AMD Ryzen 2000 (12nm Zen+) expectations

[Topweasel]

Why you think that IF is not going to be touched? higher core clock or IPC won't help much.
Even, if Ryzen 2 would have 15% higher IPC with 1067MHz DF clock at stock 2133MT/s DRam will be slower then ryzen with 2133MHz DF, DRAM 2133MT/s in games.

Because IF is inherent of their whole development process from the very beginning. It's how inter CCX communication is handled, it's how intra CCX communication is handled, it's how cross die communication is handled, it how cross socket communication is handled. It how GCU communication is handled, it how GPU memory communication is handled. It's how GPU to GPU and it's how GPU to CPU communication is handled. AMD isn't going to change that on a whim. But just look at their GPU's an increase in width and faster ram. That's how AMD got the scalable to ~500GB/s. It's link to memory is part of it's standard now and expecting AMD to change it cause a bunch guys on the Internet are absolutely sure that it is why games don't run as well as they think it should is a pipe dream. If anything AMD will make the pathway wider on future Zen archs. Won't be Zen+ and probably not Zen 2. But if AMD could or was willing to change the ratio or just plain set a speed Zen and Vega wouldn't have the exact same policy. AMD made a judgement call when developing IF that finding a bandwidth that was acceptable for CPU's and hit their goals on Vega and went with that. Otherwise you increase the complexity decrease yields on vega just to increase the performance on Zen and again that assumes what we see is caused by IF and not for example the L3 system they are using. You know the same one Intel adopted on SL-X that sees a similar drop in performance compared to expectations.

 

[Topweasel]

Pinnacle Ridge is not a one off die. Going forward server and client will have separate purpose built dies. 7nm Rome will be built at GF 7SoC 6T for maximum density and power efficiency while 7nm Ryzen will be built using 7HPC for maximum clock potential. The changes AMD makes in Pinnacle Ridge are not one off. Rather they are the beginning to a dedicated client die cadence.

But that you are talking about using the same die configuration but using different processes. That isn't the same as making a bespoke "client CPU" which the APU's are the only real ones they have (even if those get adopted into professional and server usage). Pinnacle Ridge though does have the sense that what AMD does to that wouldn't last. AMD is unlikely to use it for EPYC and while they could get a boost in using it in TR, they may not feel separating the production of TR and EPYC that much. In that sense it is a one off.

Still doesn't mean that AMD makes any changes to IF.

 

[IRobot23]

Because IF is inherent of their whole development process from the very beginning. It's how inter CCX communication is handled, it's how intra CCX communication is handled, it's how cross die communication is handled, it how cross socket communication is handled. It how GCU communication is handled, it how GPU memory communication is handled. It's how GPU to GPU and it's how GPU to CPU communication is handled. AMD isn't going to change that on a whim. But just look at their GPU's an increase in width and faster ram. That's how AMD got the scalable to ~500GB/s. It's link to memory is part of it's standard now and expecting AMD to change it cause a bunch guys on the Internet are absolutely sure that it is why games don't run as well as they think it should is a pipe dream. If anything AMD will make the pathway wider on future Zen archs. Won't be Zen+ and probably not Zen 2. But if AMD could or was willing to change the ratio or just plain set a speed Zen and Vega wouldn't have the exact same policy. AMD made a judgement call when developing IF that finding a bandwidth that was acceptable for CPU's and hit their goals on Vega and went with that. Otherwise you increase the complexity decrease yields on vega just to increase the performance on Zen and again that assumes what we see is caused by IF and not for example the L3 system they are using. You know the same one Intel adopted on SL-X that sees a similar drop in performance compared to expectations.

Still AMD knew that they will need to have answer for desktop. Memory latency is main problem here. memory latency is very important for some things.

Anyway AMD has really go IMC, so maybe with lowering that latency could give them an advantage.
AMD is very near of reaching 60GB/s on dual channel.

 

[Topweasel]

Still AMD knew that they will need to have answer for desktop. Memory latency is main problem here. memory latency is very important for some things.

Sure it is. But people act like it's the end of the world if it's not absolutely the end all be all in all avenues. AMD had goals and performance markers they needed to hit. This includes things like power usage, Yields, the ability for IF to work on every layer. Like all CPU designs that means trade-offs and compromises. Memory and core to core latency where obviously part of that and their impact on the big picture probably isn't that large.

