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

[Glo.]

Going through a benchmark suite of 21 games, 18/21 games has the 1080 ahead by 25-50%. I see 3 games where Fury X comes within 15% of a 1080, and fire strike isn't one of those 3 games. You're in denial.

What games? What site?

I use GameGPU and computerbase.de as a reference.

 

[T1beriu]

AMD is providing game drivers, linked from their Radeon Pro page for the FE release. And it's version 17.6

Apparently his using the latest drivers.

Driver from January would not even recognize Frontier. He used the latest driver from AMD website. I checked.

Source.

 

[Magic Hate Ball]

Apparently his using the latest drivers.

Source.

Wish I had my personal laptop at work so I could load up the chat session they're using.

I wonder if he did a proper DDU before swapping cards, and what he has access to for power options like max power etc.

Undervolting would be interesting too.

 

[Malogeek]

Apparently his using the latest drivers.

Yep, just saw that. Thanks.

 

[Dribble]

It can't complete in gaming (1080Ti is faster, cheaper, cooler, better o/c, ....).

It can't compete in pro benchmarks because even overpriced proper pro cards are competitively priced against it (e.g. P4000). This is aside from the fact no one is going to spend $1000 on a card that doesn't have proper certified workstation drivers.

It's too expensive for mining (at best it'll be twice as fast as a 580 or equivalent but it costs a lot more then 2 of them).

It's not going into super computers or blades or anything like that - this is clearly a consumer card.

It's doesn't have class leading performance like the titan XP (i.e. fastest in gaming) so you can't sell it to the class of people that will spend silly money just to have the best.

Really the frontier card has no market that I can see?

 

[T1beriu]

Really the frontier card has no market that I can see?

AMD Enthusiasts.

Or how I said it before, rich gamers and poor professionals, classic TITAN user.

 

[Despoiler]

Driver's INF file
;-----------------------------------------------
;----------1706261331-17.20.1035-170624a-315622C-CrimsonReLive

; AMD display information file
;
; Installation INF for the AMD display driver.
; Copyright(C) AMD 2007-2017
;-----------------------------------------------

[Version]
Signature="$Windows NT$"
Provider=%ATI%
ClassGUID={4D36E968-E325-11CE-BFC1-08002BE10318}
Class=Display
DriverVer=06/24/2017, 22.19.384.2
;;LayoutFile=layout.inf
CatalogFile=C0315622.CAT

 

[Malogeek]

It can't complete in gaming (1080Ti is faster, cheaper, cooler, better o/c, ....).

We don't know how the consumer version is going to be priced so that's not valid. Based on its performance characteristics when RX Vega is released, it will (I assume?) be priced competitively to the 1080/1080ti wherever it ends up sitting so for those who want an AMD architecture for their gaming GPU and also want to target Freesync/2 monitors, it will be a viable option.

Looks like it's going to be a power monster though, so that's going to be a consideration for many as well.

 

[Magic Hate Ball]

We don't know how the consumer version is going to be priced so that's not valid. Based on its performance characteristics when RX Vega is released, it will (I assume?) be priced competitively to the 1080/1080ti wherever it ends up sitting so for those who want an AMD architecture for their gaming GPU and also want to target Freesync/2 monitors, it will be a viable option.

Looks like it's going to be a power monster though, so that's going to be a consideration for many as well.

I assume this is going to be classic AMD, where 75%+ of the cards will be able to undervolt significantly at stock clocks.

 

[Sweepr]

I really hope these results are not representative of Radeon Vega's gaming performance, otherwise this is a disaster. Launch 1 year and 2 months after GTX 1080 to deliver similar results at 300W+ - using a much larger die (~500 mm² vs 314 mm²) + more expensive memory setup (HBM2 vs GDDR5X). Their minimum target should be matching GTX 1080 Ti. Not to mention GP102 is old news by now, pretty sure the competitor is working hard to bring a ~400-425 mm² Volta based GPU that should be faster than Titan Xp and equipped with 16GB GDDR6.

 

[Roger Wilco]

In the end, it's all going to come down to pricing. If this thing is in the ballpark of a 1080, AMD can adjust prices accordingly

We don't know how the consumer version is going to be priced so that's not valid. Based on its performance characteristics when RX Vega is released, it will (I assume?) be priced competitively to the 1080/1080ti wherever it ends up sitting so for those who want an AMD architecture for their gaming GPU and also want to target Freesync/2 monitors, it will be a viable option.

Looks like it's going to be a power monster though, so that's going to be a consideration for many as well.

