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

Page 872 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

naukkis

Senior member
Jun 5, 2002
962
829
136
It's trivial for trivial cases...

It's pretty much trivial to all cases. People here are probably too young to ever coding for vector CPU like cray. Pretty much all loops in code is easily vectorizable. SIMD hardware is the hard part as cpu hardware lacks everything needed to use vectorized models. SIMD still is tivial, it's just massive amount of work needed to code and sustain that code as little changes can broke vectorization models. With AI SIMD models finally might be worth using as they can take care of that hassle needed. On the other hand there's risc-V RVV, first consumer grade vector ISA available for vectorized models that are actually usable.
 

biostud

Lifer
Feb 27, 2003
18,846
5,706
136
Amazon had the 7800x3D for $335 a month ago. Just throwing that out there if the MSRP is $450+ for 9800x3D.
So, grab your time machine and get one or wait until the 9800X3D falls below $350, which should be 3-4 months befor the launch of the 10800X3D.

Obviously, they are never good value at launch but since they'll sell like hot cakes anyways you can hardly blame them for the pricing.
 

Aeonsim

Junior Member
May 10, 2020
13
42
91
It's pretty much trivial to all cases. People here are probably too young to ever coding for vector CPU like cray. Pretty much all loops in code is easily vectorizable. SIMD hardware is the hard part as cpu hardware lacks everything needed to use vectorized models. SIMD still is tivial, it's just massive amount of work needed to code and sustain that code as little changes can broke vectorization models. With AI SIMD models finally might be worth using as they can take care of that hassle needed. On the other hand there's risc-V RVV, first consumer grade vector ISA available for vectorized models that are actually usable.
If it was always trivial the compilers would already have solved it and everything would be SIND vectorised by default.

For pure math functions with minimal branches and large data it's certainly possible to vectorise the code relatively easily and gain performance. But once you start including conditionals or write heuristic algorithms it's much more challenging. Also many SIMD operations are architecture specific, which adds another layer of complexity, with each arch supporting different SIMD functions and data sizes.

At the moment there are numerous languages developing generic SIMD implementations to simplify and make writing cross platform SIMD code simpler. The two main reasons driving this seem to be compilers find it exceedingly hard to reliable vectorise algorithms or code. Secondly the massive diversity of SIMD implementations makes it difficult to write reliable code using native or intrinsic SIMD functions. Remember for computers built-in on the last 10 years your SIMD implementation could be one or more of the following:
SSE4.2, AVX, AVX2, AVX512(F, BW, ...), NEON, AMX, SVE, SME

That's just counting ARM and x86 and ignoring legacy ones like MMX, SSE1-3.
 

Hans Gruber

Platinum Member
Dec 23, 2006
2,369
1,259
136
So, grab your time machine and get one or wait until the 9800X3D falls below $350, which should be 3-4 months befor the launch of the 10800X3D.

Obviously, they are never good value at launch but since they'll sell like hot cakes anyways you can hardly blame them for the pricing.
We still have to see the performance numbers of the 9800x3D. The time machine is not necessary. I think a lot of people are sitting out this generation of both AMD and Intel CPU's. I know 7800x3D owners are sleeping with a smile on their faces.
 

naukkis

Senior member
Jun 5, 2002
962
829
136
If it was always trivial the compilers would already have solved it and everything would be SIND vectorised by default.
It's trivial for AI. You describe what to do and AI can vectorize it - and most importantly maintain vectorized code. Now SIMD coding overhead ain't worth doing for pretty much everything but most critical loops - AI might change it.
 

Thibsie

Senior member
Apr 25, 2017
913
1,019
136
While modern compilers are good, they don't have the understanding of the code and its intentions. But that could all change with future AI/LLM backed compilers, where compiler could understand the intention and optimize good as or better than experienced assembly programmer (even assembly programmer have to make some sacrifices to make the code more manageable which in turn can reduce performance). AI could rewrite the code to make to get maximum possible performance as well as reduce bloated applications.
Maybe in the future.
Now way for now.
 

