Discussion RDNA4 + CDNA3 Architectures Thread

Page 160 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DisEnchantment

Golden Member
Mar 3, 2017
1,747
6,598
136





With the GFX940 patches in full swing since first week of March, it is looking like MI300 is not far in the distant future!
Usually AMD takes around 3Qs to get the support in LLVM and amdgpu. Lately, since RDNA2 the window they push to add support for new devices is much reduced to prevent leaks.
But looking at the flurry of code in LLVM, it is a lot of commits. Maybe because US Govt is starting to prepare the SW environment for El Capitan (Maybe to avoid slow bring up situation like Frontier for example)

See here for the GFX940 specific commits
Or Phoronix

There is a lot more if you know whom to follow in LLVM review chains (before getting merged to github), but I am not going to link AMD employees.

I am starting to think MI300 will launch around the same time like Hopper probably only a couple of months later!
Although I believe Hopper had problems not having a host CPU capable of doing PCIe 5 in the very near future therefore it might have gotten pushed back a bit until SPR and Genoa arrives later in 2022.
If PVC slips again I believe MI300 could launch before it

This is nuts, MI100/200/300 cadence is impressive.



Previous thread on CDNA2 and RDNA3 here

 
Last edited:

branch_suggestion

Senior member
Aug 4, 2023
373
831
96
AIMD king of Compute
I may crap on AI all day but it is a legitimately amazing all-around capable HPC chip too so...yay of big stuff.
It is all but certain that MI350 and onwards will gut FP64, just as Blackwell did.
Still overall better in HPC regardless.
IIRC, the latest rumors are pointing more in the 270mm² direction.

But generally speaking, size isn't necessarily a good indicator. Not all architectural improvements cost much area, sometimes newly designed IP blocks even occupy less area.
Case in point: Despite more L0 and L1 cache, improved RT and doubled maximum FP32 throughput, Navi33 is 25mm² smaller than N23, and normally the N6 process doesn't offer enough density improvement over N7 to accomplish that, so they must've reduced the size of some units while still improving their performance.

For RDNA4, the known info and rumors suggest RT-Performance per CU per clock may more than double. That doesn't mean double the FPS, but the perf hit taken from enabling RT may get cut in half.
Basically, in games where a 7900XTX may drop from 100 to 60 fps if you enable RT, an N48 card may only drop from 100 to 80 (or 90 to 72).

Besides, if the architecture is better optimized for clocks/efficiency than its predecessor, a ~20% higher clockspeed can do a lot for performance per area, and RDNA4 is suggested to clock significantly higher than RDNA3.
I'm still banking on SkyJuice dropping the nuke outta nowhere.
 
Reactions: Tlh97

Tuna-Fish

Golden Member
Mar 4, 2011
1,474
1,966
136
Nvidia will be using essentially the same silicon as the 40 series
Will they? They could move to their own rebranded version of N4P too. I would certainly expect them to.

For AMD's sake, I hope they use 20gbit DDR6.

The rumors are that the top card of the lineup will, with the cut models cutting memory speed.

I understand that the price difference between 20GT and 18GT GDDR6 is greater than usual for such a small speed difference, and has grown recently. Apparently a huge portion of all GDDR6 made still doesn't make the cut for 20GT speeds, and the consoles helped make 20GT more affordable by being a huge volume customer for the bad bins.
 
Reactions: Tlh97 and marees

Mahboi

Golden Member
Apr 4, 2024
1,007
1,831
96
It is all but certain that MI350 and onwards will gut FP64, just as Blackwell did.
Still overall better in HPC regardless.
Sad, but not that surprising. So what, they'll have another top tier HPC chip? Or they're just not caring for now, AI takes all?
I'm still banking on SkyJuice dropping the nuke outta nowhere.
The Angstronomics guy? Is he still alive? Haven't heard a thing in forever.
 

branch_suggestion

Senior member
Aug 4, 2023
373
831
96

So, next gen GPUs will become unified. But what it means is that the UDNA GPUs will span from APUs to desktop to server/HPC products.
And now everything makes sense.
When UDNA1 or whatever is introduced, might be as early as MI350.
As was said earlier, this means Instinct gets the better WGP and memory hierarchy of RDNA, and Radeon gets some low cost Matrix accel as NV's heavier approach yields little actual performance on client due to membw bottlenecks.
Still better than an NPU, assuming it has support.
Having one core uArch with a lot of levers depending on the application is what NV has done since forever. It does slightly limit maximum PPA for a given market, but you have only one set of everything to manage instead of two.
Besides, halo parts win on how much area you are willing to commit to, so small changes here don't really matter.
Sad, but not that surprising. So what, they'll have another top tier HPC chip? Or they're just not caring for now, AI takes all?
More or less, FP64 is very hungry in every way, most HPC apps are memory bound, common theme.
The Angstronomics guy? Is he still alive? Haven't heard a thing in forever.
I believe.
 

