AMD Ryzen 9 9950X

[Dr. Dro]

Can someone please remind me why AMD needs this "Xbox game bar" nonsense for their dual-CCD CPUs to work properly, when Intel doesn't?

Because Intel has a hardware thread scheduler and AMD doesn't. The 3D Cache service uses the Xbox Game Bar to detect games and communicate this to a kernel mode driver which alters the CPU's behavior. Alder and Raptor Lake can do this transparently in hardware, and will know by itself that it should place complex workloads onto the Golden/Raptor Cove cores and issue lighter workloads to the Gracemont cores.

This is a grave deficiency of the X3D hybrid design and I fail to understand why does Zen 5 needs this at all if not for working around some aggressive form of core parking or power saving mechanism.

 

[terroralpha]

Hi again, you don't think the 12th, 13th, and 14th, well primarily the 14th generation intel processors will benchmark higher if you use higher frequency memory than the one you been using? I know we've talked about this, but it was in regards to AMD cpu's. I think if you use higher frequency memory for the 12th, 13th, and 14th generation intel processor's they will give more higher results. The thing is with 12th generation maybe 6400MHz, with 13th, something around 6800Mhz. The 14th generation can scale higher frequency memory for better results. I know you know this.

i use to have a 12900k. mine would not work with anything faster than 6200MHz. i was able to boot with memory at 6400MHz, but it just wasn't stable. i would get application hangs, occasional crashes. it was just a terrible experience all around. but my current 7950X3d runs a 96GB kit at 6400MHz just fine. a lot of gaming prebuilds from dell, lenovo, HP and others only have 6000MHz options. their BIOS simply won't allow you to use faster memory clocks.

the point i'm trying to make is that there will be people whose hardware would simply not be able to replicate results with high memory clocks because either their motherboard doesn't support it or the memory controller on their CPU can't handle it, and any benchmarks using these speeds would be even more useless than your comment. hard to believe, i know

 

[RogueSix]

Can someone please remind me why AMD needs this "Xbox game bar" nonsense for their dual-CCD CPUs to work properly, when Intel doesn't?

Intel has a hardware scheduler called 'Thread Director' and they have worked a lot with MS to tune their Thread Director with the Win 11 scheduler.

I don't get why AMD has not copied the principle yet. Thread Director is not exactly new. AMD should have long since introduced hardware scheduling as well.

 

[tfp]

Expect that contradicts what the Zen engineer said in the youtube interview with chips & cheese?

Chip & cheese is also contradicting his own interview in parts of his own review.

It totally is suppose to able to excute both pipelines decoders and predictors when running a single thread through the cores. Other wise all those parts just sits idle doing nothing predictor, decode, & pipeline.

Chips & cheese test he proved that it not doing what the engineer claimed it was supposed to do or could in the youtube interview. All three parta are idle when smt is disabled. It is only using one predictor, one pipeline, & one decoder.

I am just stating that Zen4 does not have 2 decoders per core. Zen5 mobile is not Zen4. If Zen5 decoders are working correctly or not based on an interview vs testing is something completely different.

 

[ARF]

The DDR5 throughput is quite low.

vs Core i9-14900K:

At least, the L2 cache throughput is quite high:

vs.:

Review: Ryzen 9 9950X processor - Updated with KB5041587 for Gaming Charts (Page 14)

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

Intel Core i9-14900K review (Page 15)

Join us as we review, analyse and test the Intel 14th Gen core processor, specifically the Raptor Lake-based Core i9 14900K, which has been officially released. Prior to this, there were numerous leaks and performance reviews surrounding the series. The latest model represents the advanced...

www.guru3d.com

 

[Mr_Engineer]

Those are good questions. For the charts I posted I had to go back to the setup and review but it's a bit sparse describing the setup of the Oracle VirtualBox but if I understand the test setup correctly it was a Win11 host and Win11 guest. I would think it wouldn't matter if the virtualization was a level 1 or level 2 hypervisor or the OS's involved. The potential of cross CCD latency would still be there in all situations unless the hypervisor and/or host OS was smart enough to prevent that. Perhaps I'm making a fuss over nothing but I wonder if just like for games the other tests indicated in red performed poorly for the same reasons? (edit) sorry for the word salad.

About Virtualization, it is highly critical for servers, and for sure, any regression in performance between Zen 5 and Zen 4 would never be intended by AMD.

