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

[Gon Freecss]

The average shows that the gap is only 11% for 6-core and 15% for 4 core. That is an interesting result. it looks like Ryzen scales better with extra cores.

So as we all knew six months before it released: Haswell IPC.

Actually, when it comes to computing, Zen is between Ivy Bridge and Haswell on average. When it comes to games, it's between Sandy Bridge and Ivy Bridge on average.

1600X vs 7700K
https://www.anandtech.com/bench/product/2018?vs=1826

1600X vs 2600K
https://www.anandtech.com/bench/product/2018?vs=287

I don't know if Sandy has aged that well, or if Intel has been inflating their gains over the generations.

Cheers!

You don't compare IPC with CPUs that have different core counts.

Also, Sandy Bridge is like ~25% slower than Skylake clock for clock on average.

 

[goldstone77]

jdwii the wccftech is somewhat misleading.

AMD’s EPYC Manages to Get Working on the ASUS ROG ZENITH EXTREME After a Small DIY Mod [Video Given Below]

He never actually was able to get it to post just get error codes. In a video a week earlier he actually shows that the processors are different.

 

[aldaia]

jdwii the wccftech is somewhat misleading.

AMD’s EPYC Manages to Get Working on the ASUS ROG ZENITH EXTREME After a Small DIY Mod [Video Given Below]

He never actually was able to get it to post just get error codes. In a video a week earlier he actually shows that the processors are different.

Shitty clickbait article. I suspect that an honest title like "As everybody knows, EPYC does not work on Threadripper socket" will get less clicks.
2 of the dies in the EPYC package have their memory channels unconnected. I doubt that such a thing will ever boot even if everything else was identical.

 

[YoAndy]

Spoiler

So IPC:

This is exactly what I showed with the 7700K vs the 1500X.
That is the same 8.5% advantage the 7700K has for single thread, but you lack the graphs showing that the 1500X displays superior multi-threaded performance by 8.2%. Juangra shows the difference in SMT implementation between Kabey Lake and Ryzen accounting for the multi-threaded performance gains.

 

[jdwii]

Spoiler

So IPC:

This is exactly what I showed with the 7700K vs the 1500X.
That is the same 8.5% advantage the 7700K has for single thread, but you lack the graphs showing that the 1500X displays superior multi-threaded performance by 8.2%. Juangra shows the difference in SMT implementation between Kabey Lake and Ryzen accounting for the multi-threaded performance gains.

What is the point here this has nothing to do with IPC?

 

[goldstone77]

YoAndy.... <facepalm>

 

[jdwii]

More Raven ridge leaks also with crazy GPU pricing this might be the best bet into building a budget gaming rig


https://wccftech.com/amd-ryzen-vega-cpu-new-units-new-upgrades/

 

[goldstone77]

I'm sure they will do great for the true budget builds. I can't wait to see the benchmarks on them.

 

[juanrga]

Here we go again with IPC fights. Cherrypicked applications can show from no IPC gap to 50% gap (depending on who cherrypicks). hardware.fr did a recent comparison between 4 and 6 core coffe lake vs 4 and 6 core ryzen. All of them at 3.5GHz. It's probably the most extensive IPC comparisson that I've seen (12 applications).

The average shows that the gap is only 11% for 6-core and 15% for 4 core. That is an interesting result. it looks like Ryzen scales better with extra cores.

Or they tested a CoffeLake chip with beta or first version BIOS that affected the performance...

