AMD Addresses Ryzen 9000 Performance | The Full Nerd Special Edition

Join The Full Nerd gang as they talk about the latest PC building news. In this episode the gang interviews David McAfee, Corporate VP & GM of Client Channel Business at AMD, about the Ryzen 9000 launch, benchmarking numbers, gaming performance and more. And of course we answer your questions live!

Links:

Check out the audio version of the podcast on iTunes, Spotify, Pocket Casts and more so you can listen on the go, and be sure to subscribe so you don't miss the latest live episode here on YouTube!

Follow the crew on X: @GordonUng @BradChacos @MorphingBall @AdamPMurray

=============
Follow PCWorld!
=============

This video is NOT sponsored. Some links may contain affiliate links, which means if you buy something PCWorld may receive a small commission.

#podcast #amd #pc

[arcayne]

Thanks to the Nerds and PC World Team for doing this.

That said, I am happy that I missed this one live. I would have been flooding the chat every minute with:

AMD: Admit you were wrong. Accept responsibility. Apologize for it. Commit to doing better in the future. Own it.

Done. Less than 30-seconds of air-time and no endless excuses and deflections. Zen5 is not what was communicated by AMD, and it is not going to be what was communicated by AMD. Deal. With. It.

[ricardolmendes]

Pure damage control for a poorly managed release. I honestly don't understand how the marketing team and reviews support teams could fundle things this badly. Zen5 is a good uarch for certain scenarios but the message amd passed was so mixed and wrong that only created confusion. It really is unfortunate and I hope AMD learned their lesson to not repeat this fiasco.

[TheSickness]

Too bad he didn't talk publicly to Hardware Unboxed, they already tested the "windows improvements" to also up performance on Ryzen 7000, just 9000 gets 1-3% on top. Don't be gaslight

[Takashita_Sukakoki]

So it comes back to AMD´s bad communication did the review guide not specify OS version??? Testing an unwidely available OS version so bizarre. Also running a in-game benchmark (in their automated benchmarks) is never a good CPU benchmark. Win 10 owners can kick rocks?

[CpuWaiy]

As consumers, w3 have the ultimate power. Stop buying tech when it doesnt feel right and companies will fall in line.

[Eternalduoae]

I realise the difficult position AMD is in but they need to stop doubling down on their deceit. Even in the interview, the gentleman has a lot of weasely words around the uplift and testing methodologies. I'm a bit disappointed that you guys didn't press him on the uplift of Zen 3/4 on those "double digit" improvements for Zen 5.

If it wasn't deceit on purpose then AMD:

1) Doesn't have a robust testing procedure.
2) Isn't interested in having a robust testing procedure.

There are plenty of ways of automation without using an unrealistic testing environment.

The end result of this entire period (even before Zen 5 release) is that AMD's marketing cannot be trusted even slightly.

This was not the case a few years ago. They've lost my trust entirely and the community blog post, which tried to shift blame to reviewers and, again, be weasely about the increases when putting all products in the comparable testing environment, and this interview, hasn't made the situation better.

[jellowiggler]

Just imagine if they would have launched it on time. It REALLY would have been messed up. This is the polished and ready version of the Zen5 launch. ouch.

AMD marketing needs to stop with the misinformation. They only undermine company credibility. Poor marketing not supported by their own testing let alone independent testing, the questionable benchmarking methodology at the fundamental level, not waiting for scheduler patches on their primary platform, delayed launch for QA 'issues' etc, etc, etc.

[douglasmurphy]

Am I to understand that the branch prediction gains in 24H2 are separate from the security overhead (admin) gains? Has anyone with early access to 24H2 been able to put up their own benchmarks?

[j_official]

Don’t blame gaming channels for being disappointed with Zen 5% when AMD itself published slides leading up to the launch that claimed performance uplifts that turned out to be false. With those slides, and a price premium over zen 4 that was well above the minuscule performance increase, AMD set itself up to fail. And it’s a self own, given Intel’s issues with raptor lake degradation had reached fever pitch. AMD zen 5% came along and deflated that Intel rage balloon.

Also, why did AMD need to launch zen 5 in summer 24 if the windows code regarding branch prediction in Win11 24h2 is unavailable in summer 24? AMD could’ve taken its time, worked with MS behind the scenes to publish the code, ensure reviewers are aware of the new code and ensure they’re using it, and also AMD could’ve worked on the firmware/microcode/agesa/bios for stability and so on. AMD had an opportunity when Intel was flailing in the wind and then squandered it in my opinion.

[spdcrzy]

Am I the only one that can see this being entirely in good faith? From a semiconductor engineering POV, I can see how a standard operating procedure (super admin mode) becomes almost tribal knowledge in the sense that there is a boiling frog effect. At an organizational level, that's what happened here.

[LeesChannel]

I was a little late on the live stream, but my question as I was watching: if AMD had no expectation that there was/would be a performance discrepancy, then why was there a delay on the launch of the product?

[jellowiggler]

Why the hell would you do your lab validation on a version of Windows that people can't use? Funny how they didn't mention that, and don't mention when this version will be available. Not only that, but all modern processors will get the uptake. So it probably still won't be the gains stated over older and competitor's systems.

There is no reason to advertise performance in a mode nobody can use, on a version people can't run. "Looks guys, our processor is this FAST! But we don't suggest anyone run this way, it's not safe!"

Are they going to update their marketing material and correct their claims for the product as it ships and how people can run it today, when they buy and test it?

Snake oil and excuses.

They are grasping at anything to excuse their behaviour. But you don't see them issuing corrections, or directly addressing the contradictions in their own literature re: performance.

That's OK AMD. Those who care to know, do know. Pretty much on day one. If you can be disproved in a few days by the community testers it means you did your validation wrong. You ended up with a chaotic very poor launch that underwhelmed compared to your claims by half. Same thing happened with the last gen Radeon. Are you trying to be known as the company who delivers half? The HW community will test how we will use the product. If you decide to fabricate some other test conditions and we don't see the matching results, then it makes you the problem and you will suffer for it. Why not just test how your customer will use the product on the systems they will use them with?

[benkleban]

"Our largest competitor is taking a complete thrashing in the media the past 2 months, So lets point this gatling gun directly at our feet."

[wireless]

It sounds like AMD came up with their numbers by gimping the Intel numbers through the use of a lower power profile and boosting their own figures by running in super admin mode with VBS off. They cherry-picked data, intentional or not.

[threadripper]

Nope, not buying it. And I'm a hardcore AMD fanboy. Do the interview with a panel consisting of Wendel, HUB, and GN. There would be a completely different outcome.

[AK-Brian]

It would be interesting to see if AMD's stated 5-8% average uplift (in the canned benchmarks within their test suite) rings true if tested with an RX 7900XTX rather than an RTX 4090. Their statement didn't address specifics of their test setup (other than 6000MT memory), and this may be another difference between their configuration and those of most reviewers.

It would not absolve them, but if found to be a contributing factor, may at least help explain how those original numbers were obtained.

[Kocan]

Thanks for the interview, interesting stuff, hopefully they will iron out those problems before 3d launch.
Honestly, this is very interesting after PCWOrld, HUB, GN, L1T and MLID videos, it all starts coming together.

[tomrobinson]

Some up 3 guys watch a man waffle on for a hour.

[tappy]

Interesting, I thought they dropped support for windows 10 already but it's supported until next year.

[hbdude]

AMD if full CYA mode, understandable cuz aint nobody buying these chips. literally and figuratively