Monday, April 30th 2018

AMD "Vega 20" with 32 GB HBM2 3DMark 11 Score Surfaces

With the latest Radeon Vega Instinct reveal, it's becoming increasingly clear that "Vega 20" is an optical shrink of the "Vega 10" GPU die to the new 7 nm silicon fabrication process, which could significantly lower power-draw, enabling AMD to increase clock-speeds. A prototype graphics card based on "Vega 20," armed with a whopping 32 GB of HBM2 memory, was put through 3DMark 11, on a machine powered by a Ryzen 7 1700 processor, and compared with a Radeon Vega Frontier Edition.

The prototype had lower GPU clock-speeds than the Vega Frontier Edition, at 1.00 GHz, vs. up to 1.60 GHz of the Vega Frontier Edition. Its memory, however, was clocked higher, at 1250 MHz (640 GB/s) vs. 945 MHz (483 GB/s). Despite significantly lower GPU clocks, the supposed "Vega 20" prototype appears to score higher performance clock-for-clock, but loses out on overall performance, in all tests. This could mean "Vega 20" is not just an optical-shrink of "Vega 10," but also benefits from newer architecture features, besides faster memory.
Source: VideoCardz
Add your own comment

71 Comments on AMD "Vega 20" with 32 GB HBM2 3DMark 11 Score Surfaces

#1
dj-electric
Maybe an experiment, maybe workstation \ server related.
Of course that home users won't get to taste that hefty 32GB config. Whhaayyyy too expensive and serves very little.
Posted on Reply
#2
Vayra86
Hmm. We're totally not sure if 1 Ghz was the actual core clock with an unsupported VGA. The value of this bench seems rather low.
Posted on Reply
#3
jabbadap
Both clocks are really unknown. That 3dmark SystemInfo does not recognize the card so it gives them some generic clocks.
Posted on Reply
#4
ShurikN
Those clocks mean nothing as they were most likely incorrectly read by the software.
Posted on Reply
#5
nemesis.ie
Is the throughput correct? With 32GB they might have more than 2 stacks, so the bandwidth could well be more than what's in the article?

But if they are keeping the footprint the same, maybe not. The Ryzen 2000s are the same physical silicon size as the 1000 series.
Posted on Reply
#6
RejZoR
Well, Vega Frontier has 16GB VRAM on 2 stacks. For 32GB, they'd need 4 stacks, meaning it would actually have double the bandwidth. And HBM2 on it seems to have higher clock too compared to old Vega chips. But still, gaming wise, memory bandwidth wasn't really a limiting factor, it was more GPU's throughput on pixel level.
Posted on Reply
#7
nemesis.ie
Exactly. Maybe it's a 32GB mining card, that would put 2x+ bandwidth to use. ;)

Then the consumer/game version can "just" have half the b/w and the mining folks can buy the pricier 32GB/high memory speed one ...
Posted on Reply
#8
lynx29
That is a shame my 1080 Ti still beats it in graphics score by 3000 points. I guess I will be waiting another year or two and hoping for the best, as usual.
Posted on Reply
#9
Countryside
nemesis.ie said:
Exactly. Maybe it's a 32GB mining card, that would put 2x+ bandwidth to use. ;)