So the average IPC gap between Zen and Kabylake is about 15% for applications and about 18% for games. But this is under windows. When adding linux applications the average IPC between Zen and Kabylake increases to about 30%

 

[juanrga]

Spoiler

The SMT implementation is different. You show that very well will pictures. Simplified all computing is going to have trade-offs depending on what you are trying accomplish. And then you gain IPC through uArch improvements/software optimization to perform certain tasks, or "tuning" if that is what we want to call it. I've mentioned this before in this thread with measurements look at the single thread and multi-thread test.:
http://www.tomshardware.com/forum/id-3327589/amd-ryzen-megathread-faq-resources/page-41.html#20123237

-8.5% 1500X@3.5GHz vs 7700k@3.5GHz

+8.2% 1500X@3.5GHz vs 7700k@3.5GHz

+4.3% 1500X@3.5GHz vs 7700k@3.5GHz in Physics showing Ryzens superior number crunching vs. Intel
-5.3% 1500X@3.5GHz vs 7700k@3.5GHz overall
But look at the 7600k with a base clock of 3.8GHz, which is 8.5% faster than the 1500X
+30.9% 1500X@3.5GHz vs 7600K@3.8GHz
-3.4% 1500X@3.5GHZ vs 7600K@3.8GHz

-5.2% 1500X@3.5GHz vs 7700K@3.5GHz overall
But look at the 7600k with a base clock of 3.8GHz, which is 8.5% faster than the 1500X
0.2% 1500X@3.5GHz vs 7600K@3.8GHz overall

This is first generation Ryzen, it will undoubtedly take years to optimize uArch/software to improve performance with all application. Some synthetics show a lot of raw performance that isn't being utilized by programs. AMD developed Ryzen chips to have superior multithreading clock for clock, by virtue of their implementation of SMT, which you illustrate very well with pictures. AMD also advertised that, and that is what it was trying to accomplish. Ryzen's 1800X($450) was made to compete with Intel's 6900X($1,100) which it did compete against very well despite only having dual channel RAM. The 1700 is arguably the much better buy because it was only about couple hundred MHz difference in processor speed overlocked, which could be bought for $310 and came with a cooler capable over overclocking to 3.8GHz. This is what made Ryzen so amazing.

(i)
The SMT implementation is different, not better.

(ii)
Cinebench is a favorable case for AMD. As shown the IPC gap is about 10% in Cinebench but about 30% in several other benchmarks.

(iii)
1500X vs 7770k is a favorable case for AMD. The IPC gap is only 8.5%. But choosing other models the gap increases to 10.7--11.4%.

Moreover, no one of AMD chips in that graph is stock. All them are overclocked, because memory was pushed above stock settings and this automatically overclocks the IF interconnect on Zen processors, which reduces L3 and CCX-CCX latencies and increases the IPC.

(iv)
The idea that current software is optimized for Intel and that software will run faster on Ryzen when optimized for the new muarch is easily disproved.

Current Windows software usually targets generic older processor and doesn't use all the potential in modern Intel processors. The Stilt compiled code specifically for each muarch (march=znver1 and mtune=znver1 for Zen), and the IPC gap between AMD and Intel increased. This optimization made at compile time is the reason why you can see some linux benches where Kabylake is 2x faster than Zen clock-for-clock, but you don't see those huge gaps in Windows reviews, not testing this kind of code.

(v)
Your last graph shows that 8-core Haswell is 14% faster than 1800X, despite having lower clocks 3.0/3.5GHz vs 3.6/4.0GHz.

(vi)
Ryzen isn't developed to "have superior multithreading clock for clock". As I have explained hundred of times, the Zen muarch is optimized for throughput. Intel muarchs are optimized for latency. That is the reason why AMD shines in multithreading workloads as Cinebench and POV-Ray, but loses badly in multithreaded workloads as Hitman

(vii)
Quad-channel has no impact on compute-bound workloads. The impact is on memory bound-workloads.

(viii)
Zen got momentum in early 2017 only when comparing it with older processors from Intel. Many reviews compared the 1800X with the 7700k and mentioned how Ryzen offered very superior performance in rendering, encoding, and so. Reviews mostly recommended Ryzen for well-threaded applications and Kabylake for gaming. Then Intel added two extra cores on Coffelake and the landscape changed. Now the i7-7800k can beat the 1800X even in well-threaded applications.

6C Coffee ~ 8C Zen

CoffeLake core is about 30--40% faster than Zen core. Pinnacle Ridge is expected to bring about 10% higher performance. This will reduce the gap to something about 20--30%.

