New Upcoming ATI/AMD GPU's Thread: Leaks, Hopes & Aftermarket GPU's

Yes that image is well known. Problem is that it could also be explained as a use case scenario for their conservative rasterization T3. With T3 you will have another PS instruction telling you if that pixel has been fully drawn. With this information you can do very precise per-pixel culling. The problem is that you need to build your application to support that. Also I don't believe you need tile based rasterizer to support this. It would be just much more inefficient.

 

Well, According to AMD this feature is not always enabled. It is scenario based alternative.
Regular processing is still present. So it means that it is less efficient in some situations.

 

If you mean DSBRasterizer, then this should be enabled at least for all render calls drawing primitives with shaders containing that CR T3 input( SV_InnerCoverage ). It would be interesting to see how taxing that actually is. I wouldn't be too surprised if bad usage would slow the rendering down

 

:eek2:

 

Perfect culling + drawing for: Verticles/Triangles/Texels/Pixels <- But it will have to pack some sort of next Gen prediction engine called ReliVe Advanced AI.

It's all academic :nerd:

16 dayZ

 

AMD Vega Hybrid Mod: Fixing Power Leakage for Higher OC

http://www.gamers nexus.net/guides/2986-amd-vega-hybrid-mod-results-overclocking-liquid-vs-air

 

Now that we see what 16bit support and CR T2/T3 did to VEGA, it would be interesting to see what changes will be required for Volta to support these. Luckily for NV, Pascal already supports CR T1/T2 so the transition might be less painful.

 

Here some Vega FE 1700/1100 WC benches:




Some game Tests (i did not include DOOM 4K becouse they avoid 8xAA, on ATI it has no impact THX to Async You know then VEGA will be on TOP ! so they pass )
-> Also DOOM has "only" 2 things in Async Compute: AA + AO :nerd: <-

Also on VEGA RX XTX it will be working HBCC so the minimum FPS 1%/0.1% in any Game will demolish us on 1440p & 4K !
Will be awesome if they can give us this HiTek for ~399-499€ in stores !
Like back then -> http://www.anandtech.com/show/2069
( http://www.anand|tech.com/show/2069 )

Spoiler

 

So on water and overclocked it's about reference 1080 performance?

What's wrong with the Doom results, the post above mentions it has the most impressive results from the OC.

On my phone, and none of the links above are working for me.

 

ATI/AMD will showcase its upcoming RX Vega GPU to the public for the first time on July 18th

AMD has announced that it will let the public get its hands on its new RX Vega graphics card in Budapest on July 18th. As AMD noted, on Tuesday, July 18 from 2 to 7 PM, it will be transforming the Akvárium Klub into a PC gaming zone in which it will be bringing rigs powered by Radeon RX Vega with today’s most popular games for gamers to try.

Although AMD did not reveal the games that will be showcased, our guess is that the red team will let the public test games that favour its graphics cards, such as DOOM, Sniper Elite 4, Battlefield 1 and PREY. Moreover, AMD did not reveal whether gamers will be able to use any framerate counters (you know, in order to see how these cards perform) or whether these systems will feature a single or two RX Vega GPUs in Crossfire.

AMD will also showcase the RX Vega in Portland (PDXLAN) on July 21st to July 23rd and in Los Angeles during SIGGRAPH 2017, when we expect the red team to official release it.

Spoiler

 

-> http://www.game|rsnexus.net/guides/...mod-results-overclocking-liquid-vs-air/page-3

just remove dash game|nexus

 

Am I in for rough awakening? I'm hoping the card would be priced in 450-500 range, so it would compete with both the 1080 and 1070 buyers, and I would also get insane performance for a buck. Will it be 450-500 card or a 600 card?

Pure speculation

... I guess this falls down to the "hopes" part of the topic

 

Already stated, there will be 3 cards. I expect pricing competitive with nVidia cards around similar performance.

AMD's GPUs are not in same bad spotlight CPUs had. Because Bulldozer and its children were not exactly good for gaming paired with anything above average from start. And people are used to have GPU as bottleneck, not CPU.

But, as people are used to have GPU bottleneck and strong GPUs are quite expensive, most of people do not care about AMD not having $600 competition. Therefore AMD's GPU image is far from suffering way Bulldozer did. And that's why I do not expect lower price tag for same performance as nVidia's cards.

 

Depends on how aggressive AMD wants to penetrate the market, and also how much the cost production is for them. I think there is going to be some undercutting going on, like how Vega FE undercuts Quaddro cards.

 

I'm guessing now, its possibly $599 for the top aircooled version, and $699 for the watercooled version, at least, if not $749 due to the watercooler and rad/fan.
Gimped aircooled version at $499, or higher, due to being custom boards.

 

GCN 5, the future with RX Vega

Seen in: Vega 11, Vega 10
Vega Frontier Edition, RX Vega and more!