Mahboi

Golden Member
Apr 4, 2024
1,007
1,831
96
And now everything makes sense.
When UDNA1 or whatever is introduced, might be as early as MI350.
This is lovely but calling it UDNA doesn't sit right. Sounds like Jawa language.

(best subtitles ever)

Also Radeon DNA was a meh name, Compute DNA was just silly, Unified DNA is just....nah. Unified Radeon or Ultimate Radeon?
Universal Radeon Architecture - Neural Unified Scaffold
As was said earlier, this means Instinct gets the better WGP and memory hierarchy of RDNA, and Radeon gets some low cost Matrix accel as NV's heavier approach yields little actual performance on client due to membw bottlenecks.
Still better than an NPU, assuming it has support.
Having one core uArch with a lot of levers depending on the application is what NV has done since forever. It does slightly limit maximum PPA for a given market, but you have only one set of everything to manage instead of two.
Besides, halo parts win on how much area you are willing to commit to, so small changes here don't really matter.
Yay for finally going to the deep end.
 
Reactions: Mopetar and Avalon

Kepler_L2

Senior member
Sep 6, 2020
514
2,089
136

marees

Senior member
Apr 28, 2024
374
436
96

So, next gen GPUs will become unified. But what it means is that the UDNA GPUs will span from APUs to desktop to server/HPC products.
Looks like AMD has a long term strategy / vision

I am still suspicious/skeptical of short/medium term execution

The absence of a monolithic navi '49' to replace n31 sucks. It feels like Lisa Su took her toys back & went home
 

Mahboi

Golden Member
Apr 4, 2024
1,007
1,831
96
Looks like AMD has a long term strategy / vision

I am still suspicious/skeptical of short/medium term execution
What execution? They murdered all RDNA 4 chiplets and only outputting a tiny laptop die and a monolithic small die in case all chiplets failed. There is nothing left to execute, N44/48 out, then it's everyone on board the chiplets train for MI350X and same thing but on a cheaper solution for RDNA 5.
The absence of a monolithic navi '49' to replace n31 sucks. It feels like Lisa Su took her toys back & went home
So it does.
I can accept the whole "we focus everything on compute so F gheymers" if at least the compute SW support really gets boosted hard in the next 2 years.
Just...get Blender to run decently. Please. And get some freaking early support for RDNA 4 for everything.
Literally my only request is to not freaking wait 7 months for VR to work this time, or to see that the HW has the chops but the SW just doesn't use it right.
You can't get all the software running in 2 years but you can make some heavy strides.
 

marees

Senior member
Apr 28, 2024
374
436
96
I think maybe with RDNA4?
RDNA 4 all set to be revealed


Sony is holding a PlayStation “Technical Presentation” tomorrow, just days after teasing a PS5 Pro console. The nine-minute stream will be hosted by Mark Cerny, the lead architect of the PS5 console. Cerny previously revealed the full PS5 specs ahead of its debut in 2020.
The stream will start at 8AM PT / 11AM ET / 4PM UK on September 10th and will be broadcast on the PlayStation YouTube channel. Sony hasn’t provided any further details about the technical presentation, but the announcement comes just after Sony teased the PS5 Pro in a 30th anniversary PlayStation image.


 

marees

Senior member
Apr 28, 2024
374
436
96
As per leaks today is the D-day

It is here @SolidQ @blckgrffn

RDNA 4 all set to be revealed


Sony is holding a PlayStation “Technical Presentation” tomorrow, just days after teasing a PS5 Pro console. The nine-minute stream will be hosted by Mark Cerny, the lead architect of the PS5 console. Cerny previously revealed the full PS5 specs ahead of its debut in 2020.
The stream will start at 8AM PT / 11AM ET / 4PM UK on September 10th and will be broadcast on the PlayStation YouTube channel. Sony hasn’t provided any further details about the technical presentation, but the announcement comes just after Sony teased the PS5 Pro in a 30th anniversary PlayStation image.


 

branch_suggestion

Senior member
Aug 4, 2023
373
831
96
RDNA 4 all set to be revealed


Sony is holding a PlayStation “Technical Presentation” tomorrow, just days after teasing a PS5 Pro console. The nine-minute stream will be hosted by Mark Cerny, the lead architect of the PS5 console. Cerny previously revealed the full PS5 specs ahead of its debut in 2020.
The stream will start at 8AM PT / 11AM ET / 4PM UK on September 10th and will be broadcast on the PlayStation YouTube channel. Sony hasn’t provided any further details about the technical presentation, but the announcement comes just after Sony teased the PS5 Pro in a 30th anniversary PlayStation image.


PS5 Pro is 3.5 with some 4 features, but still gives nice early insight.
 
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/    |