 

[jdwii]

Getting sick and tired of users here using outliers as their base of proof! The software that the user uses is most important and i'll argue that a wide range of software should be used in judging IPC.

Now come on i can find cases where a 1800X beats a 6900X but those cases are rare. On average Ryzen IPC is placed right between sandy and haswell. This isn't some conspiracy theory it's just pure and boring facts.

 

[juanrga]

Getting sick and tired of users here using outliers as their base of proof! The software that the user uses is most important and i'll argue that a wide range of software should be used in judging IPC.

Now come on i can find cases where a 1800X beats a 6900X but those cases are rare. On average Ryzen IPC is placed right between sandy and haswell. This isn't some conspiracy theory it's just pure and boring facts.

Excellent post!

Indeed, on average the IPC is between Sandy and Haswell, being more Haswell-like when one considers only applications and more Sandy-like when one considers only games. For games we can see how 4C/4T RyZen is like Sandy 4C/4T

R3 1300X ~ i5 2500k ~ R3 1200

 

[goldstone77]

I've said this before when comparing apps.. Comparing results and finding 1000% better IPC in windows calculator program is laughable. How often do you use the app? How much of a noticeable difference can you detect? Some people love to jump off into the deep end of the pool. Directly comparing results that matter to workloads people most often use is what really matters to 99% of consumers, and most of them are just browsing the internet, tweeting, FB, instagram, netflix, youtube, gaming, and typing with a million apps running in the background. There comes a point when pointing and clicking and typing is fairly instant, and IPC while it can be used as a measure, has become laughable looking at performance of some applications today. Tasks like rendering videos or streaming are more relevant today than they ever have been with the explosion of Twitch and YouTube. Ryzen's implementation of SMT greatly benefits those multi-threaded workloads. Ryzen supplanted the 6900K with a price tag reachable by the masses for these workloads. People can attempt to undermine Ryzen's influence with a handful of outlier IPC graphs, but it will not take away from the gift that Ryzen was to the masses. Mult-threading for a faction of the cost Intel offered!

 

[juanrga]

Synthetics as Cinebench means little, but applications as Hitman, 7-zip, Blender, WinRAR, Adobe Lightroom, x264, represent real-world usage. Many applications tested by The Stilt also represent real-life usage for people that use computers for work (physical simulation, programming, design,...) E.g. Linpack is used to measure performance of supercomputers. The famous top500 list uses Linpack to rank computers from faster to slower. No one uses Cinebench scores.

99% of consumers doing only basic consumer tasks as facebook, tweets, email, watching youtube videos, and so on don't need any RyZen chip. A phone or a tablet is enough for such tasks.

 

[goldstone77]

99% of consumers doing only basic consumer tasks as facebook, tweets, email, watching youtube videos, and so on don't need any RyZen chip. A phone or a tablet is enough for such tasks.

Exactly my point, buy the cheaper CPUs! Hello Ryzen!

 

[-Fran-]

Judgement on "what is useful" or not is a slippery slope.

I do agree "server benchmarks" are always useful, but the rest are just a "the more, the better".

Also, by that same token, Juan, they don't need anything higher than a Celeron. Well, hence why smartphones and tablets took so much terrain from PCs, I guess.

 

[goldstone77]

Synthetics as Cinebench means little, but applications as Hitman, 7-zip, Blender, WinRAR, Adobe Lightroom, x264, represent real-world usage.

Gaming performance is highly subjective. Are we talking about playing 1080p using a 1080Ti with a 60HZ monitor? Ryzen will easily achieve the performance required to fill 60HZ. Using a 1060 the difference in performance between Intel and Ryzen is cut in half. How many people actually played/play Hitman, and then how many of those people were playing that non FPS based shooter game at 144HZ? You make a weak argument for it's inclusion and relevance with 99% of the users that would play video games when most of those numbers are only academic and not having an impact on game play. 7-zip and WinRAR performance is mostly laughable. How often does anyone use these programs, and on the occasion that someone does use them, how significant is the performance impact going to be to make you doubt your purchase of the CPU. In other words, no one ever asks which is the best CPU to buy to zip and unzip my files. Blender is relevant for real world workloads. Handbakes is relevant for real world workloads I or my GF use it almost every day to encode TV shows.