The TPU review is testing Oracle Virtualbox, but most large scale enterprises use VMware for virtualization. I believe testing VMware (and maybe Hyper-V) performance on Zen 5 for virtualization would be more suitable (@W1zzard I think your testing is very thorough and awesome, please don't kill me ).

We can't just decide that VM performance sucks on Zen 5 because one app shows a regression. And it seems like some applications may need patching/recompiling to improve Zen 5 performance as there were many changes in the architecture.

 

[Assimilator]

Hi again, you don't think the 12th, 13th, and 14th, well primarily the 14th generation intel processors will benchmark higher if you use higher frequency memory than the one you been using?

W1zz has already answered this multiple times: this is not an exercise to see how high individual CPUs can benchmark, it's an exercise to compare different CPUs in as like-for-like a way as possible.

 

[Mr_Engineer]

The DDR5 throughput is quite low.

View attachment 359181

vs Core i9-14900K:

View attachment 359182

At least, the L2 cache throughput is quite high:

View attachment 359184

vs.:

View attachment 359183

Review: Ryzen 9 9950X processor - Updated with KB5041587 for Gaming Charts (Page 14)

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

Intel Core i9-14900K review (Page 15)

Join us as we review, analyse and test the Intel 14th Gen core processor, specifically the Raptor Lake-based Core i9 14900K, which has been officially released. Prior to this, there were numerous leaks and performance reviews surrounding the series. The latest model represents the advanced...

www.guru3d.com

Yes, memory bandwidth and latency always sucks on Zen (even previous generations) compared to Intel. This is a weakness that AMD still didn't improve with Zen 5 by keeping the same old IOD (big mistake).

And that's why X3D performance benefit a lot especially for games, as the largest L3 cache makes up for the slow IMC.

What is surprising to me is that games didn't benefit much on Zen 5 even with the cache bandwidth being double.

 

[W1zzard]

The TPU review is testing Oracle Virtualbox, but most large scale enterprises use VMware for virtualization. I believe testing VMware (and maybe Hyper-V) performance on Zen 5 for virtualization would be more suitable (@W1zzard I think your testing is very thorough and awesome, please don't kill me ).

Every single customer of VMWare is trying to move away from them after they got acquired by Broadcom and raised prices by like x100 or x1000 for some of their customers.

I have been using VMWare Workstation in the past for this testing, but ran into lots of issues in 2023, so I looked for an alternative. Hyper-V is integrated too tightly with the OS that it's a huge POS to control, Virtualbox just works, a single config file, you can move stuff around, no services/telemetry running in the background or through Task Manager

 

[Super XP]

AMDs new ZEN5 CPUs weren't ready for prime time. They should have waited till the software and new motherboards were "Officially" ready. So now AMD needs to play catch-up because these CPUs should be at a min 10-12% IPC over the previous gen.

ZEN5 suffers from Windows Scheduling issues. The same happened with ZEN4 when they first came out.

AMD did have an opportunity to really give it to Intel and all their major design issues. But nope lol, not yet I suppose.

 

[Zach_01]

The DDR5 throughput is quite low.

View attachment 359181

vs Core i9-14900K:

View attachment 359182

At least, the L2 cache throughput is quite high:

View attachment 359184

vs.:

View attachment 359183

Review: Ryzen 9 9950X processor - Updated with KB5041587 for Gaming Charts (Page 14)

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

Intel Core i9-14900K review (Page 15)

Join us as we review, analyse and test the Intel 14th Gen core processor, specifically the Raptor Lake-based Core i9 14900K, which has been officially released. Prior to this, there were numerous leaks and performance reviews surrounding the series. The latest model represents the advanced...

www.guru3d.com

And? Help me understand...
Ryzen CPUs always had AIDA scores lower than Intel ones. That didnt stop them from being faster overall in many cases.

What is the point of this apple-orange comparison other than showing a few arbitrary numbers that do not translate directly into overall performance. Let alone on different applications.
These may have some meaning when comparing same architecture.

Its like trying to compare shader count on 2 completely GPU architectures.
You will be driven to the wrong conclusion with mathematical precision if thats the only thing you are looking.

Yes, memory bandwidth and latency always sucks on Zen (even previous generations) compared to Intel. This is a weakness that AMD still didn't improve with Zen 5 by keeping the same old IOD (big mistake).

And that's why X3D performance benefit a lot especially for games, as the largest L3 cache makes up for the slow IMC.

What is surprising to me is that games didn't benefit much on Zen 5 even with the cache bandwidth being double.

You mean the bandwidth of L3?
But you said it right above it. What good is the wider path if the "source" (IOD) cant feed it enough.
Enter the extra cache for games...

