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

Page 80 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
What, no Bob?
wasn't that an interactive game for kids and adults? my memory of bob is hazy because around that same time when powerpc macs came out and interactive games like it became common. dinosauer safari was a popular game with one of my nieces then when they were younger. This is a good 25-26 years ago iirc.

seems like yesterday. had a fuller head of hair, in striking black too. food tasted better, no one was distracted by their phones because few could afford a cell phone. beepers were common but you had to be important to own one.

think ill go yell at the sky some today.
 

eek2121

Diamond Member
Aug 2, 2005
3,100
4,398
136
you're confusing llms with what we're discussing. These are algorithmic models that vastly speed up processes that would take longer using a traditional x86 processor and many hardware accelerators that use less power, less latency and can compute faster than the flow of the processor.

the current llm ai everyone is gahgah over is largely inflated. it's the next crypto. as it stands llms legitimate use is far smaller in scale than most people would like to admit to.

for the ai we're discussing certain functions pose a strange threat to other mediums. an example of which would be the future evolution of neural audio accelerators capable of clearing up bad audio tracks on the fly, boosting appropriate levels and making a low quality file sound better or use the processing power to change a file's structure by changing it and saving the file. you can always brute force something through shear power and speed but with that comes a lot of energy use. accelerators and co processors make it easier. there is a reason intel bought altera and movidius and amd bought xilinx. it wasn't to see who can land their piss the farthest.

I paid for a chat gpt 4 sub at the behest of some people here. I've used it a bit each day but find it largely useless. It's possibly useful if you're terrible at finding stuff online or want that imaginary friend you had as a child.
ChatGPT is a tool and it has it’s uses. I personally have used it to accelerate certain workflows. It saves more than $20/mo in productivity for me.

Windows 12 is rumored to use more “AI”. What that means is anyone’s guess, however, Microsoft is asking Intel/AMD to begin including AI specific instructions, so something is up.
 

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
ChatGPT is a tool and it has it’s uses. I personally have used it to accelerate certain workflows. It saves more than $20/mo in productivity for me.

Windows 12 is rumored to use more “AI”. What that means is anyone’s guess, however, Microsoft is asking Intel/AMD to begin including AI specific instructions, so something is up.
$20 a month? That's very specific. Is it giving you streaming links on nefarious websites to videos you would have to pay a monthly fee too or do you have certain skills and this saves you some over head per month?
 

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
That's the subscription price of ChatGPT Plus, which gets you access to the full GPT-4 model.
Oh yes I see now. That's how much I pay for the paid version. The way he phrased it made it sound like he's making an extra $20 a month or saving it by using the services. Few services exist in that price range and it's usually either name brand software or film/telly.

I used chat gpt last night to help with some math on soil grading, not bad! Maybe I'll use it in future to help me come up with slogan sayings so I can hang and post cringy signs outside in my front yard like one of those bored house wives. live love laughter and all that other crap.
 

eek2121

Diamond Member
Aug 2, 2005
3,100
4,398
136
Oh yes I see now. That's how much I pay for the paid version. The way he phrased it made it sound like he's making an extra $20 a month or saving it by using the services. Few services exist in that price range and it's usually either name brand software or film/telly.

I used chat gpt last night to help with some math on soil grading, not bad! Maybe I'll use it in future to help me come up with slogan sayings so I can hang and post cringy signs outside in my front yard like one of those bored house wives. live love laughter and all that other crap.
I save more than the $20 I pay by using it to automate some stuff I do, without getting into specifics, such as rewording content I have written.
 

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
I save more than the $20 I pay by using it to automate some stuff I do, without getting into specifics, such as rewording content I have written.
as you say this to me in a fresh post i watched the copilot feature video coming to windows 11 soon in an update. are you human or a modern ai, eek2121?
 

