20170116_210406.png

ADD:
Intel (Haswell) Core i7 4790K@4 = 181
clock for clock: same as zen

Therefore the 55% is a purely guessed number assuming that Zen has same IPC than Broadwell.

It's not a guess. its was a backward calculation based on the assumption that SMT on both platforms have the same scaling.

very similar to other claims where SMT Zen > intel. but if you insist on calling 55% a guess, we have to say the SMT theory is also a guess.

Distributed Int Schedulers, and a higher number of statically partitioned structures are the main reasons. Intel's more dynamic approach here should result in higher ST performance, but IMO is a mistake in regards to throughput perf/watt. if not perf/mm , It makes sense to me that AMD chose to avoid such a large unified scheduler.. This is after all an architecture designed to be as balanced as possible, But given it's 2017 now (Is here in Australia!) likely biased towards throughput wherever it didn't impact ST significantly,

I saw his post before and while his reasoning is sound, there are other engineering reasons to do so:
1. power saving. as he pointed out a dynamic scheduler approach, it would will cost in terms of power requirements. and amd said their goal was power.
2. I the width of each scheduler are different, AMD has narrower with per int last time I checked. meaning they can combine two of the int to process wider data assuming without losing performance (like they do for AVX2), while running narrower width calculation using less power. in theory it translates to same ST IPC. Hence they have more ints.
3. The statement "Intel's more dynamic approach here should result in higher ST performance" assumes that there are added latencies for combining two or more ints together. there is no proof so confirm this. I, with my puny and research-oriented experience have designed systems that cascade narrow units to wider ones without any added latency. I'm sure AMD can pull it off too.

 

[sarwar_r87]

However, I am getting rumors that Zen could be beyond $700.

3. Ryzen available in first half of March. Only 8C/16T model will be available at launch.

4. The 4C/8T model available one month latter.

5. "Prices are not as cheap as you might think".

I would not consider this rumours all that seriously because
3/4. AMD claimed all will be available at once on several occasion. a bios bug will not prevent them from releasing 4c8t CPU on date with 8c16t.
5. if the zen is priced similar to intel with a 50 dollar undercut, it will not translate to an already SB/IB/Haswell user to upgrade to change their whole platform. there are plenty of ppl still on SB as they donot see the need for KL. This is the problem of entering a market segment that already existed for years and is widely saturated.

For example, I am still debating if I want to get zen as I am still getting 60 fps on all games I do. Only way I will upgrade is if a 6c12t is prices similar to current 4c8ts from intel.

this is my opinion, totally unsubstantiated, but AMD needs to undercut by quite a margin so that intel will be forced to release price cuts and counter products outside their roadmaps.
a 25-50 dollar undercut will not be ground breaking as intel will enjoy some IPC superiority with KL

 

[sarinaide]

Maybe a 3.8-4.2ghz with clise to 6900 consumption. That would be a thing a beauty

From what I gather, a 125W TDP part would have a bclk at 4.0 GHz and turbo of 4.4 based on F3 stepping.

I get about 140W based in F4 steeping.

What's the chance that AMD goes nuts again (9590-like) and releases a 140W 4c/8t chip? How fast would that one be?

First, AMD and Intel don't measure TDP the same. Current 95W Zen samples consumed power similar to Intel 140W chips both in the public demos and in the CPCHarware preview. When asked about TDP, AMD's answer was "similar TDP to Broadwell-E".

Current 8C/16T Zen samples are rated at 95W, but the 4C/8T Zen samples are rated at 65W.

I see impossible a stock 140W 4C/8T Zen chip. The reason is heat density: heat per mm². You can fabricate 300W or 450W chips always and when they are large enough to maintain the heat density under control. The Piledriver FX-9000 series has 1.76x more heat density than 8000 series and the die is large. A hypothetical 140W 4C/8T Zen chip would have 2.15x more heat density than the stock 65W model, which already has higher heat density than the 95W 8C/16T Zen model.

I can imagine a 125W 'enthusiast edition' 8C/16T Zen model bundled with liquid cooling being released somewhat in 2018. The extra wattage would provide 5--10% higher clocks than standard 95W model.

Nobody said anything about a 4C/8T 140W part, but more an 80W model with higher clocks is very feasible.

RX 480 is a midrange card. Higher midrange at best.

If there is a 150 dollar difference for 200 Mhz with a wrath cooler that all self respecting enthusiasts will not use, they won't sell much of SR7 BE

The RX was intended to compete with the GTX900 series and released about 9 months before the 1000 series, with that in mind the RX480 matched the GTX980 which was a high end card and the GTX970 also a high end card that was the best selling GPU at the time. Now it is a midrange card but then it was certainly a high end card which games very well at 1080P.

About the claim Zen IPC is 55% higher than Excavator...

I believed that the 55% IPC claim that is being mentioned in several forums (including this) was a measurement made on final silicon. Posts like next one did me believe that

http://www.sweclockers.com/forum/post/16600771

It is not. Desdrenboy has pointed me that the origin of the 55% figure is a post by The Stilt on AT forums. What he said was

If Zen delivers 40% more IPC than Excavator, then it matches Sandy / Ivy Bridge. That hasn't changed and won't ever change.

However if it infact matches Broadwell-E's IPC, then the actual increase over Excavator is ~55% and not 40% like AMD stated.

https://forums.anandtech.com/threads/zen-es-benchmark-from-french-hardware-magazine.2495505/page-23#post-38656781

Therefore the 55% is a purely guessed number assuming that Zen has same IPC than Broadwell. In the same thread I found an interesting post where the author explains his hypothesis that Zen has higher SMT yields than Intel

https://forums.anandtech.com/threads/zen-es-benchmark-from-french-hardware-magazine.2495505/page-25#post-38657740

A bit more than The Stilt's original prediction, and still sounds low to some, but personally I still predict Ryzen will have greater SMT Yeild (%) than Broadwell, which means of course that throughput can be closer to Haswell/BW than it's ST performance is.

As for why I think this will be the case:- Distributed Int Schedulers, and a higher number of statically partitioned structures are the main reasons. Intel's more dynamic approach here should result in higher ST performance, but IMO is a mistake in regards to throughput perf/watt. if not perf/mm , It makes sense to me that AMD chose to avoid such a large unified scheduler.. This is after all an architecture designed to be as balanced as possible, But given it's 2017 now (Is here in Australia!) likely biased towards throughput wherever it didn't impact ST significantly,

His views agree with mines. I expect

IPC+SMT ~ Haswell
IPC < Haswell

The logic seems very wrong.

If SMT is greater than BDW-E then IPC+SMT is far greater than Haswell because Broadwell is faster than Haswell.

Stilt doesn't rule out the fact that AMD has overdone the 40% ratio, but if 40% = SB/IB then 48 = Haswell and 55% = Broadwell, 60 would likely be SKL/KBL. I am not as brazen on the up of 50% but around 48% is very realistic and likely possible. Assuming as much.

ST ~ Haswell (this metric excludes clock rate)
MT ~ Broadwell