Same IOD = lower cost, better margins and mainly opens the door for X3D parts to be more relevant than ever (=more profit). Sure... sales are not looking good for these "regular" 9000 but 7000 is in the mix and 3DVcache is coming. I can already see in the future a 9500X3D for "entry" level gaming.
I've said it again (cant remember if her or another thread, confused by so many discussions... lol) that these chiplets will be also on EPYC and there we dont know what kind of IOD AMD has cooked.
On desktop? things will pick up

AMDs new ZEN5 CPUs weren't ready for prime time. They should have waited till the software and new motherboards were "Officially" ready. So now AMD needs to play catch-up because these CPUs should be at a min 10-12% IPC over the previous gen.

ZEN5 suffers from Windows Scheduling issues. The same happened with ZEN4 when they first came out.

AMD did have an opportunity to really give it to Intel and all their major design issues. But nope lol, not yet I suppose.

You can say that. AMD is known for rushed releases. Or maybe its ready as can be for the EPYC.
In the end if you think about it, no matter how much testing is happening on a close environment the real test is out there...
You can also say that nothing is over yet. This so far was half (or less?) the Zen5 parts

And we really should want Intel to stay relevant in the industry of CPUs, so its not all that bad not striking hard on Intel... yet...

 

[AVATARAT]

@W1zzard did you try to stop core parking? As AnandTech mentioned, there may be a problem with it.
(With something like ParkControl?)

 

[Caring1]

Faster Ram is not better.

 

[Mr_Engineer]

You mean the bandwidth of L3?
But you said it right above it. What good is the wider path if the "source" (IOD) cant feed it enough.
Enter the extra cache for games...

Sorry I should have been more clear. I meant the bandwidth and latency of the memory controller for Zen.

For example, if we use the same DDR5-5200 memory with the same timings on Ryzen or Intel processors (any processor from the last few years) the bandwidth would be higher and latency would lower on the Intel platforms. The memory controller on Ryzen has always been weak, and AMD didn't try to improve it for Zen 5.

 

[AVATARAT]

Faster Ram is not better.

Did you watch the video that you link?
Or if you can explain what you mean?

 

[Zach_01]

Faster Ram is not better.

Yeah its should be common knowledge by now that decoupling UCLK:MCLK is creating more latency than what increased MT/s can compensate.
The 7800MT/s crippled memory controller with 2067:1950:3900 speeds.
What's new?

Who ever wants to squeeze performance is to get FCLK as fast as possible and keep UCLK:MCLK at 1:1 and as fast as possible.
So far the best I saw was 2133:3200:3200 (or even 2200:3200:3200) but of course not all CPUs will be able to achieve those speeds, if most not.

Also super tight timings is can make a real difference

Sorry I should have been more clear. I meant the bandwidth and latency of the memory controller for Zen.

For example, if we use the same DDR5-5200 memory with the same timings on Ryzen or Intel processors (any processor from the last few years) the bandwidth would be higher and latency would lower on the Intel platforms. The memory controller on Ryzen has always been weak, and AMD didn't try to improve it for Zen 5.

You said about double cache bandwidth?
And yeah apparently AMD didn't want to upgrade it yet... Like I said it among other in that post.

 

[ARF]

Ryzen CPUs always had AIDA scores lower than Intel ones. That didnt stop them from being faster overall in many cases.

Really? Are you sure?

Gaming performance, GPU limitations off:

Intel Core i9-13900K Review - Power-Hungry Beast

With the Core i9-13900K, Intel delivers impressive performance. Our in-depth review confirms: Raptor Lake is the world's fastest CPU for gaming. Even in applications the processor is able to match AMD's Zen 4 Ryzen 9 7950X flagship. If only power consumption wasn't so high...

www.techpowerup.com

 

[Dr. Dro]

i use to have a 12900k. mine would not work with anything faster than 6200MHz. i was able to boot with memory at 6400MHz, but it just wasn't stable. i would get application hangs, occasional crashes. it was just a terrible experience all around. but my current 7950X3d runs a 96GB kit at 6400MHz just fine. a lot of gaming prebuilds from dell, lenovo, HP and others only have 6000MHz options. their BIOS simply won't allow you to use faster memory clocks.

the point i'm trying to make is that there will be people whose hardware would simply not be able to replicate results with high memory clocks because either their motherboard doesn't support it or the memory controller on their CPU can't handle it, and any benchmarks using these speeds would be even more useless than your comment. hard to believe, i know

Sounds like you had a 4-dimm Z690 board. This is about expected, even the highest end ones max out at 6400-6800 *tops*. Higher than this will pretty much require Z790 or at least the Z690 Apex or Tachyon boards. Approaching DDR5-8000+, you need both a good CPU and a overclocking-grade Z790. AMD needs not apply, and while some Alder chips will achieve this, most of the time you will need a Raptor processor. 13th or 14th doesn't matter, 14th gen is the exact same silicon revision and stepping, only out of the box clock targets are higher with the exception of i9-13900KS = 14900K

The DDR5 throughput is quite low.

