Monday, December 12th 2016

AMD "Vega" Demoed in Sonoma, California

by
btarunr
 Discuss (120 Comments)
AMD's next-generation high-end graphics card, based on the "Vega" architecture, was showcased at an event in Sonoma CA, earlier this week. While the architecture is being debuted with the Radeon Instinct MI25 deep-learning accelerator, a prototype graphics card based on the silicon was exhibited by the company, showing Vulkan API gaming.

AMD was pretty tight-lipped about the specifications of this prototype, but two details appear to have slipped out. Apparently, the chip has a floating point performance of 25 TFLOP/s (FP16), and 12.5 TFLOP/s (FP32, single-precision). On paper, this is higher than the 11 TFLOP/s (FP32) of NVIDIA TITAN X Pascal. The other important specification that emerged is that the card features 8 GB of HBM2 memory, with a memory bandwidth of 512 GB/s. This, too, is higher than the 480 GB/s of the TITAN X Pascal. It remains to be seen which market-segment AMD targets with this card.

This article was updated on Dec 15 to accommodate AMD's request to remove all info regarding the demo system, the shown game and its performance, which has been put under NDA retroactively.
Source: Golem.de

Related News

Add your own comment

120 Comments on AMD "Vega" Demoed in Sonoma, California

#27
cdawall
where the hell are my stars
the54thvoidThreads were merged - parallel universes!!
Mods must have gotten tired of us not using the news posts :D
Posted on Reply
#28
prtskg
m1dg3t^^ Vega 10 is small Vega, Vega 11 is the 'big boy' IIRC
You remember it right. This is what AMD said. Though I wonder if the card AMD showed is Vega 10 or 11. Time to read more.
Posted on Reply
#29
Nergal
IF it turns out to be faster than the 1080 in DX11 (which I doubt); why shouldn´t they also price it more expensive?

Apparently people think that if something is more costly, it should also be better. It has worked for NV till now....


Ofcourse, AMD won´t do that and will put it 100 bucks cheaper, between the cost of the 1070 and 1080, having that score in DX11.

AMD train is arriving a 2nd time. First one was shrugged of by NV fanboys (the 1060 being 12% faster...which is no longer true!); and now this.
Even if the Zen is only somewhat decent; a lot of people will be going full red.
Posted on Reply
#30
Ferrum Master
cdawallMods must have gotten tired of us not using the news posts :D
The news were late. Some us are usually faster.
Posted on Reply
#31
EarthDog
Don't steal the thunder that hasn't clapped yet!!!!!!!!
Posted on Reply
#32
Fx
I look forward to going back to AMD for my graphics needs if they have competitive pricing, a reasonable power envelope and most importantly, a quiet card.

Basically because I like a cool and quiet man cave.
Posted on Reply
#33
john_
Ferrum MasterThe news were late. Some us are usually faster.
The newsposter writes an article. It takes time.

The forum member just throws a title and a link. Pretty easy and fast.
Posted on Reply
#34
FordGT90Concept
"I go fast!1!11!1!"
So... the 512 GiB/s HBM2 is suspicious because that's what Fury X had and HBM2 has double the bandwidth of HBM1:


Me thinks this card has two stacks of 4 GiB each which leaves the door open to a 1 TiB/s card with 16 GiB in the future (a workstation card)...
Posted on Reply
#35
prtskg
There is dual vega too. That's expected though. Dual fiji brought AMD some good contracts.
Edit - I read wrong. Dual vega is not yet shown, if there is such a thing.
Posted on Reply
#37
prtskg
2 Vegas, 1 with 12.5 TF FP32 compute, other with 8.2 TF FP32 compute. These are professional versions though. So consumer cards should have higher frequency aka more compute capability.
Edit - Only 1 vega shown, others cards shown are polaris and nano(fiji).
Posted on Reply
#38
FordGT90Concept
"I go fast!1!11!1!"
prtskgThere is dual vega too. That's expected though. Dual fiji brought AMD some good contracts.
That could theoretically have 32 GiB of RAM and 1 TiB/s each.
Posted on Reply
#39
gupsterg
NergalOfcourse, AMD won´t do that and will put it 100 bucks cheaper, between the cost of the 1070 and 1080, having that score in DX11.
Well hope it is cheaper.

Post 4 had some compares, IIRC R9 290X was against GTX 780, launch prices$549 vs $649 . Then Fury X was against 980/Ti, $649 vs $549/$649. It took a fair time before Fiji cards were cheaper than say a 980 Ti in the UK @ etailers. I was toying with the idea of getting a 980 Ti, basically the £50 cheaper price of Fury X and then say future purchase of FreeSync monitor vs G-Sync swayed it for me. So I think it will be closely priced to nVidia 1070/1080 and the lack of premium for a FreeSync monitor vs G-Sync will be the sweetener for a buyer.

The "instinct" slide shows Vega has HBM, regardless of if it's HBM 1 or 2 I would think it's more costly compared with GDDR5X, plus the interposer for GPU/HBM must add cost to card production price?
Posted on Reply
#40
Ferrum Master
john_The newsposter writes an article. It takes time.

The forum member just throws a title and a link. Pretty easy and fast.
Speed of them gossips is a remarkable phenomena.
Posted on Reply
#41
the54thvoid
Intoxicated Moderator
prtskgYou remember it right. This is what AMD said. Though I wonder if the card AMD showed is Vega 10 or 11. Time to read more.
Source?