This isn't about AMD saying gaming is unimportant. They aren't going to ignore us (though I think it would be better if they did). But they aren't going to prioritize that if it threatens their long term goals.

I expect AMD will work hard on tweaking what little they are going to touch on Zen+. I expect some decent arch changes going into Zen 2. But I doubt any of those is going to deal with IF outside increase width in the die. Not unless they are ready to change out all the impacted systems at the same time. That means new die for Ryzen, new Die for APU's, new arch for video.

 

[Zucker2k]

All SKL-X chips are ahead of all Ryzen chips in gaming.

He didn't talk about CFL , did he ? He Only said about Ryzen 1600 vs Core i7 7820X .

Forget the 7820X. ALL SKL-X chips are overall faster in gaming than any AMD chip; from Ryzen 3,5,7, and Threadripper. One could always find the corner case where an AMD chip does better than SKL-X in a few games but if anyone wants to look at it objectively, the gaming hierarchy (even for those cpu-heavy scenarios) goes like this: Coffee Lake > Broadwell-E > SKL-X > Kabylake > AMD. For cpu-light games, Coffee Lake > Kabylake > Broadwell-E, SKL-X > AMD. Even older gen Intels fare better than the AMD chips in games where single thread is king, obviously.

https://www.computerbase.de/thema/prozessor/rangliste/

Productivity-wise, Coffee Lake is slightly better than the 1800x.

 

[IRobot23]

All SKL-X chips are ahead of all Ryzen chips in gaming.

Forget the 7820X. ALL SKL-X chips are overall faster in gaming than any AMD chip; from Ryzen 3,5,7, and Threadripper. One could always find the corner case where an AMD chip does better than SKL-X in a few games but if anyone wants to look at it objectively, the gaming hierarchy (even for those cpu-heavy scenarios) goes like this: Coffee Lake > Broadwell-E > SKL-X > Kabylake > AMD. For cpu-light games, Coffee Lake > Kabylake > Broadwell-E, SKL-X > AMD. Even older gen Intels fare better than the AMD chips in games where single thread is king, obviously.

https://www.computerbase.de/thema/prozessor/rangliste/

Productivity-wise, Coffee Lake is slightly better than the 1800x.

depends on review
https://www.techspot.com/review/1450-core-i7-vs-ryzen-5-hexa-core/page9.html

 

[Zucker2k]

depends on review
https://www.techspot.com/review/1450-core-i7-vs-ryzen-5-hexa-core/page9.html

The 7800x beats the 1600x at stock clcoks, overall. So what's your point?
Edit: Plus, that site's results were always suspect anyway, since they don't match any of the reputable review site results, as is clearly evident in the Comuterbase.de link I posted. I could post more if you'd like.

 

[IRobot23]

The 7800x beats the 1600x at stock clcoks, overall. So what's your point?
Edit: Plus, that site's results were always suspect anyway, since they don't match any of the reputable review site results, as is clearly evident in the Comuterbase.de link I posted. I could post more if you'd like.

Here was comparison between R5 1600 vs i7 7800X.

Any, if AMD wants to beat Intel in gaming they need to reduce memory latency somehow.

Suspect? So computerbase is 100% legit, while techspot is fake?

 

[Zucker2k]

Here was comparison between R5 1600 vs i7 7800X.

Any, if AMD wants to beat Intel in gaming they need to reduce memory latency somehow.

Suspect? So computerbase is 100% legit, while techspot is fake?

Compare TechSpot results to the other sites yourself.

 

[IRobot23]

Compare TechSpot results to the other sites yourself.

Well, we are not here for blaming sites or even talking about i7 7800X. main problem with ryzen is IF speed. If they fix that latency somehow then AMD can clearly battle for the win in gaming.
Like I said I do not own Coffee lake.

 

[Topweasel]

Well, we are not here for blaming sites or even talking about i7 7800X. main problem with ryzen is IF speed. If they fix that latency somehow then AMD can clearly battle for the win in gaming.
Like I said I do not own Coffee lake.

For the love of god please tell me how you became soooooooooo certain that the "problem" in gaming is memory latency?