Then the consumer/game version can "just" have half the b/w and the mining folks can buy the pricier 32GB/high memory speed one ...
Yeah a 32gb hbm2 for gaming is unlikely and an overkill
keeping in mind the memory price.
Posted on Reply
#10
nemesis.ie
lynx29 said:
That is a shame my 1080 Ti still beats it in graphics score by 3000 points. I guess I will be waiting another year or two and hoping for the best, as usual.
But if this b/m is real, the card could well be running at least 70% slower than it is capable of, if you add 70% to the score, then how does it compare to the 1080Ti?
Posted on Reply
#11
jabbadap
lynx29 said:
That is a shame my 1080 Ti still beats it in graphics score by 3000 points. I guess I will be waiting another year or two and hoping for the best, as usual.
P -score is with 720p resolution, quite meaningless test for modern graphics cards I would say...
Posted on Reply
#12
lynx29
nemesis.ie said:
But if this b/m is real, the card could well be running at least 70% slower than it is capable of, if you add 70% to the score, then how does it compare to the 1080Ti?
I hope you are right my friend, I want nothing more than to support AMD with Ryzen 2 and Vega 2 in 2019. I am tired of toothpaste and gimmicks, and Nvidia's GPP bullying kind of disgusts me frankly. Nvidia has had 0 issues with branding of GPU's until now, they hold 80% of market share, why be so afraid. It's just so dumb, I hate bullying of any kind.
Posted on Reply
#13
RejZoR
Countryside said:
Yeah a 32gb hbm2 for gaming is unlikely and an overkill
keeping in mind the memory price.
That's for sure. Even 1080Ti's 11GB is rarely really utilized. So, even 16GB is overkill. This card is clearly aimed at productivity use and not gaming.
Posted on Reply
#15
Countryside
Vya Domus said:
32 GB of HBM2 is nuts.
Not for miners ;)
Posted on Reply
#16
jabbadap
Vya Domus said:
32 GB of HBM2 is nuts.
Quadro GV100 and new Tesla V100s have that much. So it's bare minimum for Vega 20 too.
Posted on Reply
#17
Slizzo
lynx29 said:
I hope you are right my friend, I want nothing more than to support AMD with Ryzen 2 and Vega 2 in 2019. I am tired of toothpaste and gimmicks, and Nvidia's GPP bullying kind of disgusts me frankly. Nvidia has had 0 issues with branding of GPU's until now, they hold 80% of market share, why be so afraid. It's just so dumb, I hate bullying of any kind.
Why is NVIDIA afraid? Because they have nothing in the iGPU space, which is where the bulk of all GPU sales go. Intel partnered up with AMD to provide some "Vega/Polaris" GPUs for their Core i7 CPUs, and now AMD has that market as well (OEMs). NVIDIA is very threatened by that, thus the GPP program.
Posted on Reply
#18
HD64G
Maybe a twin vega 10 gpu on a single pcb? Vega10 x 2=Vega 20? Makes more sense imho to have 32GB HMB for 2 GPUs and the 7nm will help to have almost double the computing power with same consumption. Which is why they will bring that mainly for the production-computing-server market as for gaming CF doesn't work properly at all times. If they did that it makes much sense as a product. I cannot see the meaning of it just as a single core gpu with that added cost for the 32GB HBM.
Posted on Reply
#19
kruk
lynx29 said:
That is a shame my 1080 Ti still beats it in graphics score by 3000 points. I guess I will be waiting another year or two and hoping for the best, as usual.
This is a machine learning oriented GPU and it's not intended for gaming, it's also an early sample. Testing in 3D Mark could mean we will get a gaming oriented RX version before Navi drops, but definitely not with 32 GB of HBM2 ...
Posted on Reply
#20
T4C Fantasy
CPU & GPU DB Maintainer
@btarunr its 1280gb/s 4096bit its double because of the 32GB memory vs 16
Posted on Reply
#21
bogami
And what is the consumption? Frustrating results .
Posted on Reply
#22
FordGT90Concept
"I go fast!1!11!1!"
If this card has 4 HBM2 stacks, this is a pretty big deal. It implies they added more memory controllers and will have double the memory bandwidth of Vega 64. Vega 64 is memory bandwidth starved so Vega20 could see a serious boost in performance.
Posted on Reply
#23
jabbadap
FordGT90Concept said:
If this card has 4 HBM2 stacks, this is a pretty big deal. It implies they added more memory controllers and will have double the memory bandwidth of Vega 64. Vega 64 is memory bandwidth starved so Vega20 could see a serious boost in performance.
Specs says maximum density for HBM2 is 8GB so to get 32GB it must have four of them(There's no Clam-shell mode for hbm). Unless it uses some unreleased HBM3 with updated specs...

T4C Fantasy said:
@btarunr its 1280gb/s 4096bit its double because of the 32GB memory vs 16
In other way to say it, it has four 8-hi 8GB 1024bit 320GB/s HBM2 vram chips. But I'm a bit skeptical for that though, it would mean 2.5 Gb/s per pin while both amd and nvidia have had hard time to achieve that current available maximum speed of 2 Gb/s per pin.
Posted on Reply
#24
T4C Fantasy
CPU & GPU DB Maintainer
jabbadap said:
Specs says maximum density for HBM2 is 8GB so to get 32GB it must have four of them(There's no Clam-shell mode for hbm). Unless it uses some unreleased HBM3 with updated specs...



In other way to say it, it has four 8-hi 8GB 1024bit 320GB/s HBM2 vram chips. But I'm a bit skeptical for that though, it would mean 2.5 Gb/s per pin while both amd and nvidia have had hard time to achieve that current available maximum speed of 2 Gb/s per pin.
all that means is lower memory clock speed, its a fact its 4096bit though
Posted on Reply
#25
jabbadap
T4C Fantasy said:
all that means is lower memory clock speed, its a fact its 4096bit though
Yep, that is given. That 3dmark 11 score run uses SystemInfo 5.1, which is from June 2017. One interesting fact on SystemInfo it has some suspiciously named dll called gpu-z.dll. So @W1zzard does it use gpu-z sdk? Either way it's obvious that old SystemInfo can't recognize that graphics card.
Posted on Reply
Add your own comment