Discussion Zen 5 Builders thread

Page 22 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Makaveli

Diamond Member
Feb 8, 2002
4,864
1,402
136
That sounds like awful price for a Prime mobo. You can find better options with more features in that price range.

$20 off: https://www.newegg.com/p/N82E16813144667

Below your price range and supports up to DDR5-8400.
Those are USD prices and I'm in Canada!!

Newegg.ca has that MSI board for $419 CAD.

8400 memory support means nothing to me as I cannot run that 1:1

I'm already looking at this kit here moving from 32GB of ram to 64GB for the next build.

G.SKILL Trident Z5 Neo RGB 64GB (2x32GB) DDR5 6000MHz CL30
 
Reactions: DAPUNISHER
Jul 27, 2020
20,921
14,495
146
8400 memory support means nothing to me as I cannot run that 1:1
But it means it has better traces so it will be more stable and may let you get to lower latencies more easily. It could possibly let you use that kit at CL28 without a hair pulling tweaking experience.
 

Makaveli

Diamond Member
Feb 8, 2002
4,864
1,402
136
But it means it has better traces so it will be more stable and may let you get to lower latencies more easily. It could possibly let you use that kit at CL28 without a hair pulling tweaking experience.
yes very true but may not be worth the price difference.

Running CL28 vs CL30 is a small gain.
 
Jul 27, 2020
20,921
14,495
146
Running CL28 vs CL30 is a small gain.
Oh well. Anyway, I tried

To me, ASUS mobos look overpriced compared to the competition. That's the only reason I have only one that I bought recently (Z97 Mark 2 for $81. Coz I wanted something better for my i7-5775C than the crappy Gigabyte it is in now).
 
Reactions: Makaveli

Makaveli

Diamond Member
Feb 8, 2002
4,864
1,402
136
Oh well. Anyway, I tried

To me, ASUS mobos look overpriced compared to the competition. That's the only reason I have only one that I bought recently (Z97 Mark 2 for $81. Coz I wanted something better for my i7-5775C than the crappy Gigabyte it is in now).
This is why the Prime board is a good balance in terms of features and price. Most other boards are all over priced. The creator board you suggested is $600 CAD.

I've looked at most of the options already

The money I will save here will go into the RAM kit.
 
Reactions: igor_kavinski

StefanR5R

Elite Member
Dec 10, 2016
6,057
9,107
136

7800X3D is the top CPU for Pytorch. No way should you go for a 9950X. 9800X3D is the answer for your workloads.
Are these training tests or inference tests?
Looks like inference: https://github.com/LukasHedegaard/pytorch-benchmark
@9950x's field would be training, specifically that of GPGPU-implemented models.

I used to have a 13900k+4090 where I found some performance improvement when using 8p+8e over 8p+0e for the Dataloader when training smaller models. Hence I'm worried that a 9800x3D would leave performance on the table, especially when using a 5090.
13900k's 8p+8e are (up to) 24 threads, with somewhat different performance between them, whereas 9800X3D has got 16 threads which all perform almost equally, have beefy SIMD at their disposal, and an extraordinary amount of cache, obviously. But as I said, I have no idea what is most effective in your use case.

BTW, dual CCX (edit) dual CCD Ryzens have greater RAM access bandwidth than single CCX Ryzens, if the workload is spread across both CCXs.
 
Last edited:
Reactions: IEC

9950x

Junior Member
Nov 17, 2024
3
1
36
Are these training tests or inference tests?
Looks like inference: https://github.com/LukasHedegaard/pytorch-benchmark
@9950x's field would be training, specifically that of GPGPU-implemented models.


13900k's 8p+8e are (up to) 24 threads, with somewhat different performance between them, whereas 9800X3D has got 16 threads which all perform almost equally, have beefy SIMD at their disposal, and an extraordinary amount of cache, obviously. But as I said, I have no idea what is most effective in your use case.

BTW, dual CCX (edit) dual CCD Ryzens have greater RAM access bandwidth than single CCX Ryzens, if the workload is spread across both CCXs.
On dual-CCD Ryzens, as well as on Threadrippers and EPYCs, we who are active in citizen science/ scientific computing occasionally use something more targeted than "core parking": When we have multithreaded applications which heavily share data between threads, we assign CPU affinity to thread groups (usually to processes a.k.a. tasks) such that the heavy sharing is confined to a CCX ( = to a last level cache domain) which is common to this thread group. On EPYCs or/and on old Zen 2 CPUs, we also sometimes arrange for such sharing across two CCXs if we want more cores per thread group. Maybe that's what you would call micro-management, but in our case, this is set-and-forget. IOW it's trivial to set up and operate once the performance characteristics of the application are known.

We have got applications in which this nets us just a moderate percentage of performance and power efficiency gain. But we also have edge case applications in which the gain can be way up in double-digit percentages on Ryzen, or even more on EPYCs with more CCXs. This specifically happens if vector arithmetic is performed on datasets that need to be kept synchronized between threads. Without scheduling hints, the cores would wait on RAM accesses a lot and also increase task energy while doing so.

However, whether or not your use case (DL training) has got similar requirements is not known to me.

Edit, PS:
Now this raises the question: Why not use CPUs from AMD's competitor, ones which have a single cache domain per processor? The answer is of course that currently AMD's CPUs are just so much better performing and efficient in computing-heavy tasks, and AMD's splitting of larger CPUs into CCXs is one of the very reasons for this advantage of theirs. And by that I specifically mean CCXs, not even CCDs.
Very much appreciate the explanation, scheduling hints & techniques are new concepts to me but make a lot of sense.