 

[tamz_msc]

My predictions for Zen+(Zen refresh)

• Wider compatibility for memory - doesn't matter whether the ICs are Samsung, Hynix or Micron - they should allow operating at their XMP speeds with some tweaking. If not then allowing them to run at 2666MHz 1DPC SR which is officially supported, provided the module itself can handle that of course.
  
  
• A modest 200-300MHz bump in base clocks.

The rumored AGESA 1.0.0.7 update already indicates that the first one is being worked on, and I expect AMD will make good progress by the time of the actual launch.

 

[IRobot23]

For the love of god please tell me how you became soooooooooo certain that the "problem" in gaming is memory latency?

Can you please tell what is the problem for ryzen?

 

[Topweasel]

Can you please tell what is the problem for ryzen?

First one must assume that there is an "issue" with Ryzen that AMD must address. Between the major use cases for the Zeppelin die, it's major selling points as Ryzen, I am not sure that is the case. Specially when you consider how little Ryzen's shortcomings actually impact gaming in normal use cases (gamers generally play GPU bottle-necked configurations). Between clock speed, IPC, general optimization around the i7 configuration, cache configuration and yes even core to core and core to memory latency. AMD is unlikely to compete on the low res, low IQ, ultra high FPS, gaming systems but that is a small niche. It means it wont be the best "eSport" system, but that doesn't make it a bad CPU for gaming. It's a bad narrative.

But if you really want to assess it's struggles in high FPS gaming I have already mentioned probably the largest reason reason for it. L3 cache config. It's probably compounded with higher memory latency, but it's no surprise that Intel adopting a move to a similar cache config on the SL-X sees a similar performance envelope compared to SL and KBL clock for clock. Sure it's faster than Ryzen at both a clock for clock and it's obvious clock advantage. But the drop off between it and it's 4 and 6 core base platform cousins, or Broadwell E are significant.

That said while I am confident of my assessment. I am doing two things you aren't. One I am not selling it as fact. It's not a fact. It's my interpretation of information available. 2. I am not pretending to enough about CPU design. What AMD's goals are. What this would affect in other use cases. To suggest that this is a must change system. I in fact think otherwise considering the switch Intel made, this is probably in the long term an important part of the design made to work better in scenarios more important than an extra 10+ FPS when you are already above 100 FPS. For gaming i would assume that if this is going to be a significant change in design going forward for both parties, games will be coded differently to minimize this impact. Finally as DX12/Vulkan and console ports more "enhanced" for the PS4pro and XboxOne X games come out, more cores are going to matter more than the max output per core.

 

[CatMerc]

For the love of god please tell me how you became soooooooooo certain that the "problem" in gaming is memory latency?

Because of the gains exhibited by Ryzen with lower latency RAM, ruling out the IF itself because of lower latency RAM > higher clock, combined with using profilers for looking at how sensitive applications are to memory latency and how they align to Ryzen's relative performance. And checking how Coffee Lake reacts to memory speed gains, to see if the gains are realized on both equally.

It really does tell the story wide and clear. Memory latency is holding Ryzen back heavily. That's not to say core clockspeed isn't, but those two are the largest culprits, and increasing clockspeed alone won't bring Ryzen to parity with Intel.

 

[IRobot23]

First one must assume that there is an "issue" with Ryzen that AMD must address. Between the major use cases for the Zeppelin die, it's major selling points as Ryzen, I am not sure that is the case. Specially when you consider how little Ryzen's shortcomings actually impact gaming in normal use cases (gamers generally play GPU bottle-necked configurations). Between clock speed, IPC, general optimization around the i7 configuration, cache configuration and yes even core to core and core to memory latency. AMD is unlikely to compete on the low res, low IQ, ultra high FPS, gaming systems but that is a small niche. It means it wont be the best "eSport" system, but that doesn't make it a bad CPU for gaming. It's a bad narrative.

But if you really want to assess it's struggles in high FPS gaming I have already mentioned probably the largest reason reason for it. L3 cache config. It's probably compounded with higher memory latency, but it's no surprise that Intel adopting a move to a similar cache config on the SL-X sees a similar performance envelope compared to SL and KBL clock for clock. Sure it's faster than Ryzen at both a clock for clock and it's obvious clock advantage. But the drop off between it and it's 4 and 6 core base platform cousins, or Broadwell E are significant.

