AMD: Speculation, Rumors, and Discussion (Archive)

[Kaotik]

Wouldn't have been the first time that this is solved by a simple shader rewrite. I don't know the precise software architecture, but it is safe to assume that there is at least some type of hardware abstraction / driver layer, right?

The problem comes from the fact that Neo is supposed to play unpatched PS4-games too

 

[Ext3h]

The problem comes from the fact that Neo is supposed to play unpatched PS4-games too

Live patching?

 

[Deleted member 2197]

RX 460 AMD Labs benchies

http://www.3dcenter.org/news/amd-ei...en-zu-viel-performance-fuer-die-radeon-rx-460

 

[Deleted member 13524]

1.7 to 2x faster than a full Bonaire, which has the same number of CUs and clocks at 1.1GHz?

Unless this P11 is clocking close to 2GHz, those numbers are ridiculous..


EDIT: Looking at these Overwatch scores, I wonder what settings they used to bring the R7 260X down to 30 FPS. That card does 40 FPS at 1440p and ultra settings..

 

[3dilettante]

Live patching?

If code for the current PS4 uses functions and formats dropped by GCN3, doing it without a more global developer evaluation might break something. The performance or behavior might not come out as expected, although AMD must have felt the formats and features it removed were not critical.

One comparatively arcane distinction between the PS4's IP level and GCN3/4 is that wavefront GPR allocation granularity changed.
http://gpuopen.com/amdgcn-assembly/

In current GCN3 hardware, VGPRs are allocated in groups of 4 registers and SGPRs in groups of 16. Previous generations (GCN1 and GCN2) have a VGPR granularity of 4 registers and an SGPR granularity of 8 registers.

It would be awkward in some high-occupancy case to find that the SGPR load for a scalar-light shader has automatically doubled, although that might be what hardware customization is for.

 

[Alexko]

RX 460 AMD Labs benchies

http://www.3dcenter.org/news/amd-ei...en-zu-viel-performance-fuer-die-radeon-rx-460

If true, that would be pretty decent given that the 260X draws about 90~100W.

 

[xEx]

Good performance for a 100 bucks 75W card. Although iirc Pros like to use 144Hz monitors, ergo play at 144FPS. I also dont really understand the "Esport" focus on a budget card.

 

[Silent_Buddha]

Good performance for a 100 bucks 75W card. Although iirc Pros like to use 144Hz monitors, ergo play at 144FPS. I also dont really understand the "Esport" focus on a budget card.

Not all Pro's or aspiring Pro's make a lot of money. The ones that are successful can get sponsorships. Depending on their sponsor they can get free hardware (sponsored by an IHV or AIB, for example). But most "professional" players don't make a lot of money, and people that are attempting to break into the pro leagues don't often have much money (high school and college students).

Hack many of the Korean Pro players got their start in PC Bangs because they couldn't afford their own PC.

Regards,
SB

 

[kalelovil]

1.7 to 2x faster than a full Bonaire, which has the same number of CUs and clocks at 1.1GHz?

Unless this P11 is clocking close to 2GHz, those numbers are ridiculous..

1GB 260X running into memory issues? Or maybe power limited through the driver to 75W.

 

[xEx]

1GB 260X running into memory issues? Or maybe power limited through the driver to 75W.

IDK in OW but for LoL and Dota 1GB is more than enough. LoL is a DX9 game...

 

[Deleted member 13524]

It seems the non-reference RX 480 models are coming with substantial factory overclocks, which makes sense since they're not bundling that tiny feeble heatsink from the reference models:
http://rog.asus.com/23932016/gaming-graphics-cards-2/overview-rog-strix-rx-480/

They're even taking advantage of the PEG power thing:

 

[Silent_Buddha]

And as we all know from the numerous Nvidia users saying it over and over. The reference design doesn't matter, only custom AIB cards should be used for benchmarking and comparison. So maybe the PCIE thing will finally be put to rest.

Probably not. And I myself still think reference should be compared to reference in benchmarks with custom boards ignored for launch reviews.

Regards,
SB

 

[CSI PC]

I am a supporter of custom AIB over reference models myself, whether AMD or Nvidia.
One alarm bell with that Asus marketing, they specifically only show 4k.....
And with interesting fps even for reference in that Asus slide; majority will be using this card at 1080p, and some at 1440p, with niche (in comparison) being mGPU 4k or high fps 1440p.
Cheers

 

[Malo]

Note: GPU core power comes from the PCIe slot and through the external PCIe power connector

Isn't that in contention with the STRIX claim of 0% power from PCIe slot for GPU?

 

[AnarchX]

Some one made accidentally a die shot of Polaris 10: https://imgur.com/a/ad01x
So there only 36 CUs in silicon.

 

[Grall]

That will buff out. No worries.

 

[DieH@rd]

Easy fix. Just wrap GPU in a big towel and let it running for a few minutes.

 

[Moloch]

Some one made accidentally a die shot of Polaris 10: https://imgur.com/a/ad01x
So there only 36 CUs in silicon.

Some people shouldn't be allowed near computers

 

[Razor1]

WOW how the hell......

 

[fellix]

The chap used the included thermal "cement" glue in the cooler's kit, that wasn't meant to be applied on the GPU at all, but only for the memory chips and VRMs.
This cooler (Accelero Xtreme IV) is notorious for the rather difficult installation procedure and it's definitely not recommended for the inexperienced users.