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

Page 580 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Mopetar

Diamond Member
Jan 31, 2011
8,004
6,444
136
Would be ironic and a bit strange.
AMD's SMT was already far better than Intel's.
AMD vastly improved it even more, while...

... Intel want to give up on SMT entirely?

Still, would be a questionable choice outside of servers as for most other things the ST crown is a huge seller.

Server is where AMD is making most of its money, so it's not surprising that they would make design choices that favor it over other markets.

Intel dumping SMT seems shortsighted, but they have their E-cores that are there to tackle the same problem and validation for SMT seemed to be an issue to the point where not having it may have allowed for faster development times and no security surprises a year down the road.
 
Reactions: KompuKare

Nothingness

Platinum Member
Jul 3, 2013
2,722
1,357
136
Then there's me, who looks with deep suspicion on any SIMD beyond 128b. VMX got a lot of things right the first time.
I'd argue being 128-bit is not necessarily what makes VMX/AltiVec a good SIMD ISA.

IMHO a good SIMD ISA is good independently of the width. The problem starts when the good parts of the ISA are attached to widths useless to most people as is the case for AVX-512. AVX10.2 looks like a step in the right direction (at last).
 
Jul 27, 2020
17,797
11,596
106
You need to convert the fame into money :-D
Nah. I'm useless with money. I converted my $6000 lottery into 12700K + Sonic Z790 mobo, used M1 Macbook Air (still cost me $1000 coz I got it in return for new 3060 Ti and 3080 10GB cards), a used MSI Z690 DDR4 mobo with DDR4-4000 sticks, two Dell Precision used laptops (one Xeon with eDRAM and other Core 8th gen), four SODIMMs to upgrade the Precision laptop to 128GB RAM but never got around to doing that, an ASUS 6800 XT LC card (NIB and unused), used AORUS RX 6800 and who knows what other crap I bought.

I look at that old movie where the guy has 30 days to spend millions or he loses and I'm like, 30 days is a bit too long. I could spend that much in a week or so. Two weeks tops!
 

SarahKerrigan

Senior member
Oct 12, 2014
593
1,427
136
I'd argue being 128-bit is not necessarily what makes VMX/AltiVec a good SIMD ISA.

IMHO a good SIMD ISA is good independently of the width. The problem starts when the good parts of the ISA are attached to widths useless to most people as is the case for AVX-512. AVX10.2 looks like a step in the right direction (at last).

Of course - I wasn't saying that 128b width is what made VMX good, but I meant to allude to it as an example of "features over width" (ie, rich permutes, lanewise integer rotate, etc.)

That being said, I have deep skepticism of 256b or 512b vectors as a good use of silicon area for anything that doesn't look like HPC.
 

Nothingness

Platinum Member
Jul 3, 2013
2,722
1,357
136
It's pretty annoying that GB6 dropped the split between INT, FP and crypto scores. It makes deciphering performance gains difficult, and we don't get SPEC results that often.
I posted a list of what tests are INT and FP in GB6, reverse-engineered by playing with the geomean total and the formula Primate Labs gave.

But yeah you have to extract scores and put them in a spreadsheet
 

Mopetar

Diamond Member
Jan 31, 2011
8,004
6,444
136
The next step will be "GB6 is so great" because it uses AVX-512 and demonstrates AMD is faster than the competition that lacks it.

Intel had it when AMD didn't and for whatever reason disabled it. The only real issue is that it used a lot of power, but for the workloads it could be used in, the performance gains were worth it.

I think they just wanted higher base clock or all core boost numbers or something, which I don't think anyone cares about since the 1T boost number is what gets marketed.

It is funny how the tables have turned. There was a time where the only saving grace for the Intel chips was the AVX-512 benchmarks.

It is more of a server feature that consumers get just by virtue of AMD having a single design (excluding their compact core here, but it's the same ISA anyways) for all of their product categories.

I could also see why AMD would push for using it. The whole point of AVX is that you issue far fewer instructions for the equivalent amount of work performed. Having a front-end that's still effectively the same for a single thread means there's a lot more pressure on it and that it will bottleneck the bigger backend.
 

Nothingness

Platinum Member
Jul 3, 2013
2,722
1,357
136
That being said, I have deep skepticism of 256b or 512b vectors as a good use of silicon area for anything that doesn't look like HPC.
I think that AVX2 has good usage now. The question is: is it because of its width, or because it's richer than SSE/AVX? Or both?

For AVX-512, I agree. 512-bit looks too wide for general usage.
 

KompuKare

Golden Member
Jul 28, 2009
1,069
1,108
136
If it only were this way around. I remember people talking about the utmost importance of AVX-512 on Ice-Lake laptops rendering all current and near future AMD chips useless
Specifically a few years ago posters who were fairly and talked as if they had just bought a whole bunch of Intel shares and kept going on about how with AVX-512 Intel would conquer the world.
Haven't seen any of those posters in a while but I guess AVX-512 was not the latest thing after all, at least not for Intel!
 

CouncilorIrissa

Senior member
Jul 28, 2023
295
1,034
96
I posted a list of what tests are INT and FP in GB6, reverse-engineered by playing with the geomean total and the formula Primate Labs gave.

But yeah you have to extract scores and put them in a spreadsheet
That's a massive help, thanks. The .gb6 file looks like a json, so I'll use my skills as a webshit to parse it and calculate the scores.

Maybe. When I'm bored.
 
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/    |