That said while I am confident of my assessment. I am doing two things you aren't. One I am not selling it as fact. It's not a fact. It's my interpretation of information available. 2. I am not pretending to enough about CPU design. What AMD's goals are. What this would affect in other use cases. To suggest that this is a must change system. I in fact think otherwise considering the switch Intel made, this is probably in the long term an important part of the design made to work better in scenarios more important than an extra 10+ FPS when you are already above 100 FPS. For gaming i would assume that if this is going to be a significant change in design going forward for both parties, games will be coded differently to minimize this impact. Finally as DX12/Vulkan and console ports more "enhanced" for the PS4pro and XboxOne X games come out, more cores are going to matter more than the max output per core.

Ryzen with DDR4 3200MHz CL14 is prety capable of 144Hz gaming.

Let me give you and example :
They say kaby/skylake/coffee IPC is around 6-7% higher, depends on instruction.
1. So to get same ST performance we have to have lower clock on skylake platform.
2. Ryzen plaform is 4+0 so whole CCX is disabled.
3. So both platform perform same in ST and MT
4. Both running 2400MHz DDR4, while ryzen will have higher or same bandwidth.
5. Ryzen will loose

No, I do not think that L3 cache is main reason for Skylake-X or Ryzen. Both bandwidth and latency are very important of getting higher FPS, some games are less dependent some more. That are my thoughts based on my benchmarking. Anyway I am no expert so I could agree with you or at least respect your thoughts.

Just for informations, even if 70ns vs 45ns sounds small difference, it is still x1.55.

If I get coffee lake I will try to get higher DRAM latency and test it.

PS: I can do a simple test.
I can do 2+2 vs 4+0 both configuration with single channel 3200MT/s+ (lower bandwidth, lower memory latency) vs dual channel 1866/2133MT/s (higher bandwidth, higher latency).

 

[Topweasel]

Ryzen with DDR4 3200MHz CL14 is prety capable of 144Hz gaming.

Let me give you and example :
They say kaby/skylake/coffee IPC is around 6-7% higher, depends on instruction.
1. So to get same ST performance we have to have lower clock on skylake platform.
2. Ryzen plaform is 4+0 so whole CCX is disabled.
3. So both platform perform same in ST and MT
4. Both running 2400MHz DDR4, while ryzen will have higher or same bandwidth.
5. Ryzen will loose

No, I do not think that L3 cache is main reason for Skylake-X or Ryzen. Both bandwidth and latency are very important of getting higher FPS, some games are less dependent some more. That are my thoughts based on my benchmarking. Anyway I am no expert so I could agree with you or at least respect your thoughts.

Just for informations, even if 70ns vs 45ns sounds small difference, it is still x1.55.

If I get coffee lake I will try to get higher DRAM latency and test it.

PS: I can do a simple test.
I can do 2+2 vs 4+0 both configuration with single channel 3200MT/s+ (lower bandwidth, lower memory latency) vs dual channel 1866/2133MT/s (higher bandwidth, higher latency).

Memory tests have been done between Ryzen and KabyLake. KabyLake displays nearly the change in performance, Ryzen slightly more, but almost within margin of error. Which makes sense. There is something in IF offering more bandwidth or lower latency. But if it was solely in memory latency and not Memory bandwidth. The gains with Ryzen would be much higher. It isn't. Cross CCX latency and memory latency have been the boogeymen since the beginning. Problem is the fact that it is deficient on Zen it's been locked on to being the issue. But if that was really the case why would SL-X with nearly the same latency tank so much in comparison to SL or KBL? If core to core latency is at fault why isn't SL-X below Ryzen in gaming performance considering every Core has a 100+ns latency between each other compared to the 4+4 configuration with Ryzen?

 

[IRobot23]

Of course DRAM is very important for Coffee lake also.

 

[Topweasel]

Of course DRAM is very important for Coffee lake also.

The point being that Kaby saw nearly the same increase in performance, sure Ryzens was slightly larger, suggesting it wasn't latency as much as it was bandwidth. Sure some of this extra increase be by increasing IF bus speed. But it does suggest that IF isn't as responsible as people imply it is.