DisEnchantment

Golden Member
Mar 3, 2017
1,747
6,598
136
I see some rumors floating around of a 96 Core Zen 4 based CPU in SH5 socket. I admit such a thing did not cross my mind before but it seems like something that makes too much sense to not materialize. This should take technical computing one step above Genoa-X/Turin-X
I don't know if AMD had already conceptualized this tiered memory layout with HBM, but if it materializes it is going to be a monster.
L3--> V-Cache --> Infinity Cache --> HBM --> DDR5
Execufix caught wind of it as usual in the past.
 

moinmoin

Diamond Member
Jun 1, 2017
5,064
8,032
136
I see some rumors floating around of a 96 Core Zen 4 based CPU in SH5 socket. I admit such a thing did not cross my mind before but it seems like something that makes too much sense to not materialize. This should take technical computing one step above Genoa-X/Turin-X
I don't know if AMD had already conceptualized this tiered memory layout with HBM, but if it materializes it is going to be a monster.
L3--> V-Cache --> Infinity Cache --> HBM --> DDR5
Execufix caught wind of it as usual in the past.
Yeah, had been mentioned in the RDNA4 + CDNA3 thread.

The memory layout will be really interesting with this thing.
 

jamescox

Senior member
Nov 11, 2009
644
1,105
136
I see some rumors floating around of a 96 Core Zen 4 based CPU in SH5 socket. I admit such a thing did not cross my mind before but it seems like something that makes too much sense to not materialize. This should take technical computing one step above Genoa-X/Turin-X
I don't know if AMD had already conceptualized this tiered memory layout with HBM, but if it materializes it is going to be a monster.
L3--> V-Cache --> Infinity Cache --> HBM --> DDR5
Execufix caught wind of it as usual in the past.

It has beeen speculated for a while. An SH5 article from Sept. 2021 on wccftech mentions the possibility of a Zen 4 cpu in SH5. The bandwidth would be ridiculous; I am not sure a Zen 4 chiplet can really make full use of it, so it might be overkill. Zen 5, if it has significant FP upgrades, may make better use of it. Since it will likely have significantly higher max power, it may be able to run even a 96-core at relatively high clock, so that should be able to consume a lot of bandwidth, especially with AVX512 in use. If they exist, they are going to be very expensive and high power consumption, but I am not sure what would come close to competing with it.

I have been wondering for a long time about the “line” or “page” size for HBM as cache. Cpu caches use (commonly) 64-byte cache lines. This is really tiny for a gpu though, so I have been assuming that the HBM cache functions on probably 4K pages. Anyone seen specifics anywhere?
 
  • Like
Reactions: Tlh97

Joe NYC

Platinum Member
Jun 26, 2021
2,535
3,466
106
It has beeen speculated for a while. An SH5 article from Sept. 2021 on wccftech mentions the possibility of a Zen 4 cpu in SH5. The bandwidth would be ridiculous; I am not sure a Zen 4 chiplet can really make full use of it, so it might be overkill. Zen 5, if it has significant FP upgrades, may make better use of it. Since it will likely have significantly higher max power, it may be able to run even a 96-core at relatively high clock, so that should be able to consume a lot of bandwidth, especially with AVX512 in use. If they exist, they are going to be very expensive and high power consumption, but I am not sure what would come close to competing with it.

I have been wondering for a long time about the “line” or “page” size for HBM as cache. Cpu caches use (commonly) 64-byte cache lines. This is really tiny for a gpu though, so I have been assuming that the HBM cache functions on probably 4K pages. Anyone seen specifics anywhere?

AVX512 power consumption has not gone off the charts in Zen 4 as is the case with Intel CPUs. But that was with limited bandwidth into the chiplet. Who knows what happens if the bandwidth goes up by orders of magnitude.

Looking forward to Zen 5 Turin Dense chiplet, which will likely have 16 core on N3, and will minimize L3 size, some of that old L3 bandwidth demand will move to the system level cache on base die. This would optimize the use of limited and expensive N3 silicon to only the most dense logic, and pretty much all of the low density SRAM, IO, analog would be on N6 base die.
 
  • Like
Reactions: Tlh97