 

[goldstone77]

Judgement on "what is useful" or not is a slippery slope.

I do agree "server benchmarks" are always useful, but the rest are just a "the more, the better".

Also, by that same token, Juan, they don't need anything higher than a Celeron. Well, hence why smartphones and tablets took so much terrain from PCs, I guess.

Mobiles have absolute devastated the PC market.

 

[juanrga]

99% of consumers doing only basic consumer tasks as facebook, tweets, email, watching youtube videos, and so on don't need any RyZen chip. A phone or a tablet is enough for such tasks.

Exactly my point, buy the cheaper CPUs! Hello Ryzen!

Ryzen doesn't fit in a tablet even less is a phone and it is expensive compared to the cheap processors used for those tasks.

 

[juanrga]

Judgement on "what is useful" or not is a slippery slope.

I do agree "server benchmarks" are always useful, but the rest are just a "the more, the better".

Also, by that same token, Juan, they don't need anything higher than a Celeron. Well, hence why smartphones and tablets took so much terrain from PCs, I guess.

Actually #1 in Amazon CPU bestselling sales is a Celeron processor.

 

[juanrga]

Synthetics as Cinebench means little, but applications as Hitman, 7-zip, Blender, WinRAR, Adobe Lightroom, x264, represent real-world usage.

Gaming performance is highly subjective. Are we talking about playing 1080p using a 1080Ti with a 60HZ monitor? Ryzen will easily achieve the performance required to fill 60HZ. Using a 1060 the difference in performance between Intel and Ryzen is cut in half. How many people actually played/play Hitman, and then how many of those people were playing that non FPS based shooter game at 144HZ? You make a weak argument for it's inclusion and relevance with 99% of the users that would play video games when most of those numbers are only academic and not having an impact on game play. 7-zip and WinRAR performance is mostly laughable. How often does anyone use these programs, and on the occasion that someone does use them, how significant is the performance impact going to be to make you doubt your purchase of the CPU. In other words, no one ever asks which is the best CPU to buy to zip and unzip my files. Blender is relevant for real world workloads. Handbakes is relevant for real world workloads I or my GF use it almost every day to encode TV shows.

Remember I initially mentioned Hitman, 7-zip, Blender, WinRAR, Adobe Lightroom,... because someone else said that the maximum IPC gap measured between Intel and AMD was 10%, which of course isn't true, as benchmarks show 30% and higher.

Any of those applications is more related to real usage than Cinebench, which is not even an application, but a synthetic benchmark.

There are broad consensus among reviewers and people than Intel chips are much better for gaming. It is very interesting how the same people that claims that AMD is good enough for gaming, however, doesn't apply the same criteria when talking about applications. Apparently spending 3 seconds more on encoding a video is something terrible!

 

[goldstone77]

Synthetics as Cinebench means little, but applications as Hitman, 7-zip, Blender, WinRAR, Adobe Lightroom, x264, represent real-world usage.

Gaming performance is highly subjective. Are we talking about playing 1080p using a 1080Ti with a 60HZ monitor? Ryzen will easily achieve the performance required to fill 60HZ. Using a 1060 the difference in performance between Intel and Ryzen is cut in half. How many people actually played/play Hitman, and then how many of those people were playing that non FPS based shooter game at 144HZ? You make a weak argument for it's inclusion and relevance with 99% of the users that would play video games when most of those numbers are only academic and not having an impact on game play. 7-zip and WinRAR performance is mostly laughable. How often does anyone use these programs, and on the occasion that someone does use them, how significant is the performance impact going to be to make you doubt your purchase of the CPU. In other words, no one ever asks which is the best CPU to buy to zip and unzip my files. Blender is relevant for real world workloads. Handbakes is relevant for real world workloads I or my GF use it almost every day to encode TV shows.

