Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

[DisEnchantment]

Speculate at will

 

[Cllaymenn]

 

[StefanR5R]

Will just running games in Admin mode (click "run as administrator) help, or do you need to ran from that hidden super admin account?

it's sufficient

Link?

e.g. computerbase.de, translated:

The same effect as described above by activating the administrator account can also be achieved more easily. All you have to do is use the "run as administrator" option, which can be found by right-clicking on the game's shortcut. Hardware Unboxed points this out in a tweet and also explains that many CPUs (including Intel) will potentially deliver better performance as a result, as this is not a bug on AMD systems, but rather a Windows feature. This behavior can also be observed on Windows 10.

edit, original HWUB tweet:

Just a quick update for you on the Windows 11/Zen 5 testing, as I don't think I will get a video out on this over the next few days.

- The same gaming performance uplift using the administrator account is seen when using Windows 10, so this looks far less likely to be a bug as AMD suggested, and just the way Windows works.
- If you run the game "run as administrator" you will see the same performance uplift as seen in the system administrator account, so that's a much easier way to test this :D
- I'm still yet to test other CPUs but all indications point to this performance uplift applying to any and all CPUs. Trusted members in our Discord server have report similar findings with Intel CPUs for example.
- Changing the application priority level from low, medium and real-time doesn't change performance.

I'll keep testing when I can, but there is some other content I want to get done first, such as Black Myth Wukong GPU Benchmarking for example.
Thanks guys.

 

[coercitiv]

I can already see the outcry on social media about artificial segmentation to milk customers

There would be smarter ways of doing this without sparking any outcry, for example making the support mandatory on high-end chipsets, optional on B and A. We already have something similar for PCIe support.

That being said, power delivery is just one of the problems with higher core counts. The second one was already discussed to death in the past few days > mem bandwidth. I don't want to rehash a discussion which sparks on this forum every 6 months or so, the main takeaway is we always end up in the same spot: some prosumers would like to use the mainstream platform as HEDT replacement for their jobs, because Threadripper is also a Walletripper.

 

[StefanR5R]

Running games as administrator:
I haven't been following this, because I am not interested in video games myself. Now having read up on it, I am disturbed that AMD¹ made this suggestion to a youtuber (¹ according to this youtuber). From there, the news that entertainment software runs better with elevated privileges appears to spread like wildfire, well illustrated for the laymen by longer bar graphs. Teaching Joe and Jane to "solve" privilege separation related issues on their home computers through privilege escalation is a… questionable move.

 

[igor_kavinski]

Teaching Joe and Jane to "solve" privilege separation related issues on their home computers through privilege escalation is a… questionable move.

Next update, Microsoft needs to show some popup message to home users saying something like "Admin mode may expose you to serious security related issues. Tread with caution!".

Otherwise, this could be used as an attack vector by malicious game authors (particularly but not limited to, the Chinese).

 

[Heartbreaker]

Running games as administrator:
I haven't been following this, because I am not interested in video games myself. Now having read up on it, I am disturbed that AMD¹ made this suggestion to a youtuber (¹ according to this youtuber). From there, the news that entertainment software runs better with elevated privileges appears to spread like wildfire, well illustrated for the laymen by longer bar graphs. Teaching Joe and Jane to "solve" privilege separation related issues on their home computers through privilege escalation is a… questionable move.

I don't have an issue with using "Run as Administrator" for the occasional game from trusted sources. Some older games I have require "Run as Administrator" just to function at all.

But as it was first described, running in the secret Admin account, was a terrible, very risky idea... That elevates everything you run.

 

[positivedoppler]

So far, it has been a Jekyll & Hyde story for zen 5

Zen, 5 excuses vs zen 5$/day crypto mining

https://twitter.com/x/status/1824693836230345079

AMD Ryzen 9 7950X used by cryptominers, making $3 per day in profit from their CPUs

AMD's current-gen Ryzen 9 7950X processor is using used for cryptomining, as it's more profitable than GPU mining while Bitcoin rides new highs.

www.tweaktown.com

T
Annoying when hwu casually dismiss their viewer questions or points with a mere "nope"

 

[Heartbreaker]

T
Annoying when hwu casually dismiss their viewer questions or points with a mere "nope"

They have covered some of those in great detail. Which one do you think deserves more than a "nope" at this point?

 

[PJVol]

Hmm, do you have a pointer to a description of this functionality? Was this one for CPU initiated accesses perhaps? SDCI is for device initiated accesses, the device being the writer.

