AMD Vega (FE and RX) Benchmarks [Updated Aug 10 - RX Vega 64 Unboxing]

[davide445]

I want to ask why this is to be considered a gaming card.
From the presentation, price, drivers I did consider it a card for professional content creators.
In this case his price/performance appear to me really competitive considering benchmark results on Professional applications.
I also suppose game content designers didn't test on the 0.5% user base with the fastest available cards, so also didn't understand the performance problem here.
I'm probably missing something so will be really pleased for some others opinion.

 

[Tup3x]

I want to ask why this is to be considered a gaming card.
From the presentation, price, drivers I did consider it a card for professional content creators.
In this case his price/performance appear to me really competitive considering benchmark results on Professional applications.
I also suppose game content designers didn't test on the 0.5% user base with the fastest available cards, so also didn't understand the performance problem here.
I'm probably missing something so will be really pleased for some others opinion.

Well, they are comparing this to Titan Xp instead of Quadro so...

 

[R0H1T]

Well, they are comparing this to Titan Xp instead of Quadro so...

There's no (Pascal) quadro they could pit it against, AFAIK at the Vega FE "reveal" event, so they went with the next best (or worst) thing to compare it to.

 

[wanderica]

I guess it is time to pull out popcorn, enjoy the doom and gloom.

Ryzen has not taught people to not judge new architecture after first glimpses of performance, I guess.

Im wondering also, when people will get that AMD did not send the GPUs to reviewers, because - the drivers are not ready, to show final product?

I get your point, but Ryzen is a bit different, I think. If you're referring to the low clock rumors, then the difference here is that Vega FE is a released product while those clock rumors were well before Ryzen launch. That's a pretty big distinction. If you're talking about the new platform growing pains that Ryzen is still experiencing to some degree, then I feel it necessary to point out that Ryzen was trading blows with Intel's best at less than half the cost. This Vega FE performance is embarrassing compared to Nvidia's products.

For me, at least, it's not the performance itself that has me rolling my eyes here. It's the way it was handled from the beginning. AMD took great efforts to compare Vega FE to Titan Xp, yet now that it's out, it is FAR below that in gaming performance to the point that the unbelievably overpriced Titan Xp might actually be the better value for gaming. The p4000 beats this thing in professional workloads at less money. I really hope this is the previously mentioned driver issue. Otherwise AMD might have really stepped in it with Vega.

 

[tential]

I guess it is time to pull out popcorn, enjoy the doom and gloom.

Ryzen has not taught people to not judge new architecture after first glimpses of performance, I guess.

Im wondering also, when people will get that AMD did not send the GPUs to reviewers, because - the drivers are not ready, to show final product?

If the drivers are not ready to show the final product after the product is released then how you can you say the final product is ready? You're suggesting amd released an unfinished product and the only reason to do so is to meet the 1H deadline they communicated to investors.

 

[Dribble]

There's no (Pascal) quadro they could pit it against, AFAIK at the Vega FE "reveal" event, so they went with the next best (or worst) thing to compare it to.

Of course they could have compared it to a quadro card but then it would have looked rubbish. They could have compared it gaming but then it would have looked rubbish. The only way they could make it look half decent was by taking the halo overpriced gaming card and then using it (with it's gaming drivers) in professional benchmarks, which is stupid as no one buys a titan to run cad that's what quadro is for.

 

[Glo.]

If the drivers are not ready to show the final product after the product is released then how you can you say the final product is ready? You're suggesting amd released an unfinished product and the only reason to do so is to meet the 1H deadline they communicated to investors.

Why do you expect game ready drivers from a product that is slated for professionals, in the first place?

 

[Phynaz]

What has been rumored at the time of Vega presentation(January, the driver 17.1.1 is also from January) that in that moment it was ONLY Fury X drivers, tuned only to work "just" with Vega architecture.

I find the claim that Vega was running Fury drivers difficult to believe. If Vega is truly a new architecture with a new instruction set as I believe AMD has claimed, then there's no way Fury drivers could function. All the hardware the driver interfaces would be different.

 

[Arachnotronic]

I find the claim that Vega was running Fury drivers difficult to believe. If Vega is truly a new architecture with a new instruction set as I believe AMD has claimed, then there's no way Fury drivers could function. All the hardware the driver interfaces would be different.

Agreed.

 

[Glo.]

I get your point, but Ryzen is a bit different, I think. If you're referring to the low clock rumors, then the difference here is that Vega FE is a released product while those clock rumors were well before Ryzen launch. That's a pretty big distinction. If you're talking about the new platform growing pains that Ryzen is still experiencing to some degree, then I feel it necessary to point out that Ryzen was trading blows with Intel's best at less than half the cost. This Vega FE performance is embarrassing compared to Nvidia's products.

For me, at least, it's not the performance itself that has me rolling my eyes here. It's the way it was handled from the beginning. AMD took great efforts to compare Vega FE to Titan Xp, yet now that it's out, it is FAR below that in gaming performance to the point that the unbelievably overpriced Titan Xp might actually be the better value for gaming. The p4000 beats this thing in professional workloads at less money. I really hope this is the previously mentioned driver issue. Otherwise AMD might have really stepped in it with Vega.

First performance benchmarks from Ashes of Singularity were slating the CPU performance at around 4 core Cire i5 with Haswell architecture.

Everybody jumped to conclusions about how rubbish it is. Reality turned out to be different. Why? Because gaming software was never optimized for Ryzen architecture.

And what we have seen at release?

Keep calm. And do not jump to conclusions about hardware that can use game drivers from January(!).

 

[tential]

I find the claim that Vega was running Fury drivers difficult to believe. If Vega is truly a new architecture with a new instruction set as I believe AMD has claimed, then there's no way Fury drivers could function. All the hardware the driver interfaces would be different.

I'm no expert on that but I think it's irrelevant what amds excuse is. They released a product with a massive price tag and marketed it as being able to game, showed it against a titan xp, and now we're hearing every excuse under the sun as to why the gaming experience is poor.

You can't catch the moving goal post for how long you're supposed to wait for amds product to be ready. You're always supposed to wait for xyz.

 

[tamz_msc]

I find the claim that Vega was running Fury drivers difficult to believe. If Vega is truly a new architecture with a new instruction set as I believe AMD has claimed, then there's no way Fury drivers could function. All the hardware the driver interfaces would be different.

I don't know if it is a "Fury driver", but it is an older one:

Maybe @rys could shine a light on the matter, just one question: is the Vega FE gaming driver so far gimped compared to the Vega RX driver coming a month later?

It's not gimped (that would be completely ridiculous), but it is older.

https://forum.beyond3d.com/posts/1989522/

 

[Glo.]

I find the claim that Vega was running Fury drivers difficult to believe. If Vega is truly a new architecture with a new instruction set as I believe AMD has claimed, then there's no way Fury drivers could function. All the hardware the driver interfaces would be different.

You can put the Graphics IPv9 into compatibility mode, but your software will lack all of new features: FP16, Memory paging system, redesigned memory architecture(Pixel Engine client of L2 cache). The only thing that would work, just, is the load balancing, because it is the core of the architecture(Hardware Schedulers).

In GPv9 we have the same structure of CU's. Throughput of the cores should be the same, if AMD has not changed in any way the Registry File Sizes available to those cores. What is different is what is around the architecture, and what is responsible for graphics(geometry performance).

I can compare Vega architecture from High level layout to GP100/GV100 and Consumer Pascal GPUs, but this is not the thread for this.

What matters is this: somehow in current state, Frontier edition Radeon Vega GPU is not performing as it should based on paper specs, and architecture layout.

 

[Kenmitch]

I'm no expert on that but I think it's irrelevant what amds excuse is. They released a product with a massive price tag and marketed it as being able to game, showed it against a titan xp, and now we're hearing every excuse under the sun as to why the gaming experience is poor.

You can't catch the moving goal post for how long you're supposed to wait for amds product to be ready. You're always supposed to wait for xyz.

....Or gaming performance will be exposed when RX Vega is launched. They did state other features of Vega FE edition will be revealed when RX launches. What that means only time will tell.

 

[Armsdealer]

... The good news is amd will have to price it cheap given they have a tiny time window in which to sell before it becomes totally obsolete by volta. The better news is the cut die vega 56 has 87.5% of the compute of the full die. Better than 1070 performance at 300 USD or less is my guess.

 

[davide445]

There's no (Pascal) quadro they could pit it against, AFAIK at the Vega FE "reveal" event, so they went with the next best (or worst) thing to compare it to.

I didn't understand also this. Quadro P5000 and P6000 are the obvious benchmarks to be compared with for a card such as Vega FE considering price and target users.
From the SPEC benchmark I read appear to me it's at about 90% of P5000 performance at 1/2 of the price. Or again I'm missing something.

 

[Cloudfire777]

Maybe @rys could shine a light on the matter, just one question: is the Vega FE gaming driver so far gimped compared to the Vega RX driver coming a month later?

https://forum.beyond3d.com/posts/1989520/

It's not gimped (that would be completely ridiculous), but it is older.

https://forum.beyond3d.com/posts/1989522/

 

[Malogeek]

I want to ask why this is to be considered a gaming card.

The only thing that really differentiates it from the supposed RX Vega is 16Gb vs 8Gb memory capacity. It's the same chip otherwise.

 

[Bacon1]

https://forum.beyond3d.com/posts/1989520/


https://forum.beyond3d.com/posts/1989522/

Right, its not "gimped" as in they aren't intentionally sabotaging it. But it is "gimped" in that its much older and missing out on a lot of performance improvements.

 

[Glo.]

The only thing that really differentiates it from the supposed RX Vega is 16Gb vs 8Gb memory capacity. It's the same chip otherwise.

8 GB GPUs will have 2 Gbps HBM2 chips, with 512 GB/s bandwidth.

 

[IllogicalGlory]

Right, its not "gimped" as in they aren't intentionally sabotaging it. But it is "gimped" in that its much older and missing out on a lot of performance improvements.

Sounds like wishful thinking to me. What performance improvements is it missing, and why do you believe that?

 

[Glo.]

Sounds like wishful thinking to me. What performance improvements is it missing, and why do you believe that?

I guess, we have to wait and see.

 

[Despoiler]

I find the claim that Vega was running Fury drivers difficult to believe. If Vega is truly a new architecture with a new instruction set as I believe AMD has claimed, then there's no way Fury drivers could function. All the hardware the driver interfaces would be different.

The demo of Vega on Star Wars Battlefront that was done 6 or so months ago was done using Fury drivers put into a debug mode. That was straight from AMD themselves.

 

[swilli89]

The demo of Vega on Star Wars Battlefront that was done 6 or so months ago was done using Fury drivers put into a debug mode. That was straight from AMD themselves.

I wouldn't waste your breath

 

[Kenmitch]

Sounds like wishful thinking to me. What performance improvements is it missing, and why do you believe that?

Sounds more logical than wishful too me at least. The evidence or lack of points more towards unoptimized drivers than a hardware issue.