Remember I initially mentioned Hitman, 7-zip, Blender, WinRAR, Adobe Lightroom,... because someone else said that the maximum IPC gap measured between Intel and AMD was 10%, which of course isn't true, as benchmarks show 30% and higher.

Any of those applications is more related to real usage than Cinebench, which is not even an application, but a synthetic benchmark.

There are broad consensus among reviewers and people than Intel chips are much better for gaming. It is very interesting how the same people that claims that AMD is good enough for gaming, however, doesn't apply the same criteria when talking about applications. Apparently spending 3 seconds more on encoding a video is something terrible!

You are confusing seconds for hours more encoding.

 

[goldstone77]

AMD Tech Day at CES: 2018 Roadmap Revealed, with Ryzen APUs, Zen+ on 12nm, Vega on 7nm
by Ian Cutress on February 1, 2018 8:45 AM EST
https://www.anandtech.com/show/12233/amd-tech-day-at-ces-2018-roadmap-revealed-with-ryzen-apus-zen-on-12nm-vega-on-7nm/8

For the new processors, AMD is claiming a 10% boost in performance per watt overall. This is going to be taken as higher clocks for the same power at the high-end and lower power for the same frequency for more power sensitive products. AMD’s slide from Jim Anderson’s presentation above specifically says ‘higher clocks’, which when combined with Mark Papermaster’s presentation which states ‘10%+ performance vs. 14LPP’, we could take this to mean that we should expect anywhere from an 8%+ increase in frequencies with 2nd Generation Ryzen for desktop (not 10%, accounting for the fact that the power/efficiency curve gets worse the faster you are).
he 10% figure is corroborated by GlobalFoundries, who back in 2017 stated that its 12LP process will offer a 10% better performance. GF also stated that the 12LP process also offers a 15% area reduction against 14LPP. This is done, according to GlobalFoundries, by using 7.5T libraries rather than 9T libraries. This requires its customers to ‘recompile’ their 14LPP designs for the adjusted 12LP process.

To put this into context, this means we might see 4.3-4.5 GHz processors with the 2nd Generation Ryzen sticker where we used to see 4.0 GHz processors. It will be interesting to see how AMD and GlobalFoundries have moved that seemingly hard overclock limit we saw on processors like the Ryzen 7 1800X, which had a hard time moving past 4.2 GHz on all cores.

We are having a tour around GlobalFoundries Fab 8 in New York this week, and are set to spend some time with CTO Gary Patton, talking about the new process. While he is unlikely to quote specifically about how AMD has used the new design rules to its advantage, he might tell us some more information about the process in general.

As for 2nd Generation Ryzen, we are told that the major motherboard vendors and OEMs already have near-final engineering samples to hand to ensure updated compatibility.

It will be interesting to hear what Ian Cutress finds out from the tour at GlobalFoundries!

 

[juanrga]

AMD Tech Day at CES: 2018 Roadmap Revealed, with Ryzen APUs, Zen+ on 12nm, Vega on 7nm
by Ian Cutress on February 1, 2018 8:45 AM EST
https://www.anandtech.com/show/12233/amd-tech-day-at-ces-2018-roadmap-revealed-with-ryzen-apus-zen-on-12nm-vega-on-7nm/8

For the new processors, AMD is claiming a 10% boost in performance per watt overall. This is going to be taken as higher clocks for the same power at the high-end and lower power for the same frequency for more power sensitive products. AMD’s slide from Jim Anderson’s presentation above specifically says ‘higher clocks’, which when combined with Mark Papermaster’s presentation which states ‘10%+ performance vs. 14LPP’, we could take this to mean that we should expect anywhere from an 8%+ increase in frequencies with 2nd Generation Ryzen for desktop (not 10%, accounting for the fact that the power/efficiency curve gets worse the faster you are).
he 10% figure is corroborated by GlobalFoundries, who back in 2017 stated that its 12LP process will offer a 10% better performance. GF also stated that the 12LP process also offers a 15% area reduction against 14LPP. This is done, according to GlobalFoundries, by using 7.5T libraries rather than 9T libraries. This requires its customers to ‘recompile’ their 14LPP designs for the adjusted 12LP process.

