'

Linux DRM driver update reveals Vega specifications

Linux DRM driver update reveals Vega specifications

Linux DRM driver update reveals Vega specifications

Linux DRM driver update reveals Vega specifications

 
In a new update that has been submitted for the Linux 4.12 kernel, there has been a huge update to the Digital Rendering Manger (DRM) which includes support for AMD's Vega architecture. 
 
These updates include clear references to Vega's specifications, with entries like "gfx.config.max_shader_engines = 4," which tells us that Vega has four shader engines and "Adev-> gfx.config.max_cu_per_sh = 16" which signified that Vega has 16 GCN (or in the case of Vega NCU) compute units. 
 
If Vega has the same number of stream processors per compute unit as Fiji (64), this will give Vega a total of 4096 shaders. Other entries indicate that each shader engine has two Asynchronous Compute Units, one render back-end and 16 Render Output Units, which will give Vega a total of 64 ROPs. 

When compared to AMD's Polaris 10 RX 570, Vega 10 will feature 2x the Stream processors, texture mapping units and render output units, which when combined with Vega's architectural changes should create a very powerful GPU.

 

 RX 570RX 580RX Vega
GPUPolaris 10 ProPolaris 10 XTVega 10
Process node14nm14nm14nm
Shade Engines444
CUs per shader Engine8916
Shaders Per CU646464 (expected)
Stream Processors204823044096
Performance (FP32)5.095 TFLOPs6.175 TFLOPs12.5 TFLOPs
Performance (FP16)5.095 TFLOPs6.175 TFLOPs25 TFLOPs
Render Output Units323264
Texture Mapping units126144256
Hardware Threads448
Memory Interface256-bit256-bit2048-bit
Memory4/8GB GDDR54/8GB GDDR58GB HBM2

 

Looking at AMD's previously announced Vega Instinct MI25, we see that AMD claims that the GPU offers around 12.5TFLOPs of FP32 compute performance given AMD's MI naming scheme.   

If AMD's new NCUs come with the same stream processor count as AMD's GCN/Polaris GPUs, this will place Vega's clock speeds at around  1526MHz, which is a significant boost when compared to both Polaris and Fiji.  As you can see below, AMD's MI25 accelerator is said to be passively cooled, which means that Vega should run cool and that higher clock speeds should be possible if users are willing to increase Vega's heat output. 

As you can see below, AMD's MI25 accelerator is said to be passively cooled, which means that Vega should run cool and that higher clock speeds should be possible if users are willing to increase Vega's heat output. 

 

Linux DRM driver update reveals Vega specifications  

At this time these specifications remain unconfirmed by AMD, but there is no reason for AMD to lie in their Linux submissions. For the most part, these specifications confirm what was already expected from Vega, which is a high GPU core count and clock speed that are higher than what is available with AMD's existing Polaris offerings. 

 

You can join the discussion on Vega's specifications on the OC3D Forums

 

«Prev 1 2 Next»

Most Recent Comments

03-05-2017, 19:14:20

Kleptobot
These numbers seem pretty promising. Would be interesting to see what the tdp was tooQuote

03-05-2017, 20:29:13

NeverBackDown
Rumored for 225watts. Less than Fury X but I suspect it'll be 2x-2.5x faster. So it's power efficiency will be way way up.Quote

04-05-2017, 01:22:10

Kleptobot
I wonder how much extra performance per cycle these will have over the rx580. Seeing as a 1080ti is about twice as fast as a 580 I think this is an interesting line of questioningQuote

04-05-2017, 01:59:37

NeverBackDown
Quote:
Originally Posted by Kleptobot View Post
I wonder how much extra performance per cycle these will have over the rx580. Seeing as a 1080ti is about twice as fast as a 580 I think this is an interesting line of questioning
Well leaked slideshows say 2x performance per clock. So it should be an interesting battle. And with supposedly much increase clockspeeds along with other improvements, should hopefully be closeQuote

04-05-2017, 04:57:39

WYP
Quote:
Originally Posted by NeverBackDown View Post
Well leaked slideshows say 2x performance per clock. So it should be an interesting battle. And with supposedly much increase clockspeeds along with other improvements, should hopefully be close
By leaked slideshow I hope you don't mean that WccfTech April Fools nonsense. Most early slides are official now.

Are you referring to the performance improvements on the Geometry pipeline? That does not mean 2x the throughput of shaders.

https://www.overclock3d.net/gfx/arti...124027868l.jpg

https://www.overclock3d.net/news/gpu...ral_analysis/1Quote
Reply
x

Register for the OC3D Newsletter

Subscribing to the OC3D newsletter will keep you up-to-date on the latest technology reviews, competitions and goings-on at Overclock3D. We won't share your email address with ANYONE, and we will only email you with updates on site news, reviews, and competitions and you can unsubscribe easily at any time.

Simply enter your name and email address into the box below and be sure to click on the links in the confirmation emails that will arrive in your e-mail shortly after to complete the registration.

If you run into any problems, just drop us a message on the forums.