AMD: Navi Speculation, Rumours and Discussion [2019-2020]

Status
Not open for further replies.
It's occurred to me that PS5 and Navi 2x might both have the "Infinity Cache" architecture (both seemingly being 256-bit) while XSX doesn't.

Microsoft designed a GPU/CPU split in the memory architecture, with 10GB dedicated to the GPU while receiving the full 560GB/s, but Sony decided to use a "flat" design. So Infinity Cache could be seen as salve for the "relatively low" 448GB/s PS5 has, and presumably Navi 2x (though Navi 21 with a 256-bit bus could have as much as 512GB/s).
 
The biggest question at the moment is is whether that USB-C is practically dead VirtualLink or something else using USB-C
 
xEx said:
Eh53l3-XsAADDTj


This is the new cooler AMD will use. I like it.
Click to expand...

I like the three R's. Educational. Reading, 'riting and 'rithmatic.
 
About limited bandwidth.
Could AMD use some kind of texture compression?

Sony bragged about compressing textures, to half size, lossless.
They use it to transfer data, from NVMe to Vram, at double speed.

I wonder if same technique could be used to transfer data from Vram to GPU?
Som kind av hardware circuit, to decompress, with very low latency, and maybe a memory pool, for decompressed data.

Sounds plausible, that AMD do something similar in the Consoles.
Is this doable, or have my thoughts walked astray?
 
Last edited:
Missed 10 minutes edit window, "Som kind av" should have been "Some kind of", sorry for my "Swenglich".

I see latency as the biggest problem, but I wonder if that can be solved by a very fast decompressing circuit?
Will solve most of Direct Storage too, and 16 GB Vram will be more than enough.
 
Wesker said:
I may be the only person on the net that prefers blower style reference coolers. I want that heat dumped outside of my case, damn it!
Click to expand...
Blowers are inevitably loud, even if they cool well. And most of the time, they can't cool as well as axial designs. There are two main arguments for blower coolers... Low power GPUs and Multi-GPU configuration. And the latter is debatable.

AMD has been slammed for their blower coolers for ages. That includes the 5700 cards, which aren't even that power hungry. Going this route is a good thing. I am even considering buying a reference card, depending on the reviews. I normally never buy reference cards.
 
I the latest totally-not-a-youtube video from totally-not-RedGamingTech, totally-not-Paul doubled down on the 256bit GDDR6 + 128MB of Infinity Cache. He now claims he has more trustworthy sources backing that up.
He also claims the 128MB Infinity Cache is not L2, and it isn't counting with L2 to reach the 128MB capacity. It's L3 victim cache or something exclusively dedicated to specific blocks of the GPU (ROPs? framebuffer only?) or something else entirely.
Also, the maximum clock speeds are reportedly close to the PS5's 2.23GHz, because like Cerny stated going above that creates problems with GPU functionality (not related to power-clock curves).

There's also word of a RDNA2 GPU with HBM for mobile, though I think that might be little more than conjecture based off Vega 12 and Navi 12.



xEx said:
Isn't that just a render?
Click to expand...
Well at the moment I'm pretty sure it's just pixels on my monitor.


Kaotik said:
The biggest question at the moment is is whether that USB-C is practically dead VirtualLink or something else using USB-C
Click to expand...
[Wishfull mode=ON]
It's a USB4 port at 40Gbps that people can use to directly connect external NVMe drives and make use of the embedded hardware decompressor and DirectStorage to get I/O results similar to the new consoles. The GPU can take advantage of both data coming from the PCIe connector and the dedicated USB4 connection.
 
I am quite skeptical both about the "Infinity cache" than the "clock instability" over PS5 clocks, also because in the same video of RGT there are screens of the PS5 presentation where it is explicitely stated that PS5's CUs are diffrent respect to desktop RDNA2 Cu's and that the latter are quite "beefier" than the PS5's ones. So as they are different, they are quite unlikely to have the same issues if not process related.
 
ToTTenTranz said:
[Wishfull mode=ON]
It's a USB4 port at 40Gbps that people can use to directly connect external NVMe drives and make use of the embedded hardware decompressor and DirectStorage to get I/O results similar to the new consoles. The GPU can take advantage of both data coming from the PCIe connector and the dedicated USB4 connection.
Click to expand...

I dont think you have to wish, NV claimed speeds/IO tech similar and above what new consoles have. RDNA2/AMD sure does that too, in special considering they designed most of the consoles.

Leoneazzurro5 said:
I am quite skeptical both about the "Infinity cache" than the "clock instability" over PS5 clocks, also because in the same video of RGT there are screens of the PS5 presentation where it is explicitely stated that PS5's CUs are diffrent respect to desktop RDNA2 Cu's and that the latter are quite "beefier" than the PS5's ones. So as they are different, they are quite unlikely to have the same issues if not process related.
Click to expand...

Maybe at 36CU (low to mid end) dGPU parts we might see 2.3Ghz or higher. A 80+CU gpu is never going to clock that high.
 
Leoneazzurro5 said:
I am quite skeptical both about the "Infinity cache" than the "clock instability" over PS5 clocks, also because in the same video of RGT there are screens of the PS5 presentation where it is explicitely stated that PS5's CUs are diffrent respect to desktop RDNA2 Cu's and that the latter are quite "beefier" than the PS5's ones. So as they are different, they are quite unlikely to have the same issues if not process related.
Click to expand...
What screens where now?
 
XSX has 76 MB of cache, if you take out the CPU cache it is around 64 MB for the 2 Shader Engines.
For a 4 SE 80 CU Navi21, 128 MB is nothing extra compared to the 64MB for 2 SE (32MB/SE) that XSX has.
At Hot Chips though, MS never said a thing where those 64MB of cache were located or used for.
 
ethernity said:
XSX has 76 MB of cache, if you take out the CPU cache it is around 64 MB for the 2 Shader Engines.
Click to expand...
Why are you assuming 12MB of cache for the CPU? 8MB L3 + 512KB L2 per core? Don't the L1 count for ESRAM?

Desktop Zen2 has 32MB L3 + 8x (512KB + 64KB) = 36MB "GameCache".
That would leave 40MB total cache for the 52CUs iGPU, which when compared to Navi 10's 4.5MB total ESRAM it's already a beast.
IMO there's little reason to believe the CPU in the SeriesX doesn't have all 32MB L3 of the desktop/server version, especially since Microsoft will be using the SoC for Azure and XCloud.


EDIT: I was corrected. Microsoft said they're using the "Renoir" version of Zen2 L3 arrangement.
 
Last edited by a moderator:
ethernity said:
At Hot Chips though, MS never said a thing where those 64MB of cache were located or used for.
Click to expand...
There's maybe about 40MB of cache that are more easily accounted for between the CPU and GPU bits, but at least on the die shot, you can make out a fair bit of cache-like structures on the (Infinity) Fabric marked areas. There's a bunch of other cache for the multimedia blocks as well.


edit: maybe 54MB, including register file
 
Last edited:
Status
Not open for further replies.
Back
Top