I loaded sixteen gigs of HBM2 ...

Subject: Graphics Cards | February 7, 2019 - 03:30 PM |
Tagged: VRAM, video card, Vega 20, Vega, radeon vii, radeon, pcie, opencl, HBM2, graphics card, gaming, compute, amd, 7nm, 16GB

While enjoying the pictures and tests Sebastian ran on the new AMD Radeon VII, was there a game that we missed that is near and dear to your heart?  Then perhaps one of these reviews below will solve that, the list even includes Linux performance for those on that side of the silicon.  For instance, over at The Tech Report you can check out Monster Hunter: World, Forza Horizon 4 and the impressive results that the new 7nm card offers in Battlefield V. 

Check those results here.

View Full Size

"AMD's Radeon VII is the first gaming graphics card powered by a 7 nm GPU: Vega 20. This hopped-up Vega chip comes linked up with 16 GB of HBM2 RAM good for 1 TB/s of memory bandwidth. We put this potent combination to the test to see if it can beat out Nvidia's GeForce RTX 2080."

Here are some more Graphics Card articles from around the web:

Graphics Cards

 

Video News


February 7, 2019 | 05:02 PM - Posted by MoreDieAndWaferActionAt7nmForRed (not verified)

That's a few more DIEs/Wafer at 7nm with the DIE shrink so the usable die count will go up. Let's hope there can be some use for even the Vega 20 DIEs that do not have enough working shader cores to even bin out the Radeon VII.

So maybe a lower cost Vega 20 variant that's closer to the Vega 56 in Shaders:TMUs:ROPs that's still ok for gaming and uses less power. Maybe even a Dual Die on a single PCIe card variant like the Radeon Pro V340 that's based on 2 Vega 10 DIEs that are each similar to Vega 56. I'm sure that the current Vega 10 DIEs based V340 will be updated to with some Vega 20 bins and maybe even a Dual Vega 20 die based GPU SKU for the consumer gaming market is also possible.

That's what AMD did that for the Radeon Pro Duo(Fiji) and Radeon Pro Duo(Polaris) based SKUs. So why not a Radeon Pro Duo/Vega 20 based SKU once the binns begin to fill up with more defective DIEs. And maybe AMD could interface the 2 Vega 20 DIEs across the PCIe card's PCB with some xGMI links instead of PCIe/Protocol links.

And looking at the 60 nCU based Radeon VII's performance relative to the 64 nCU based Vega 64 the Radeon VII's gaming performance was not hurt by having less shader cores. So Maybe 2, 50 nCU enabled Vega 20 die bins that still have 64 working ROPs per Die and a better ROP to shader ratio.

I'll bet that would put AMD on TOP what with the xGMI Links being basd on the Infinity Fabric Protocol. 32GB of VRAM and tons of room for more gamng assets that do not have to be swapped out as often, with 2TB/s of total VRAM Bandwidth across 2 GPU DIEs to top it all off.

February 7, 2019 | 05:05 PM - Posted by meettooo (not verified)

You loaded sixteen gigs, and what didya get? Not even a 1080ti an' deeper in debt.

February 8, 2019 | 01:15 PM - Posted by Jeremy Hellstrom

Not bad at all ...

February 8, 2019 | 11:30 AM - Posted by chipman (not verified)

"Taking Vega 20 out of the data center means AMD is first to market with a gaming GPU fabricated on TSMC's cutting-edge 7 nm FinFET process."

It is simply G A R B A G E S also called Radeon Instinct MI50 rebranded as Radeon VII!

"There's a shocker. Despite being rated for a similar board power to that of the Radeon RX Vega 64, the Radeon VII actually lets our test system consume quite a bit less power while delivering much higher performance. We wouldn't be surprised if the actual improvement in performance per watt for the Radeon VII greatly exceeds that 25% figure.
"

There's no surprise since the Doom test doesn't give the performance but only the system power consumption.

Thus, it is NOT enough to determine the power efficiency.

https://techreport.com/review/34453/amd-radeon-vii-graphics-card-reviewed

It seems AMD paid reviewers to promote their G A R B A G E S even with dumb measures!

February 8, 2019 | 03:11 PM - Posted by ReallyGamersDoTheyEvenMatterInTheRealWorld (not verified)

Really not and some undervolting was done on some review samples with 70+ Watts more power savings.
But Radeon VII is an even lower binned part than that MI50 even if the Shaders:TUMs:ROP figures are the same because the MI series parts have better thermal metrics because GPU makers bin on power usage metrics also. The best electrically and thermally performant Vega 20 DIEs with proper numbers working units are never getting in the hands of gamers, Never Ever!

Radeon VII's DP FP Metrics have been clipped back also to 1/4 the rate of single percision FP rates instead of the MI50's/MI60's 1/2 DP FP to 1 SP FP rates. Radeon VII also does not support xGMI because there are no xGMI connections provided on the consumer Vega 20 die based variants.

But if you are a prosumer Radeon VII does still offer higher DP FP(3,360 GFLOPS) rates than the Vega 10 based GPUs and Nvidia's consumer variants of TU102. The TU102 based RTX 2080TI only offers: 420.2 GFLOPS of DP FP performance at 1:32 its SP FP Rates.

So we can see where Nvidia is getting some of its power savings right there with less DP FP enabled. The Radeon VII still has the prosumer aspect and AMD will galdly sell Radeon VIIs to coin miners that want the 16GB of VRAM for the more memory intensive coin hashing algorithms, prosumers also for other Non Gaming Graphics uses in addition to compute.

Just you look at the larger number of numbers of Quadro variants that Nvidia gets from its TU102 Base Die tapeout before It bottom bins the poorest performing TU102 DIEs for is RTX 2080Ti production.

You can definitely Game on Radeon VII and its competative with the RTX 2080 and Radeon VII will still need more driver tweaks on some gaming titles this early in its release cycle. But AMD will measure its GPU sucess buy the numbers if units sold for whatever usage the customer intends.

And AMD does not have to be all GPU only focused as Nvidia what with AMD's Zen Server Products taking ever increasing server market share numbers with each business quarter that goes by and taking those Radeon Instinct MI50/MI60 unit sales along for the Epyc ride as GPU Compute/AI accelerators for any clients that need the extra FP/AI power.

Really What Can Flagship Gamers expect but the non perfortmant Die binns of some Vega 20 or TU102 Die Tapeouts! That's because gamers can not afford to pay their own way in the GPU market by paying the proper markups like the Pro Market pays to cover those billion dollar+ R&D and other development costs. All that GPU IP R&D/Other cost is paid for via the Pro GPU revenues/markups with gamers leaching on the Pro Customers dime. That's alright you self entitled gamers because the pro customers get to deduct the cost of the Pro GPUs from their Taxes as a business expense and now gamers are leaching off of the Taxes Paied by businesses.

Maybe U-SAM should add an extra consumer GPU Tax on there to create more Government Research Grants for exascale systems, and that research benifits anyone that uses CPUs, GPUs, or any other types of processors.

chipman's single enfeebled Cell of Gray is twitching again in its Triggered state and heating up that sea of lard ocean that dwarfs that tiny Cell that's barely fuctional! Maybe chipman is undervolting that cell a bit too much or it could be poor genetic variation that's the problem.

February 10, 2019 | 01:25 PM - Posted by Anonymousnameisalreadyused (not verified)

^ does every bot think we all suffer from eds

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.