In fact, Anandtech had an article a while back saying 4096 cores is about as much as the architecture can take, Fiji or otherwise. So, if this Vega 10 is touted at 12.5 T/flops at a certain speed (1525Mhz), it works out at 4096 shaders.
Instinct MI25 is advertised as 25TFLOPs half-precision computing monster that will compete with Tesla P100 deep-learning computing accelerator. This means that single-precision (FP32) raw performance is 12.5 TFLOPs. We concluded that if Instinct MI25 would be clocked at 1525 MHz, this would give us 4096 Stream Processors.
videocardz.com/64706/amd-vega-doom-4kultra-gaming-performance-demo-possible-specs

I don't see Vega 11 being bigger than 4096 shaders.
Posted on Reply
#42
TheGuruStud
the54thvoidSource?

In fact, Anandtech had an article a while back saying 4096 cores is about as much as the architecture can take, Fiji or otherwise. So, if this Vega 10 is touted at 12.5 T/flops at a certain speed (1525Mhz), it works out at 4096 shaders.



videocardz.com/64706/amd-vega-doom-4kultra-gaming-performance-demo-possible-specs

I don't see Vega 11 being bigger than 4096 shaders.
Probably as big as it can be on 28nm. Although, I doubt they would go too big. Yields, yields, yields.
Posted on Reply
#43
renz496
NergalIF it turns out to be faster than the 1080 in DX11 (which I doubt); why shouldn´t they also price it more expensive?

Apparently people think that if something is more costly, it should also be better. It has worked for NV till now....


Ofcourse, AMD won´t do that and will put it 100 bucks cheaper, between the cost of the 1070 and 1080, having that score in DX11.

AMD train is arriving a 2nd time. First one was shrugged of by NV fanboys (the 1060 being 12% faster...which is no longer true!); and now this.
Even if the Zen is only somewhat decent; a lot of people will be going full red.
in general they probably still is. hardware canucks made conclusion from average combined tittle of DX11. since they add more titles that favoring amd cards the end result was swaying towards AMD cards more for the average score. it does not mean RX480 are now as fast as 1060 in every DX11 titles. i just check TPU numbers for RX480 and GTX1060 from launch review and recent review that include both card. while i see some performance increase on RX480 the same also happen with 1060. so in the end 1060 still end up being ahead of RX480 in TPU relative performance chart.
Posted on Reply
#44
RejZoR
iOIts an early prototype running with unoptimized drivers from Fiji so theres quite some untapped performance to gain till June.
Plus all case vents were taped to keep people from taking pictures from the insides so the card might be even throttling a bit....
Wait a second, June? JUNE? Not another Fiji perpetually delayed thing all over again while NVIDIA will be almost rolling out Volta based graphic cards at that point...
Posted on Reply
#45
iO
RejZoRWait a second, June? JUNE? Not another Fiji perpetually delayed thing all over again while NVIDIA will be almost rolling out Volta based graphic cards at that point...
Vega10's milestones are pretty similiar to Polaris: First tape-out in summer. First "final silicon" end of November/beginning of December. And launch at Computex...

Lets hope they can speed things up this time...
Posted on Reply
#46
HD64G
prtskg2 Vegas, 1 with 12.5 TF FP32 compute, other with 8.2 TF FP32 compute. These are professional versions though. So consumer cards should have higher frequency aka more compute capability.
MI8 is a Nano, not any new core. Only new one is the MI25, probably 2XVega10 cores into it for 12,5GFlops/s for each one of them.

And I beg to differ about the launch date of Vega 10. Most possibly in January 2017.
Posted on Reply
#47
Xzibit
HD64GMI8 is a Nano, not any new core. Only new one is the MI25, probably 2XVega10 cores into it for 12,5GFlops/s for each one of them.

And I beg to differ about the launch date of Vega 10. Most possibly in January 2017.
MI25 is not a dual gpu

Raja Koduri answered that on twitter
Posted on Reply
#48
renz496
HD64GMI8 is a Nano, not any new core. Only new one is the MI25, probably 2XVega10 cores into it for 12,5GFlops/s for each one of them.

And I beg to differ about the launch date of Vega 10. Most possibly in January 2017.
if they really going to launch in january (which depending on time could be less or more than 1 month from now) then why they showcase the product using Fiji drivers? if they really that close to launching there should be at least more specific driver for it? at least that what i would expect.
Posted on Reply
#49
the54thvoid
Intoxicated Moderator
HD64GMI8 is a Nano, not any new core. Only new one is the MI25, probably 2XVega10 cores into it for 12,5GFlops/s for each one of them.

And I beg to differ about the launch date of Vega 10. Most possibly in January 2017.
The 25GFlops is double precision, the 12.5GFlops is touted as Vega's single power (higher than Titan XP). I think, if my reading was correct.
Posted on Reply
#50
W1zzard
cdawallMods must have gotten tired of us not using the news posts :D
I did it manually, had to time travel a bunch of posts :)
Posted on Reply
Add your own comment
Apr 26th, 2024 16:31 EDT change timezone

Latest GPU Drivers

New Forum Posts

Popular Reviews

Controversial News Posts