 

[PeterScott]

Expectations?

Moaning that AMD didn't do enough.

 

[Topweasel]

Expectations?

Moaning that AMD didn't do enough.

Probably. "OMG AMD is screwed because they had the perfect time to one up Intel and they failed".

 

[CatMerc]

I linked a DF video previously. Coffee Lake doesn't care much for RAM beyond 2666. Ryzen has a LOT more to gain.

Either way, even if it did gain from lower latency, that doesn't invalidate Ryzen's gains. Matter of fact is that from profiling, we know games are memory latency sensitive, and we can easily measure that Ryzen has higher memory latency. This is holding it back in games, I don't see any other way to look at it.

As long as this remains true, core for core, clock for clock, Ryzen will have a deficit in gaming.

 

[IRobot23]

I don't know why you guys like to use thread for purposes of fanboys. OP did good job, you could give him little more respect.

Anyway hopefully we will have more CPU improvements that we had.

 

[Topweasel]

I linked a DF video previously. Coffee Lake doesn't care much for RAM beyond 2666. Ryzen has a LOT more to gain.

Either way, even if it did gain from lower latency, that doesn't invalidate Ryzen's gains. Matter of fact is that from profiling, we know games are memory latency sensitive, and we can easily measure that Ryzen has higher memory latency. This is holding it back in games, I don't see any other way to look at it.

As long as this remains true, core for core, clock for clock, Ryzen will have a deficit in gaming.

That is silly though. We know Ryzen doesn't have the memory Latency an i7 has, so that means that it's Ryzen's issue? It also doesn't have the L3 configuration. It also doesn't have a ring bus. Hell it's CPUID isn't GenuineIntel, maybe that is the reason? I am all for figuring out what is causing the performance rift and having AMD fix that as long as it doesn't hurt them in another way. This "I know what the answer is even though I have never built a CPU, or developed a game, just saw a difference (and I'll admit there is a large amount of people that agree) and used that as fact".It is bothersome because it could be blinding people from the real issue.

I haven't checked out as many Coffee lake reviews as I probably should. I do know KabyLake within about 2% or so saw the same increases in performance as Ryzen when using faster memory. I don't know but have no reason to believe otherwise the CoffeeLake as KBL with 2 extra cores would behave any differently. As for your link I am not going to watch the whole thing (this reviewer isn't my style) but I skimmed through the video and didn't find a single test with the same CPU using 2 speeds of memory, can you give me the time stamp of when he did?

 

[CatMerc]

That is silly though. We know Ryzen doesn't have the memory Latency an i7 has, so that means that it's Ryzen's issue? It also doesn't have the L3 configuration. It also doesn't have a ring bus. Hell it's CPUID isn't GenuineIntel, maybe that is the reason? I am all for figuring out what is causing the performance rift and having AMD fix that as long as it doesn't hurt them in another way. This "I know what the answer is even though I have never built a CPU, or developed a game, just saw a difference (and I'll admit there is a large amount of people that agree) and used that as fact".It is bothersome because it could be blinding people from the real issue.

I haven't checked out as many Coffee lake reviews as I probably should. I do know KabyLake within about 2% or so saw the same increases in performance as Ryzen when using faster memory. I don't know but have no reason to believe otherwise the CoffeeLake as KBL with 2 extra cores would behave any differently. As for your link I am not going to watch the whole thing (this reviewer isn't my style) but I skimmed through the video and didn't find a single test with the same CPU using 2 speeds of memory, can you give me the time stamp of when he did?

Plenty of theories have been tested and either disproven or proved to be so minor they're of no real concern.
The Ringbus can't be the reason because within a single CCX, there's lower latency than the ringbus system, so a 4+0 configuration should do much better, but it doesn't. The L3 victim cache could be the reason, but I don't really see theoretically how that would cause the performance deficit we're seeing. Then again the pros and cons of caching policies is something I'm not super well versed in. CPUID can be hidden through a VM, no such thing is causing an issue.

As for when in the video it's shown: https://youtu.be/9f5JQrnOwTE?t=14m7s
Super minor, not at all like the graph I've shown from The Stilt's RAM testing.
The LL bars mean Low Latency, having very aggressive timings to showcase the effects of lower latency.