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

[DisEnchantment]

Speculate at will

 

[jdubs03]

This guy from Canada keeps selling 9900Xs for under $400 US. I was tempted, but looks sus as hell. What do yall think?

AMD Ryzen 9 9900X - Ryzen 9 9000 Series Granite Ridge (Zen 5) 12- Core 4.4 GHz | eBay

Find many great new & used options and get the best deals for AMD Ryzen 9 9900X - Ryzen 9 9000 Series Granite Ridge (Zen 5) 12- Core 4.4 GHz at the best online prices at eBay! Free shipping for many products!

www.ebay.com

Seller doesn’t accept returns. Hard pass.

 

[jpiniero]

I would say at best they were illegally attained. "If it looks to good to be true...".

It "Fell Off the Back of a Truck"

 

[whoshere]

Not sus at all, https://www.ebay.com/usr/dizord-39?_tab=about

Member since: Jun 01, 2024

Wait a few months and buy it off Amazon/NewEgg for the same price if not lower.

 

[Heartbreaker]

Seller doesn’t accept returns. Hard pass.

Especially since Fake CPUs are out there. Der8auer did a video just today:

 

[inf64]

Computerbase has some preliminary benchmarking data with the new updates:

Benchmarks: Windows 11 23H2 + KB5041587 & 24H2 HVCI an/aus

Explodiert die Leistung von AMD Ryzen mit Windows 24H2 und Update KB5041587 für Windows 23H2 im CPU-Limit in Spielen – oder nicht?

www.computerbase.de

 

[igor_kavinski]

Computerbase has some preliminary benchmarking data with the new updates:

Benchmarks: Windows 11 23H2 + KB5041587 & 24H2 HVCI an/aus

Explodiert die Leistung von AMD Ryzen mit Windows 24H2 und Update KB5041587 für Windows 23H2 im CPU-Limit in Spielen – oder nicht?

www.computerbase.de

The results seem to suggest that 24H2 has a newer kernel or something coz the backported patch goodness isn't as good as the 24H2 goodness.

Dang it. Now Microsoft is keeping good stuff away from regular users while those with the latest Ryzen AI laptops get it first. I don't want to subscribe to some dang insider channel to get the good stuff.

 

[biostud]

The results seem to suggest that 24H2 has a newer kernel or something coz the backported patch goodness isn't as good as the 24H2 goodness.

What are you seeing that I'm not?

 

[igor_kavinski]

What are you seeing that I'm not?

Look at the results for F1 24 and Outcast. There could be other games like these.

 

[Det0x]

The results seem to suggest that 24H2 has a newer kernel or something coz the backported patch goodness isn't as good as the 24H2 goodness.

Dang it. Now Microsoft is keeping good stuff away from regular users while those with the latest Ryzen AI laptops get it first. I don't want to subscribe to some dang insider channel to get the good stuff.

You can get a iso for the newest version here, without any accounts of any kind

Windows 11, version 24H2 (26100.1591) amd64

Select language for Windows 11, version 24H2 (26100.1591) amd64

Select language for Windows 11, version 24H2 (26100.1591) amd64 on UUP dump. UUP dump lets you download Unified Update Platform files, like Windows Insider updates, directly from Windows Update.

uupdump.net

 

[MS_AT]

Computerbase has some preliminary benchmarking data with the new updates:

Benchmarks: Windows 11 23H2 + KB5041587 & 24H2 HVCI an/aus

Explodiert die Leistung von AMD Ryzen mit Windows 24H2 und Update KB5041587 für Windows 23H2 im CPU-Limit in Spielen – oder nicht?

www.computerbase.de

Nice, detailed setup. I wish they mentioned the BIOS settings by name too. I also hope for 9950x/9900x they will provide information if core parking was used or not. One could wish that such detailed setups would be a norm going forward, I mean every review containing detailed BIOS + Windows settings listing in addition to drivers used + version of monitoring tools used + versions of game / software that was tested. That at time would allow to avoid some internet drama over differing results...