View attachment 359181

vs Core i9-14900K:

View attachment 359182

At least, the L2 cache throughput is quite high:

View attachment 359184

vs.:

View attachment 359183

Review: Ryzen 9 9950X processor - Updated with KB5041587 for Gaming Charts (Page 14)

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

Intel Core i9-14900K review (Page 15)

Join us as we review, analyse and test the Intel 14th Gen core processor, specifically the Raptor Lake-based Core i9 14900K, which has been officially released. Prior to this, there were numerous leaks and performance reviews surrounding the series. The latest model represents the advanced...

www.guru3d.com

Raptorlake achieves far higher memory bandwidth than Ryzen, this is nothing new. The other numbers aren't comparable between architectures, however, Zen 5 does have very high L1 and L2 bandwidth.

Faster Ram is not better.

Faster RAM is always better, unless you are going beyond the platform's capabilities - which is the case with socket AM5 systems thus far. 6200 C28 is very fast and very tight memory that will be great in any platform that handles it, but on Intel the 7800 C38 option would be better as the platform can make use of the extra bandwidth without having to reduce bus speed like Ryzen does.

 

[Dyatlov A]

AMDs new ZEN5 CPUs weren't ready for prime time. They should have waited till the software and new motherboards were "Officially" ready. So now AMD needs to play catch-up because these CPUs should be at a min 10-12% IPC over the previous gen.

ZEN5 suffers from Windows Scheduling issues. The same happened with ZEN4 when they first came out.

AMD did have an opportunity to really give it to Intel and all their major design issues. But nope lol, not yet I suppose.

It has the 10% IPC improvement,
View attachment IMG_0011.webp

It is just not clocking well because too thick IHS and tight Wattage.

 

[DemonicRyzen666]

Faster Ram is not better.

I don't feel like that was good benchmarking at all.
He chose all cpu bound games that mostly prefer low latency & high clock speeds, that don't care about ram bandwidth, also he's running 1080p low quality.
You don't need fast ram for that low resolution you need quickness for that. You need more bandwidth with high textures & higher resolutions. That whole benchmarking suite he did is just a cpu latency benchmark.

 

[Melvis]

We truly gone full circle with the AMD fanboys for years crying about Intel barely doing any performance uplifts each new gen only for AMD to come and do the same.

And its still a bigger jump in performance then what intel did for over a decade! Let alone 13th to 14th gen

 

[DemonicRyzen666]

I am just stating that Zen4 does not have 2 decoders per core. Zen5 mobile is not Zen4. If Zen5 decoders are working correctly or not based on an interview vs testing is something completely different.

That was my point, Zen 5 is basically broken currently.

 

[Mr_Engineer]

MSI are claiming that Ryzen 9000 CPUs achieve 4-9% performance increase vs stock settings by using the latest AGESA update and their enhanced PBO settings:

MSI Exclusive “PBO Enhanced Mode” Offers 22% Boost To AMD Ryzen 9000 CPUs, Thermal Limit Features

 

[W1zzard]

MSI are claiming that Ryzen 9000 CPUs achieve 4-9% performance increase vs stock settings by using the latest AGESA update and their enhanced PBO settings:

MSI Exclusive “PBO Enhanced Mode” Offers 22% Boost To AMD Ryzen 9000 CPUs, Thermal Limit Features

"The applications used were CPU-Z, Cinebench R20, Cinebench R23, and Cinebench 2024"

This looks like some premade profile that does what you do manually when changing PBO settings, not seeing any innovation here.

 

[Frank_100]

These cpus really just seem like wastes of sand. What apps are using avx-512 to really benefit from this architecture ?

Waves plug-ins. R, Mathematica or anything using the Intel Math Kernel Library. It really is everywhere in scientific/creative code.
Alot apps will run without it, but they run much slower.

Certainly the GCC compiler and all Intel compilers would be able to write commands for it and make use of it.