This new design is built around the ‘Next-Generation Compute Unit’ (NCU) and is expected to increase instructions per clock as well as offer higher clock speeds in general. The NCU is a refinement of GCN’s compute units and not a completely new design.
Vega also supports High Bandwidth Memory 2 (HBM2), which can offer (if used with 4 stacks!) a mind-boggling 1 terabytes per second of pure bandwidth and 32 GBs of VRAM. Whether its full capacity will be used on new cards isn’t yet known.
A new, larger memory address space, and the High Bandwidth Cache Controller are also slated to appear.

Additionally, the new GPUs are expected to include improvements in the Rasterization and Render output units. This should in theory greatly help Virtual Reality and high resolution and refresh rate gamers. The stream processors are heavily modified from the previous generations to support packed math Rapid Pack Math technology for 8-bit, 16-bit, and 32-bit numbers.
With this, there is a significant performance advantage when lower precision is acceptable (for example: processing two half-precision numbers at the same rate as a single precision number). In fact, it almost doubles the effective GPU power if used correctly and can lead to tremendous gains in performance!
This technology is for now unused in games, but it will be of great service to gamers down the line as well as content creators.
A look at the future

There are many unknowns about Vega, which is set to be the biggest revision GCN has ever gotten. We also have almost no idea what AMD will be introducing with Navi. GCN is a storied architecture with impressive engineering prowess. Sure, the core behind the beast is the same, but many enhancements and improvements have added up over time, making for a faster, cheaper, more efficient and feature-rich product.

 

At 1700mhz the FE is pulling around 400watts from the PCI-e cables, not accounting for the PCI-e slot. That is insane power usage.

All that cleared away, the power leakage reduction under liquid was non-trivial, and granted us our additional ~40MHz of OC headroom on the core (by way of granting ~30W more to the core). Under max consumption and overclock, we measured current at the PSU cables to be ~33.3A at 12V, or just under 400W (+/-2% error). Given that there’s also power going through the PCIe slot, this is one of the most power-hungry overclocks we’ve managed without modding BIOS to increase the power limit. The gains in thermals were tremendous, ultimately affording the reduction in power leakage (and overall increased power available to the core).

http://www.***********.net/guides/2986-amd-vega-hybrid-mod-results-overclocking-liquid-vs-air/page-3

 

Found the memory straps for the FE based on a unverfied bios on tpu


memory ic listed KHA883901B_CL17_WL6

starts around 95a6

167mhz

Spoiler

3C 41 00 00 0A 06 03 03 09 09 03 03 02 02 00 05 08 00 00 00 03 03 05 00 04 00 00 00 00 02 00 00 03 02 01 03 03 02 01 03 01 0D 00 00 8C 02 00 00 08 0C 00 00 2E 00 00 00 00 00 11 00 04 00 00 00 55 14 0D 01 1B 00 07 00 04 00 02 05 09 01 0C 01 00 00 02 00 3B 00 00 00 44 04 00 00

300mhz

Spoiler

30 75 00 00 0B 0A 05 03 0F 0F 05 05 02 02 00 05 08 00 00 00 03 03 05 00 06 00 00 00 00 03 00 00 03 02 01 03 03 02 01 03 01 0E 00 00 92 04 00 00 08 0C 00 00 51 00 00 00 00 00 11 00 04 00 00 00 55 14 0E 01 30 00 07 00 04 00 02 05 0A 01 0C 01 00 00 02 00 69 00 00 00 44 04 00 00

500mhz

Spoiler

50 C3 00 00 0C 11 08 05 18 18 07 07 02 03 00 05 08 00 00 00 04 03 05 00 09 00 00 00 00 04 00 00 03 02 01 03 03 02 01 03 01 0E 00 00 9E 07 00 00 08 0C 00 00 87 00 00 00 00 00 11 00 04 00 00 00 55 14 11 01 50 00 07 00 04 00 02 05 0B 02 0C 01 00 00 02 00 AF 00 00 00 44 04 00 00

800mhz

Spoiler

800mhz
80 38 01 00 11 1B 0D 08 27 27 0C 0C 04 05 00 05 0D 00 00 00 06 03 07 00 0D 00 00 00 00 07 00 00 03 02 01 03 03 02 01 03 01 11 00 00 30 0C 00 00 08 0C 00 00 D8 00 00 00 00 00 11 00 06 00 00 00 86 20 18 01 80 00 07 00 06 00 02 06 10 04 0C 01 00 00 02 00 18 01 00 00 44 04 00 00

1000mhz

Spoiler

A0 86 01 00 11 21 10 0A 2F 2F 0E 0E 04 06 00 05 10 00 00 00 06 04 09 00 10 00 00 00 00 08 00 00 03 02 01 03 03 02 01 03 01 11 00 00 3C 0F 00 00 08 0F 00 00 0E 01 00 00 00 00 11 00 08 00 00 00 A8 28 1C 01 A0 00 07 00 08 00 02 08 10 04 0C 01 00 00 02 00 5E 01 00 00 44 04 00 00

will be interesting to see how these compare to the RX variants

 

@user1: Looking at those, I think that reverse engineering those values will be considerably easier than HBM1, as there we had effectively just 3 straps and some AMD's settings made no sense once compared to tech sheet.

 

^^ YES -> New MOD for VEGA HBM coming