 

[marees]

In case anyone has not seen this:

https://twitter.com/x/status/1830515620644815267

 

[Josh128]

In case anyone has not seen this:

https://twitter.com/x/status/1830515620644815267

So how does Windows 10 fit into all this? I cannot access the core isolation function in my system, says unavailable. Is Win10 by default at the top of this list, or the bottom?

 

[marees]

So how does Windows 10 fit into all this? I cannot access the core isolation function in my system, says unavailable. Is Win10 by default at the top of this list, or the bottom?

Run win10 in super admin mode, I guess !?

 

[eek2121]

So how does Windows 10 fit into all this? I cannot access the core isolation function in my system, says unavailable. Is Win10 by default at the top of this list, or the bottom?

presumably slower than win11 with the patch.

 

[Josh128]

presumably slower than win11 with the patch.

Thats not at all a given. See video from August 15 below.

 

[DZero]

This guy from Canada keeps selling 9900Xs for under $400 US. I was tempted, but looks sus as hell. What do yall think?

AMD Ryzen 9 9900X - Ryzen 9 9000 Series Granite Ridge (Zen 5) 12- Core 4.4 GHz | eBay

Find many great new & used options and get the best deals for AMD Ryzen 9 9900X - Ryzen 9 9000 Series Granite Ridge (Zen 5) 12- Core 4.4 GHz at the best online prices at eBay! Free shipping for many products!

www.ebay.com

Definately a scam. Better to report now.

 

[inf64]

Another mega test of the 24H2 update impact:

 

[Panino Manino]

The AMD Zen 5 Gaming postmortem: Larger generational gains than many reported, game-boosting Windows Update tested, Ryzen 5 7600X3D gaming benchmarks, too

Objects in mirror are larger than they appear.

www.tomshardware.com

Yes, as Tom's says, the performance is better than what we initially thought, after all that was done to "fix" it, but the performance is still lower than what AMD promised, isn't it?
IMO the disappointment remains.

 

[SteinFG]

The AMD Zen 5 Gaming postmortem: Larger generational gains than many reported, game-boosting Windows Update tested, Ryzen 5 7600X3D gaming benchmarks, too

Objects in mirror are larger than they appear.

www.tomshardware.com

Yes, as Tom's says, the performance is better than what we initially thought, after all that was done to "fix" it, but the performance is still lower than what AMD promised, isn't it?
IMO the disappointment remains.

The difference to previous gen is still small after all all those updates, 5-6%. That's equivalent to like, 230 MHz clock increase on some 5.5 GHz part. Worth something like 30-40 extra dollars, but not much more. Currenly 9700X is 80 dollars more expensive than 7700X.

 

[Det0x]

Another mega test of the 24H2 update impact:

Pretty good test from Toms this time 👍

Performance increase from the windows update (do note lower than other outlets since they tested with VBS disabed from the get-go)

Additionally, many say that the Windows update boosts AMD’s performance not only because of branch prediction optimizations but also because of a correction that reduces the overhead of Meltdown mitigations for a security vulnerability. We asked AMD, and the company refused to comment, which should tell you all you need to know — they could have simply denied it but didn’t. Technically, an ‘improvement to the branch prediction unit’ would completely align with reduced overhead from a mitigation because Meltdown attacks speculative execution functions. That means that larger gains could be seen when testing with the performance-sapping virtualization-based security (VBS) feature turned on.

However, because Microsoft has posted instructions on how to disable the feature if you prize gaming performance more than security, we always conduct all testing with VBS off to extract the full performance potential of the chips. So, you could see a few more percentage points of performance gains based on your configuration.

Sadly there is no benefit for Intel with the new windows update


End result

We wrap up here with only the results using the new version of Windows. Note that due to time constraints, the Intel chips are tested here at true stock memory settings, whereas the AMD chips have Expo memory overclocking enabled, so Intel will perform better with tuned memory. I’ll add the overclocked memory results for Intel in an update to this article. We also have the as-yet-unreviewed Ryzen 5 7600X3D, a Microcenter exclusive, in the test pool, but this chip is only tested with Expo memory settings.

The Ryzen 5 7600X3D with Expo overclocked memory settings delivers about the performance we expect, beating out the $530 Intel Core i9-14900K flagship with stock memory by a small amount. That’s impressive for a $300 chip, but it does come with drawbacks in application performance. We’ll post a review in the coming days to suss out the true differences, but this is a great first showing for the chip.