I can't recall the exact doc right now, but the underlining technical base certainly was the L3 partitioning. But you can "google" something like RAM disk on Genoa-X V-cache. As for the "initiator question", may be certain L3 slices in this case are seen as the device with "dram" MTR value, or at the IF layer, one of the SDP ports could be configured as "coherent slave" for global visibility. Honestly, idk.

 

[tsamolotoff]

Hmm, initially i did not observe any difference between 'true' admin and normal admin in cp2077 integrated benchmark, but now there is some extra performance (although this time i've tested it with low settings instead of high settings, could be that the actual "game cpu"-limited part (which is where the fps is lowest, somewhere after the camera flies out of some alleyway into a plaza) doesn't really change with switching to admin account, but the part where fps depends on how fast the driver can process frames actually does, idk.

 

[positivedoppler]

They have covered some of those in great detail. Which one do you think deserves more than a "nope" at this point?

I have not watch all their videos. There are already tons of content on the 9000 series. I view their tweet as condescending and dismissive so I wont bother with the rest of their videos.

 

[fastandfurious6]

Guru3d is expecting some type of X3D announcement at Gamescom on Aug. 23.

Review: Ryzen 9 9950X processor - Updated with KB5041587 for Gaming Charts (Page 29)

We review the ZEN 5 processor series, highlighting the powerful 16-core Ryzen 9 9950X. This model boasts the highest number of cores for consumers and operates at the fastest speeds. We're moving beyond high-end to the enthusiast tier of processors.

www.guru3d.com

gamers rise up!!!! ✊✊✊✊✊

 

[Cllaymenn]

Sorry, folks. I created the table for the ES 9950X DDR5-8000 configuration in LibreCalc and I guess I don't know how to make formulas work flawlessly in that AND I was in a hurry to post.

Here's the updated percentage improvement table and I hope I got it right this time:

View attachment 101856

So looking at that, maybe the ST perf isn't that hot but in MT, it's a beast?

If you are perplexed by these numbers, you are not alone

Any mistakes are all mine!

Does anyone still remember those results from 6-7 weeks ago that appeared here in this thread from a igor_kavinski friend who had access to the ES 9950X?

I wonder if these results nowadays would be confirmed in the store version? From these engineering sample tests on the older AGESA it appeared that 9950X is AES - 96%, FP32 - 74.5%!, FP64 - 77% faster than 7950X! (and 140%+, 190%+, 188%+ than 13900K!)

Can any 9950X owner here verify/check this 3 results AES, FP32 and FP64 in new AIDA64 (this take 2min)?

Logically thinking more perfected final store version of 9950X + newer AGESA should have results even better

 

[Heartbreaker]

I have not watch all their videos. There are already tons of content on the 9000 series. I view their tweet as condescending and dismissive so I wont bother with the rest of their videos.

Your loss.

 

[Markfw]

gamers rise up!!!! ✊✊✊✊✊

Considering the most people were upset about Zen 5 was the gaming performance . This should make that and other concerns go away. They already rule on productivity.

 

[linkgoron]

Oh yeah Bulldozer was a marketing failure. Calling their modules dual core was a failure and it cost them. And rightly so.

Bulldozer was an engineering failure that no amount of marketing could save. Yes, marketing failed of course, but in the end it was in many cases slower than Phenom II and ST performance was abysmal.

Spoiler

What can you do when Intel is 45% faster...

RDNA3, for example, could have been released in a much better way. They should've used the 7800XT moniker, they should've given it a slightly lower price (especially the 7900xt). Really, against the 4080 it's pretty competitive (sans RTX), but bad pricing and confusing naming just ruined the 7900 line. Oh, and last but not least, they shouldn't have lied in the launch presentation performance numbers...

They messed up in this launch as well. However, Ryzen 9xxx are still probably the best x86 consumer CPUs in the world. They had to deliver better pricing. 5-10% performance lift for better TDP won't cut it with the current prices. Perf/watt isn't even that much better than Zen4 non-x SKUs. I don't see many people getting a Ryzen 9000 instead of a 7000 for more money, and the zen4 x3D variants are still by far the best for gaming. At this point, you've just got to ask yourself how/what went terribly wrong with Zen5. As the awesome insiders in this forum have been so spot-on with Zen5 performance numbers, they might be able to shed some more light in the future on what went wrong with Zen5.

gamers rise up!!!! ✊✊✊✊✊

x3D is both AMD's doom and hope. I think that the 7800x3D/7950x3D make Ryzen 9000 look much worse than it probably would, but the Ryzen 9000x3D lineup might also be its saving grace if they manage to bring it quickly.

 

[Ranulf]

Hardware unboxed mentioned I think in a Twitter post that some of their discord members were reporting that they see the same thing in Windows 10 as well.

