Custom-design Radeon RX Vega Cards by Mid-October

[lewis007]

That "everybody" of yours include the reviewer of GPUs in TPU named @W1zzard also? If not, take a look at his findings when testing on the launch-day driver of Vega GPUs. And come again in 2-3 months to discuss again if Vega is by then so big of a failure you and others talk about so much. I am talking ONLY about performance at present. Power consuption can become MUCH better with 2 clicks in the driver btw.

 

[lewis007]

You have to admit it is power hungry and expensive.

 

[W1zzard]

Thanks for clarification. I wonder what makes your results so different. Guru3D shows 9% at 4K, the site I linked shows 10%, this video shows anywhere from no difference to 15%, which would maybe amount to 10% overall and fall in line with the other two. HardCOP /1440p only available/ - 10% as well. Your new test shows 22%, more than twice as big as the ones I quoted, and in line with the initial Pascal DOOM Vulkan performance on old drivers, where 1070 gets beat even by Fury X.

Probably different test scene. That video shows the game start which isn't that good a benchmark, not sure what the others use.

 

[cucker tarlson]

What scene are you using ?

 

[Casecutter]

Does anyone know how many Vega chips AMD is contracted to supply Apple for their Mac stuff?

That's what thought... nobody wants recognize there's an elephant in the room!

Just like why the Tonga XT never showed, and 380X was not brought in until late. AMD is working to supply the corporate custom clients first, it provides for a known ROI.

 

[cucker tarlson]

We don't recycle results, all results have the same age. Last VGA rebench was around July time.

I'm gonna put a question mark under that statement.

then why is 1070Ti reference faster than 1080 reference ? And 1070Ti 30% faster than 1070 can't be right. Both 1070 and 1080 results are old.

You DO recycle results, I knew your results were off cause the moajorty of other reviews showed half the gap between 1080 an V64 then you reported.

 

[W1zzard]

As listed in the review set up page which you apparently didn't bother to check:

GTX 1070 Ti: 388.13 WHQL
GTX 1070, 1080, 1080 Ti: 388.00 WHQL
All RX Vega: 17.10.1 Beta
NVIDIA other cards: 384.76 WHQL
AMD other cards: Catalyst 17.6.2 Beta

I retested 1070, 1080, 1080 Ti and both Vegas before the review

 

[cucker tarlson]

1070 and 1080 got a driver boost in Vulkan far earlier than 388.00 came out. Guru 3D has 1070 beating Fury X in their review from August. 388.13 is a Wolfenstein II game ready driver.

 

[nemesis.ie]

I suppose if you want to be picky/get a retest we also need to use 17.11.1 for Vega as it ups the SoC clock to 1200MHz too.

I can understand W1zzard having to draw a line/pick a version at some point given the complexity of testing and the frequency of driver updates at the moment.

I suppose another option would be to have a "live benchmark" tab on the main page that "someone" continuously updates as many cards as possible with the newest drivers as time goes by?

That might actually be a cool feature ... "what cards are winning in your favourite game this week?" could be a thing.

 

[W1zzard]

1070 and 1080 got a driver boost in Vulkan far earlier than 388.00 came out. Guru 3D has 1070 beating Fury X in their review from August. 388.13 is a Wolfenstein II game ready driver.

Then all is good and I'm using the right drivers. Guru3D might be using a different test scene, test duration, maybe not warming up the card (thermal throttling)

 

[cucker tarlson]

Still, 30% between 1070 and 1070Ti is a weird result.

 

[eidairaman1]

So much for this only xfx has one in the works...

 

[EarthDog]

??????

https://www.tweaktown.com/reviews/8383/asus-rog-strix-vega-64-review-refined-kinda/index.html

 

[nemesis.ie]

A lot of very odd stuff/statements in that "review" ...

 

[eidairaman1]

A lot of very odd stuff/statements in that "review" ...

yeah and they don't even give you a board shot

 

[EarthDog]

Point is.. there is more than one.

 

[eidairaman1]

Point is.. there is more than one.

Only a board shot compared to a reference design would determine it being custom.

 

[EarthDog]

Only a board shot compared to a reference design would determine it being custom.

Oooook....
http://www.guru3d.com/articles-pages/asus-radeon-rog-rx-vega-64-strix-8gb-review,4.html

If unable to spot the differences now, you can go look up the reference design for comparison.

 

[eidairaman1]

Oooook....
http://www.guru3d.com/articles-pages/asus-radeon-rog-rx-vega-64-strix-8gb-review,4.html

If unable to spot the differences now, you can go look up the reference design for comparison.

yeah I thought I'll go ahead and compare thanks for the link too appreciate it