Bigos

Member
Jun 2, 2019
159
397
136
It's trivial for AI. You describe what to do and AI can vectorize it - and most importantly maintain vectorized code. Now SIMD coding overhead ain't worth doing for pretty much everything but most critical loops - AI might change it.

Please, demonstrate an AI that automatically vectorizes code and does that with 100% accuracy (i.e. preserves the original meaning of the code). The AI I know of has troubles writing even non-vectorized code that is accurate and requires humans to review it and fix things.
 

naukkis

Senior member
Jun 5, 2002
962
829
136
Please, demonstrate an AI that automatically vectorizes code and does that with 100% accuracy (i.e. preserves the original meaning of the code). The AI I know of has troubles writing even non-vectorized code that is accurate and requires humans to review it and fix things.

Making AI to code is the hard part. Converting that code to SIMD model is trivial. AI has all time needed to produce and maintain SIMD - people don't.
 

eek2121

Diamond Member
Aug 2, 2005
3,134
4,479
136
AMD should release a dual 3D V-Cache part just so the have the opportunity to bring back the x2 branding. Also a consumer oriented 4 chiplet threadripper part for the x4. 🤣

I was just to post this. That's just tomshardware being doing their british tabloid level articles.

I'm still salty they kept financing that thing and shut down Anandtech
I’ve been wanting to start a review site. However, I lack funds currently for hardware, but I have considered reaching out to other reviewers to partner up. Still on the fence. My goal would be to remain independent and probably rely on optional subscriptions/donations.
So, grab your time machine and get one or wait until the 9800X3D falls below $350, which should be 3-4 months befor the launch of the 10800X3D.

Obviously, they are never good value at launch but since they'll sell like hot cakes anyways you can hardly blame them for the pricing.
I suspect AMD will rebrand after the 9000 series like they did for mobile.
 

DAPUNISHER

Super Moderator CPU Forum Mod and Elite Member
Super Moderator
Aug 22, 2001
29,888
25,676
146
I am predicting that the stock out of the box experience for 9800X3D won't be blowing anyone's socks off in this forum; however, I also think that with a bit of tweaking, and good memory, this will be a gamers dream.
That's the selling point right there. 7800X3D is in the top spot on Amazon U.S. at $476. No amount of Ferengi math from reviewers is going to keep the 9800X3D from claiming that spot if there is sufficient stock over the next 3 months.

Did I see correctly that PBO is 5450MHz? That would track with all of the vanilla Zen 3 I've used. With $30 air cooling they hit 250MHz with ECO mode and +200MHZ PBO.
 

Josh128

Senior member
Oct 14, 2022
511
865
106
That's the selling point right there. 7800X3D is in the top spot on Amazon U.S. at $476. No amount of Ferengi math from reviewers is going to keep the 9800X3D from claiming that spot if there is sufficient stock over the next 3 months.

Did I see correctly that PBO is 5450MHz? That would track with all of the vanilla Zen 3 I've used. With $30 air cooling they hit 250MHz with ECO mode and +200MHZ PBO.
The recent Geekbench run of ST 3475 shows ~5460MHz auto-boosting, which is likely achieved by a +100-150 MHz boost override and unknown CO voltage offset. AMD conservatively chose to avoid a 5.4 or 5.5 listed max boost to protect themselves from reliability issues and thus liability for any unforeseen problems that may arise with this new technology. But it would appear that with a bit of PBO tweaking, 5.45 auto boosting is going to be readily achievable, and indeed should be easily achievable on air.

I blew all the 'tards minds over at WCCFTECH by telling them Im running my 9900X on air. They think you need an 240mm AIO even for a sub 100W 7800X3D over there. They've been tricked by business. The reality is that liquid cooling doesnt even need to be considered unless you plan to be running a chip at >200W PPT for extended periods of time.
 
Last edited:
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/    |