Their tweet:

https://twitter.com/x/status/1824619686874714522

 

[StefanR5R]

At this point, you've just got to ask yourself how/what went terribly wrong with Zen5.

Something went terribly wrong? That's a rather selective view on the matter.

5-10% performance lift for better TDP

This too.

won't cut it with the current prices.

They are launch prices. Ryzen 7000 supply still seems to be in the channel, let's see for how long.

 

[HurleyBird]

Everybody keeps denying the change in avx-512 applications. First it was in PURE avx-512 benchmark show 98% faster than Zen 4. Now we have at least one application where it is 30-40% faster that is actually used (in the application areas). See the benchmarks in the DC forum here. These are not imaginary gains. In scientific and other areas, it should see these huge gains. This being the case, I doubt that it will die at Zen 5.

Not denying it. But I doubt AMD went through the trouble of full and half speed AVX-512 on Zen 5 just to abandon the concept after one generation. And it's likely that mobile and desktop converge quite a bit in Zen 6.

What I really want to see from AMD, especially if the server parts start to diverge with better features, if the return of Threadripper. But I'm still totally turned off from the whole "we'll selectively release parts sometimes, sometimes very late, and sometimes we'll just skip generations and leave the segment to lie fallow" from before. The only way I would ever buy into a future TR platform is if AMD promised to pursue a specific cadence and swore off their previous behavior.

 

[Markfw]

Not denying it. But I doubt AMD went through the trouble of full and half speed AVX-512 on Zen 5 just to abandon the concept after one generation. And it's likely that mobile and desktop converge quite a bit in Zen 6.

What I really want to see from AMD, especially if the server parts start to diverge with better features, if the return of Threadripper. But I'm still totally turned off from the whole "we'll selectively release parts sometimes, sometimes very late, and sometimes we'll just skip generations and leave the segment to lie fallow" from before. The only way I would ever buy into a future TR platform is if AMD promised to pursue a specific cadence and swore off their previous behavior.

You forget the Server is the real moneymaker. desktop is a pittance compared to server, thats why they base desktop from server chipps binned differently. That also saves money. Ask yourself. Right now who is doing hands and feet above the other, and making money instead of bleeding money ? Companies are there for one reason, $$$$$, nothing else.

Is the company a success ? Yes. The best in server ? YES. do they have the best in desktop ? Yes(by a small margin) mobile ? close

And when Turin comes out, they will be so far ahead on server, intel will take ages to catch up.

 

[DavidC1]

Bulldozer was an engineering failure that no amount of marketing could save. Yes, marketing failed of course, but in the end it was in many cases slower than Phenom II and ST performance was abysmal.

Research and Development often needs to try out moonshot architectures and ideas to see if they can at least be salvaged for future use. What is a constant struggle(in life too) is finding the right balance, not too hot and not too cold. That's the real challenge. In case of Bulldozer they shot too far in one direction, a wrong one.

If AMD benefitted from it and resulted in Ryzen, in the long term it would merely be a lesson.

 

[soresu]

If AMD benefitted from it and resulted in Ryzen

May be splitting hairs here but Ryzen is the brand (like Radeon or Instinct).

Zen(n) is the µArch.

Phenom did not persist as a brand past Thuban, Bulldozer temporarily resurrected FX and simultaneously killed it due to associated bad mojo, but Athlon remains I think as an APU budget brand, or did until fairly recently.

 

[DavidC1]

May be splitting hairs here but Ryzen is the brand (like Radeon or Instinct).

Zen(n) is the µArch.

Phenom did not persist as a brand past Thuban, Bulldozer temporarily resurrected FX and simultaneously killed it due to associated bad mojo, but Athlon remains I think as

Let me try a second time.

If AMD benefitted from it and resulted in Ryzen.

 

[biostud]

Their tweet:

https://twitter.com/x/status/1824619686874714522

So probably bypassing all kinds of windows security etc. when running in elevated mode freeing ressorces for the games.

 

[zir_blazer]

Not "extra memory" but more efficient paging. Normally, all memory on x86 windows is accessed through 4kB pages. If you want to access 4GB of space, you need to set up a million PTEs, which is a problem because the TLB can only cache a few thousand. The hardware also supports 2MB pages, which are a lot more reasonable, and on zen use the same tlb entries so the cache can cover 6GB on Zen 4.

You forgot 1 GiB Huge Pages, supported since Nehalem.
The only software that I used them in was QEMU. In general performance increase by using static 1 GiB Huge Pages is not great (About 1%) compared to Transparent Huge Pages (2 MiB), and they require to be reserved at boot time, but it is free performance. Can't find the QEMU benchmarks...