To put this into context, this means we might see 4.3-4.5 GHz processors with the 2nd Generation Ryzen sticker where we used to see 4.0 GHz processors. It will be interesting to see how AMD and GlobalFoundries have moved that seemingly hard overclock limit we saw on processors like the Ryzen 7 1800X, which had a hard time moving past 4.2 GHz on all cores.

We are having a tour around GlobalFoundries Fab 8 in New York this week, and are set to spend some time with CTO Gary Patton, talking about the new process. While he is unlikely to quote specifically about how AMD has used the new design rules to its advantage, he might tell us some more information about the process in general.

As for 2nd Generation Ryzen, we are told that the major motherboard vendors and OEMs already have near-final engineering samples to hand to ensure updated compatibility.

It will be interesting to hear what Ian Cutress finds out from the tour at GlobalFoundries!

Zen+ is a marketing rebrand for Zen. 12nm is a marketing rebrand for 14nm+.

The 10% better performance and 15% area reduction aren't using 14LPP as baseline, but unnamed TSMC 16nm node.

The known Pinnacle Ridge qualification samples and Raven Ridge commercial processors on 14nm+ have 100--200MHz higher than the 14nm counterparts. 500MHz on top of 4GHz looks a pipedream, but it cannot be ruled out completely.

Interesting that Ian mentions "past 4.2GHz" for the 1800X, when the average overclock is 4.0GHz.

 

[8350rocks]

is it really an IPC gap or an optimization issue? Ofcourse programmers tend to optimize their code according to the market sales. and Intel has been the market leader for nearly a decade, surpassing AMD sales with huge margin. I am not an expert but the cause of this IPC gap can be defined like "Hardware IPC+Software Efficiency". And if we take this into consideration, how much performance can be gained if the code is to be optimized for Ryzen? will it be still the same? or do we have some opportunity here?

Windows code is usually optimized for ancient CPUs, so it cannot use the newest features in modern Intel CPUs and that is why the average IPC gap with RyZen is lower ~15% in windows reviews.

Linux code is usually more optimized for modern CPUs. Moreover, The Stilt compiled the linux code for the latest architectures (including RyZen) and that is why the average IPC he measured increased to ~30%.

In Windows reviews you can find few benchmarks where the IPC gap is around ~30%. I gave some examples above (Audacity, Hitman, 7-zip, WinRAR, Adobe Lightroom). In Linux reviews you can find many benchmarks where the IPC is around 30% and you can benchmarks where the IPC gap is higher than 200%.

Those benchmarks are suspect because Ryzen typically outperforms Intel clock for clock in linpack due to the considerably better SMT implementation.

Additionally, I notice that you are only using benchmarks where AVX advantage is heavily leveraged to exacerbate your claims.

There is no better SMT implementation. The Zen core extracts less ILP from code and thus leaves more execution units free for executing a second thread

The Stilt just proves what we know about the microarchitectures. Haswell and Kabylake are 16FLOP/core, whereas Zen is 8FLOP/core. So the performance gap is about the double and The Stilt measured 1.89 and 2.22. Note that Skylake-X and SP are 32FLOP/core and increase still more the performance gap with Zen.

I have given both AVX and non-AVX workloads. Moreover, there is nothing wrong on giving AVX workloads. Would us stop using the last versions of Handbrake and Blender because support AVX and put AMD CPUs in a worse position? Obviously, no.

FLOPs are AVX workloads, how many consumer applications are compiled to use AVX code? Zero in a windows environment?

Ok.

When AVX is literally as prevalent as SSE4 code, then we can start talking about them being relevant. AVX is not even a regular in games...so pretending that blue sky is anything else is just fallacy.