My specific use case for the CPU is:
1. Retrieve a sample from a dataset on SSD
2. Optional: apply some function on the sample
3. Feed the sample into the GPU

^This process is completely parallelizable and require no synchronization between threads. Surprisingly though, I've found 8p+8e > 8p+0e > 4p > 0p+16e = 0p+8e.
For some reason, E-cores on their own are really slow, even for just retrieving samples from disk. Using 16e cores is significantly slower than just using 4p cores.

This was especially problematic with my 13900k, as if I were to use the computer for anything such as simple web browsing or coding, all processes would be pushed to the E-cores and I see a 40% slowdown in training... Preventing this would require micro-management (process IDs change every training run). That's why I'm avoiding the 285k altogether.


7800X3D is the top CPU for Pytorch. No way should you go for a 9950X. 9800X3D is the answer for your workloads.
This does not exactly match my use case (data sample retrieval from disk with low latency) but seems promising nonetheless. I've also seen that 9800x3D achieves the best python performance which is positive.
 
Reactions: igor_kavinski

biostud

Lifer
Feb 27, 2003
18,846
5,711
136
This is why the Prime board is a good balance in terms of features and price. Most other boards are all over priced. The creator board you suggested is $600 CAD.

I've looked at most of the options already

The money I will save here will go into the RAM kit.
View attachment 111798
What thing do you need?

Have you looked at any of the ASRock boards?

 
Reactions: Gideon

Golgatha

Lifer
Jul 18, 2003
12,258
672
126
Did some tuning on my 9800X3D I just got today. Getting 5.30Ghz all-core and around 5.47 single core while running Cinebench R23. Set PPT limit to 130w, TDC at 110A, EDC at 120A. +200 manual overclock with -20 all core on curve optimizer. Max temp with a Scythe Fuma 3 using a graphite thermal pad is around 86°C on all cores. At the desktop, it hums along at around 46°C. I manually limited the max temp to 90°C just to make sure I didn't cook my new shiny. Pretty happy with the performance, but is there anything else I should try or change to fully unlock the potential of this CPU?

 
Reactions: lightmanek

DaaQ

Golden Member
Dec 8, 2018
1,568
1,141
136
Did some tuning on my 9800X3D I just got today. Getting 5.30Ghz all-core and around 5.47 single core while running Cinebench R23. Set PPT limit to 130w, TDC at 110A, EDC at 120A. +200 manual overclock with -20 all core on curve optimizer. Max temp with a Scythe Fuma 3 using a graphite thermal pad is around 86°C. I manually limited the max temp to 90°C. Pretty happy with the performance, but is there anything else I should try or change to fully unlock the potential of this CPU?

View attachment 111824
I forgot about the Kryosheet, (pad) for the CPU, I am using one on the 7900XTX, what size did you use for the 9800x3d? I used PK2 or 3 on my current cpu. Which runs hotter than any of the Ryzens. Still stay under 55 degrees.

EDIT: I also used them Kritical thermal pads on the VRM side of my XTX, seem to be working quite well, as I have not seen over 68 degree on memory junction temps. Which have been the hottest temp according to hwinfo.
 
Reactions: igor_kavinski

Golgatha

Lifer
Jul 18, 2003
12,258
672
126
I forgot about the Kryosheet, (pad) for the CPU, I am using one on the 7900XTX, what size did you use for the 9800x3d? I used PK2 or 3 on my current cpu. Which runs hotter than any of the Ryzens. Still stay under 55 degrees.

EDIT: I also used them Kritical thermal pads on the VRM side of my XTX, seem to be working quite well, as I have not seen over 68 degree on memory junction temps. Which have been the hottest temp according to hwinfo.
Just a relatively inexpensive 40x40mm pad. I've reused it a couple of times now too. I do have one of those metal Thermalright AM5 CPU shims installed as well.
 
Reactions: DaaQ

DaaQ

Golden Member
Dec 8, 2018
1,568
1,141
136
Just a relatively inexpensive 40x40mm pad. I've reused it a couple of times now too. I do have one of those metal Thermalright AM5 CPU shims installed as well.
Can you link the shim please. Thank you for the link.

Seems like them Kritical pads have vaporwared. I used 1.0mm on my GPU memory and VRM with better than expected results. Maybe the 9800X3D will change that. I am Currently PCIE2.0
 

Shmee

Memory & Storage, Graphics Cards Mod Elite Member
Super Moderator
Sep 13, 2008
7,817
2,783
146
Has anyone here been able to go from AM4 to AM5 without having to reinstall Windows 10? I am wondering what the experience was like. I would prefer to stay on Windows 10 and not have to reinstall.
 
Reactions: Makaveli

Gideon

Golden Member
Nov 27, 2007
1,842
4,380
136
Did some tuning on my 9800X3D I just got today. Getting 5.30Ghz all-core and around 5.47 single core while running Cinebench R23. Set PPT limit to 130w, TDC at 110A, EDC at 120A. +200 manual overclock with -20 all core on curve optimizer. Max temp with a Scythe Fuma 3 using a graphite thermal pad is around 86°C on all cores. At the desktop, it hums along at around 46°C. I manually limited the max temp to 90°C just to make sure I didn't cook my new shiny. Pretty happy with the performance, but is there anything else I should try or change to fully unlock the potential of this CPU?

View attachment 111824
Other than memory timings, it seems to be pretty well setup for a 24/7 config.

ScatterBencher has an interesting video on the subject, discussing most major avenues to take:

Bare in mind, the ECLK strategies are probably very far from a 24/7 stable build (unless someone has experience with this and claims otherwise?)

From there you could probably try Precision Boost Overdrive Scalar (that boosts voltage sightly though) + play around with the curve optimiser but it's certainly diminishing returns.

What memory timings are you running?

Buildzoid has an older list for low-effort ones that are known to work on most Hynix A/M dies:

If you want to tune the max out of it i suggest this video though:
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |