[Updated with final Review] Vega FE Benchmarking by PCPER

[Ekin]

The final pcper review is here: 

https://www.pcper.com/reviews/Graphics-Cards/Radeon-Vega-Frontier-Edition-16GB-Air-Cooled-Review

Very Interesting and worth a look!

Quote

This isn’t a gaming card.

[...]

Calling this “not a gaming card” is a fair statement, as long as you also agree than the GTX Titan, Titan X, Titan Xp are also “not gaming cards.” But they are, despite NVIDIA segmenting it off as well. Plenty of professionals will buy this hardware, but discerning gamers that want the best of the best will also be purchasing Titans and FEs well into the future. “Professional graphics cards” have certified drivers and specific code paths in place for applications like 3ds max, Maya, etc. Neither Titan nor Vega FE have that and instead will depend on the driver stacks we are used to seeing in GeForce and Radeon systems.

 

The drivers are old.

[...] the driver we used in our testing, and the ONLY driver that supports Vega FE today, is not old. To be clear though, the driver IS from a different branch than the currently available Radeon RX 500-series driver, but the exact time of that branch, and how it affects performance on games or other workloads, isn’t information AMD is interested in sharing at this time.

 

The driver isn’t optimized for gaming.

The games we are using for this review were not released in the last 30 days or anything like that. [...]

The one caveat to this is that the Vega architecture itself is still unoptimized in the driver. But this would affect gaming and non-gaming workloads most of the time. So if the driver isn’t ready for ANYTHING, then that’s a valid concern, but it applies to ALL workloads and not just games.

 

It didn’t hit the 1600 MHz it should have.

The rated clock speed of the Radeon Vega Frontier Edition is 1600 MHz, but in our testing the GPU never really got to that, settling in the 1440 MHz range the majority of the time. 

 

Quote

From a gaming standpoint, the performance of the Radeon Vega Frontier Edition did not light the stage on fire. In general, our game benchmarks put the Vega FE somewhere between the GeForce GTX 1070 and the GTX 1080.

 

The Fallout 4 Benchmark was corrected everything else seems to be the same.

 

And finaly from the professional standpoint it looks a bit better.

Quote

From a professional workload angle, the Radeon Vega Frontier Edition fares much better, splitting the line between the features, performance and support that NVIDIA offers between its GeForce and Quadro products. By enabling those features, and in doing so giving the Vega FE the ability to battle a much higher performance GPU (otherwise) in this space, AMD is hoping to convince those creators to take notice and invest. To be clear, in most cases, the P5000 Quadro card will offer performance that exceeds Vega FE, but at double the price. 

 

Professional Benchmarks: 

Spoiler

 

 

 

Bonus: Nude Vega
https://www.pcper.com/news/Graphics-Cards/Radeon-Vega-Frontier-Edition-GPU-and-PCB-Exposed

Spoiler

There is a LOT of empty PCB space on the Vega FE card. This is likely indicative of added area needed for a large heatsink and fan to cool 300-375 watt TDP without throttling.

 

-----------------------------------------------------

Old Post

-----------------------------------------------------

 

Today pcper gave us a behind the scenes Livestream of their Vega Frontier Edition benchmarking and at the end we got a sneak peek of their gaming results. 
Please visit pcper.com tomorrow to get the full Story. I will link it here when it goes up!

 

Here are some early gaming results. The benchmarks are not 100% done and more cards & productivity benchmarks will be added.

 

Gaming: 2560x1440:

Spoiler

Fallout 4 1440p will be retested before the article goes up on pcper.

 

Gaming 3840x2160:

Spoiler

 

Quote

 

Power draw was at ~280 Watt in gaming, with only ~25 Watt from the PCIe slot. 

Mining performance was verry low.

 

 

Source Video with some timestamps:

• 1:19h ethminer
• 1:36h excavator mining
• 1:42h gaming benchmark results
• 1:44:45h Ryan Shrout's closing thoughts

The first half of the stream is missing.

 

In my opinion the gaming results are not great. It will be hard to price Vega FX competitivly if it uses HBM and only performces a bit better than a GTX 1070.

[Max_Settings]

So maybe 3-5% better performance than a GTX 1070, for double the TDP? No thank you.

[LienusLateTips]

14 minutes ago, Max_Settings said:

So maybe 3-5% better performance than a GTX 1070, for double the TDP? No thank you.

But this is Frontier, not targeted at gamer.

[ARikozuM]

Just now, JDE said:

But this is Frontier, not targeted at gamer.

So, it'll use more POWAH?!

[leadeater]

17 minutes ago, ARikozuM said:

So, it'll use more POWAH?!

[You_are_a_cunt]

this is exactly what happened with the 1800x. These cards are not designed or targeted towaeds being gaming cards. They are workstation cards, similar to quadro or firepro cards. Vega gaming gpus will launch at a later date.

the gaming benchmarks are interesting, but ultimately useless

[Taf the Ghost]

7 minutes ago, revsilverspine said:

this is exactly what happened with the 1800x. These cards are not designed or targeted towaeds being gaming cards. They are workstation cards, similar to quadro or firepro cards. Vega gaming gpus will launch at a later date.

the gaming benchmarks are interesting, but ultimately useless

Given the general ability to remember things by most people, everyone will forget about this by the time the RX Vega launches.

[ARikozuM]

1 minute ago, Taf the Ghost said:

Given the general ability to remember things by most people, everyone will forget about this by the time the RX Vega launches.

Does anyone here remember how bad the Titan Z was?

[You_are_a_cunt]

Just now, Taf the Ghost said:

Given the general ability to remember things by most people, everyone will forget about this by the time the RX Vega launches.

true, but it still kind of annoys me that people nitpick and rage about things like workstation ccomponents not performing well in gaming.

i've yet to see someone complain about an 1800x not performing up to par in cpu intensive tasks

[Taf the Ghost]

1 minute ago, ARikozuM said:

Does anyone here remember how bad the Titan Z was?

Most even techies forgot the Titan X exists until the FE launched.  Non-mainstream products get put in the Memory Hole really fast.

[Hunter259]

43 minutes ago, JDE said:

But this is Frontier, not targeted at gamer.

It's a prosumer card with a Gaming mode. It's a titan for AMD. The have actual pro cards on the way. Quadro's aren't "meant" for gaming but run just as well as their GTX counterparts. This is such a weak argument for this card. You might get 10-15% on drivers but this still is such a disappointing card.

[Phate.exe]

So it's barely outperforming a lightly overclocked Fury X, with similar power draw, and WAY higher clockspeeds?

 

Yeah, something's not right.  Wouldn't have taken them nearly this long if all they were doing was making a 14nm Fiji card.

 

There's "this card is underperforming because the drivers are not optimized well" (see:  Polaris) and then there's "this card is barely functioning".  I'm thinking we're a lot closer to the second option.

[Taf the Ghost]

8 minutes ago, Hunter259 said:

It's a prosumer card with a Gaming mode. It's a titan for AMD. The have actual pro cards on the way. Quadro's aren't "meant" for gaming but run just as well as their GTX counterparts. This is such a weak argument for this card. You might get 10-15% on drivers but this still is such a disappointing card.

 

So, is the FE like the 1070 or the 1080 Ti?  If it gets another 15% from the drivers, is that going to make it 10% faster than the 1080 Ti?

 

Which is it? 

 

Plus, you're assuming AMD is using the exact same approach as Nvidia where all Vega cards are just cut-down versions of the top-tier professional one. It's got new tech with a new driver architecture. Until we have the RX version, we don't know.  That's an actual argument: we don't know yet.

[Hunter259]

Just now, valdyrgramr said:

I mean the card was aimed at pros, can game, but Raja said the card is slower than the RX ones.  But, I'm not sure how well this card will perform in a pro sense either.  So, who was this aimed at?

It's not going to be much slower unless they are holding back some god tier drivers for the RX version but I can't thing of anytime a card has gained over 40% from a driver. This seems like they knew it was a failure and waited as long as they could but that just seems so stupid. Based on this, VEGA is worse than Polarius Perf/Watt

[Taf the Ghost]

3 minutes ago, valdyrgramr said:

I mean the card was aimed at pros, can game, but Raja said the card is slower than the RX ones.  But, I'm not sure how well this card will perform in a pro sense either.  So, who was this aimed at?

People that want $5000 USD Nvidia cards, minus validated drivers, for design projects.  Think small business CAD stuff. 

[leadeater]

3 minutes ago, Phate.exe said:

So it's barely outperforming a lightly overclocked Fury X, with similar power draw, and WAY higher clockspeeds?

 

Yeah, something's not right.  Wouldn't have taken them nearly this long if all they were doing was making a 14nm Fiji card.

 

There's "this card is underperforming because the drivers are not optimized well" (see:  Polaris) and then there's "this card is barely functioning".  I'm thinking we're a lot closer to the second option.

Plot twist, AMD is kneecapping Vega FE on purpose until Vega RX comes out.

 

For the love of god don't take this seriously.

[Hunter259]

1 minute ago, Taf the Ghost said:

 

So, is the FE like the 1070 or the 1080 Ti?  If it gets another 15% from the drivers, is that going to make it 10% faster than the 1080 Ti?

 

Which is it? 

 

Plus, you're assuming AMD is using the exact same approach as Nvidia where all Vega cards are just cut-down versions of the top-tier professional one. It's got new tech with a new driver architecture. Until we have the RX version, we don't know.  That's an actual argument: we don't know yet.

It got close in that one game. That's an outlier. Currently it's closer to a 1070/1080 so it will be somewhere near those two even with 15% from drivers. 

 

Lol. Dude. Everyone does this. You make 1 GPU and cut it from there. That is top end VEGA with some of the pro software stuff enabled. It's still a full fat VEGA chip we have been waiting on.

 

You can run Kepler on drivers never made for it and it only gained maybe 10% in most games when drivers came out for it. This isn't like something the industry has never seen before.

[Cartoons Plural]

25 minutes ago, ARikozuM said:

Does anyone here remember how bad the Titan Z was?

I remember it was two titans for the price of three.

[ARikozuM]

2 minutes ago, leadeater said:

For the love of god don't take this seriously.

[Taf the Ghost]

7 minutes ago, valdyrgramr said:

I mean the card was aimed at pros, can game, but Raja said the card is slower than the RX ones.  But, I'm not sure how well this card will perform in a pro sense either.  So, who was this aimed at?

See the benchmarks there.  It mostly seems Vega FE toasts AMD's own previous offerings in actual Production-type tasks.  Which seems the actual point of the card.  (Which also brings up a question of how many monitors can it run.)

 

[Ekin]

4 minutes ago, Taf the Ghost said:

 

So, is the FE like the 1070 or the 1080 Ti?  If it gets another 15% from the drivers, is that going to make it 10% faster than the 1080 Ti?

 

Which is it? 

 

 

Ryan believes there was an error in testing Fallout, thats why I wrote he will retest Fallout again.

[Taf the Ghost]

9 minutes ago, Hunter259 said:

It got close in that one game. That's an outlier. Currently it's closer to a 1070/1080 so it will be somewhere near those two even with 15% from drivers. 

 

Lol. Dude. Everyone does this. You make 1 GPU and cut it from there. That is top end VEGA with some of the pro software stuff enabled. It's still a full fat VEGA chip we have been waiting on.

 

You can run Kepler on drivers never made for it and it only gained maybe 10% in most games when drivers came out for it. This isn't like something the industry has never seen before.

Again, you're pointing to Nvidia can proclaiming it is identical. AMD's design problems until Raja got there is that they designed really good iGPUs and then scaled them up. (Part of the comparative power usage issues is that the Maxwell architecture scales down really well.)  I fully expect that the Vega die is going to be the same, but GPUs are far more about getting the raw calls to the processing units.  We don't know how far the new tech in the cache can be pushed, or if the FE edition is running some compatibility approach at the base-level which allows it to switch modes without too much issue. (Or do both at the same time.) 

 

There's a ton we don't know and everyone posting about it doesn't know either.  I still expect the RX Vega to slot in roughly between the 1080 and 1080 Ti in most situations for the top SKU while being cheaper than the 1080 Ti. Except in DX12 on Ryzen, where it might toast the 1080 Ti by 30% because there's something wrong with Nvidia's DX12 driver there.

[Hunter259]

Just now, Taf the Ghost said:

Again, you're pointing to Nvidia can proclaiming it is identical. AMD's design problems until Raja got there is that they designed really good iGPUs and then scaled them up. (Part of the comparative power usage issues is that the Maxwell architecture scales down really well.)  I fully expect that the Vega die is going to be the same, but GPUs are far more about getting the raw calls to the processing units.  We don't know how far the new tech in the cache can be pushed, or if the FE edition is running some compatibility approach at the base-level which allows it to switch modes without too much issue. (Or do both at the same time.) 

 

There's a ton we don't know and everyone posting about it doesn't know either.  I still expect the RX Vega to slot in roughly between the 1080 and 1080 Ti in most situations for the top SKU while being cheaper than the 1080 Ti. Except in DX12 on Ryzen, where it might toast the 1080 Ti by 30% because there's something wrong with Nvidia's DX12 driver there.

AMD would have to be purposely gimping the card for what you are saying to be true. They said they were working hard on the drivers and the drivers that were used were VEGA drivers. 

 

And now I know you are just being stupid.

[Taf the Ghost]

1 minute ago, Hunter259 said:

AMD would have to be purposely gimping the card for what you are saying to be true. They said they were working hard on the drivers and the drivers that were used were VEGA drivers. 

 

And now I know you are just being stupid.

I'm so taken aback. I see the light now! 

 

There's a serious issue with Ryzen + DX12 + 1080/TI that kept coming up in gaming benchmarks. We were discussing it a lot after the Ryzen 5 dropped and we started seeing game optimizations.  It's a Nvidia problem in DX12. (I'd need to dig back to grab what seems to be the full issue, as it's been a couple of months.)

 

As to the Vega FE, it works at its actual role (as a Tweener CPU for professional work), but the gaming performance leaves a lot to be desired. However, we don't know why and we also don't know if there's some specific reason it might be limited.  (Baseline compatibility is a "thing" and important in a lot of settings.)

[goodtofufriday]

I just want to see what it does computationally. I dont see it as an impossibility that this card is design in such a way that, while it can game, it just wont be anything terribly amazing. But then in comp tasks itll kill everything