The Ryzen 7 7800X3D continues to dominate our gaming benchmarks easily, while the Ryzen 7 7700X with Expo enabled beats the 14600K but still lags behind the bare stock 14900K config (remember, this is a proxy for the 14700K as well, which would be about 2% slower than the 14900K). The Ryzen 5 9600X has pulled into parity with the 14600K, but the latter would take the lead with tuned memory.

Even after the Windows update and with Expo memory overclocking removing Zen 5’s memory speed advantage, we still see decent generational gains for Zen 5 over Zen 4. Measuring across our entire test suite, the Zen 5 Ryen 5 9600X is 9% faster than the prior-gen Ryzen 5 7600X, while the Ryzen 7 9700X is 7% faster than the prior-gen Ryzen 7 7700X.

The big takeaway here is that, regardless of memory speeds, the type of benchmarks used, or whether the Windows patch has been applied, the generational gains for Zen 5 gaming are larger than we’ve seen in many reports. Whether or not that justifies purchasing one of these chips over the Intel comparables is a different matter, but we’ll update our reviews with updated buying advice over the coming days.

 

[MS_AT]

Pretty good test from Toms this time 👍

Performance increase from the windows update (do note lower than other outlets since they tested with VBS disabed from the get-go)

Hardware Unboxed is also testing with disabled VBS

Sadly there is no benefit for Intel with the new windows update

That at least shows consistent messaging as both rumors and official communication suggested Intel should see nothing as the issue was, if we believe rumors, with Zen/Zen2 mitigations being applied to all Zen cpus.

 

[DisEnchantment]

Speculation time,

gcc.gnu.org Git - gcc.git/blobdiff - gcc/config/i386/x86-tune-sched.cc

+ /* For znver5 decoder can handle 4 or 8 instructions per cycle,
+ op cache 12 instruction/cycle, dispatch 8 instructions
+ integer rename 8 instructions and Fp 6 instructions.
+
+ The scheduler, without understanding out of order nature of the CPU
+ is unlikely going to be able to fill all of these. */
+ case PROCESSOR_ZNVER5:

This patch is really cryptic. Z5 is 4 wide or 8 wide decode?

It currently is 4 wide x2, so there are some chips that are really 8 wide? Is there some issue with the current stepping that they deactivate the chicken bits?

 

[Nothingness]

Speculation time,

gcc.gnu.org Git - gcc.git/blobdiff - gcc/config/i386/x86-tune-sched.cc

This patch is really cryptic. Z5 is 4 wide or 8 wide decode?

It currently is 4 wide x2, so there are some chips that are really 8 wide? Is there some issue with the current stepping that they deactivate the chicken bits?

I don't think the gcc guys know more than us. They base their changes on the optimization guide.

 

[zeropride]

The difference to previous gen is still small after all all those updates, 5-6%. That's equivalent to like, 230 MHz clock increase on some 5.5 GHz part. Worth something like 30-40 extra dollars, but not much more. Currenly 9700X is 80 dollars more expensive than 7700X.

It beats zen4 in workloads and ai stuff.

 

[ikjadoon]

What's points per GHz got to do with it? Different architectures have different ways to get the same answer. Points per watt I'd understand. Otherwise just rank by performance.

As others have rightly noted, AMD / Intel / Arm calculate generational IPC uplift with Geekbench 1T. Basically every "IPC" relative uplift claim includes either or both GB5 and GB6. That different CPUs use different frequencies is an orthogonal discussion.

Overall perf is still critical and I always keep it in the chart. Ranking by Pts / GHz is not a serious choice; it simply makes the chart easier to read (e.g., newer CPUs are near the top, by a given firm). Notebookcheck doesn't always test the highest-end SKU, either, so it's less helpful to rank by perf alone.

But it might make sense to only share AMD CPUs in the AMD thread.

Sources that GB is an important "IPC" data source for basically the entire industry:


(of course, we'd all object to cherry-picked sub-tests, like the GB5 crypto tests)





Anyways,