Yes; in the end, it's all going to come down to pricing. If this thing is in the ballpark of a 1080, AMD can adjust prices accordingly and promote this as a direct competitor to the 1080/1070. Additionally, there are some pretty massive price gaps in GPUs beyond the 1060/470/480/570/580 range. I think AMD could do well bridging those gaps with aggressive pricing and strategic placement of the various Vega models. Once Volta comes around, however, AMD is going to need to adjust pricing and retreat back to the mid-range market.

 

[EXCellR8]

+1 for build quality, -1 for color scheme

http://wccftech.com/amd-radeon-rx-vega-excellent-price-performance/

info is from bits and chips

 

[Mopetar]

I really hope these results are not representative of Radeon Vega's gaming performance, otherwise this is a disaster.

Hopefully that should suggest they aren't. Otherwise it would be better for AMD to not release RX Vega at all, as it wouldn't sell at good margins and it would just be a waste of Vega dies that could have been put towards professional or HPC products that have much higher margins. With Ryzen being a compelling product, it isn't as though AMD doesn't have better use for spare wafers.

The best case scenario for RX Vega is that gaming drivers will be good enough in a month to justify its existence. I question if AMD can hit that goal simply because they should be much further along at this point. I'd be more confident in that assessment if AMD had done a better job of managing expectations, but they haven't been doing a good job at all. They knew in advance where the performance would land, and because it's terrible they needed to do a much better job of making it clear that the gaming drivers are very rough and not finalized.

Presumably Vega should be rather good for mining, so if AMD drops prices to make Vega competitive in terms of gaming, miners will scoop up all of the cards. At that point it makes far more sense for AMD to just sell them as mining cards and price them according to their value there which is going to be far more favorable for them from a revenue perspective, especially with the shortage in Polaris cards right now and their elevated prices.

 

[jpiniero]

What games? What site?

I use GameGPU and computerbase.de as a reference.

Tom's had a review of a 1080 Ti and the 1080 FE was something like 25-45% in games. This includes like Doom Vulkan and Hitman DX12.

What concerns me is that base clock. 1382 is not going to get it done.

 

[Cloudfire777]

What a disaster this architecture is if RX Vega ends up like 10% ish faster than GTX 1080.

No wonder why AMD didnt want any reviews of FE Vega even though it has gaming mode that enables gaming drivers.

This is what they have been hyping up and working on for 2-3 years? Just fire Raja and the marketing team of Radeon and start all over again.

Vega would have been great a year ago. Not now

 

[Glo.]

Tom's had a review of a 1080 Ti and the 1080 FE was something like 25-45% in games. This includes like Doom Vulkan and Hitman DX12.

What concerns me is that base clock. 1382 is not going to get it done.

Well I am asking not about reviews of the GPU, but game release reviews of performance. In most games of 2016 and 2017 I have seen Fury X 10-15% behind GTX 1080.

 

[Bouowmx]

Disqus user #define added 3DMark Fire Strike Extreme (10585 graphics score) and Ultra (5356 graphics score): https://disqus.com/home/discussion/...aunches_today_for_1200_us/#comment-3389895192

Spoiler: Images

 

[Head1985]

Disqus user #define added 3DMark Fire Strike Extreme (9449) and Ultra (5126): https://disqus.com/home/discussion/...aunches_today_for_1200_us/#comment-3389895192

Spoiler: Images

Yeah its 10% faster than reference GTX1080 and on par with aftermarket one.Time to move from GCN AMD really.

 

[Cloudfire777]

Disqus user #define added 3DMark Fire Strike Extreme (9449) and Ultra (5126): https://disqus.com/home/discussion/...aunches_today_for_1200_us/#comment-3389895192

Spoiler: Images

Thanks

GTX 1080 Firestrike Extreme: 10200
Vega FE: 10600
GTX 1080 Ti: 13600

GTX 1080 Firestrike Ultra: 5200
Vega FE: 5350
GTX 1080 Ti: 6700

 

[[DHT]Osiris]

Thanks

GTX 1080 Firestrike Extreme: 10200
Vega FE: 10600
GTX 1080 Ti: 13600

GTX 1080 Firestrike Ultra: 5200
Vega FE: 5350
GTX 1080 Ti: 6700

Day late and a dollar short. AMD needs to pull a rabbit out of its hat to avoid continuing to wallow in its inadequacy.

At least they'll still sell a million to miners, for whatever that's worth to a graphics card company.

 

[railven]

Woof. Welps, still got the caveat of wait for RX Vega. But...not looking good. Not sure how AMD (or us as consumers) expect improvements if this thing is say still slower than a GTX 1080 Ti with "proper" drivers, probably cost more to make, and will possible (as one poster said) sell for less or just not exist in the "gamer" side with all dies going to the "professional" side.

Well, at least it almost being 2018, I'd give a little more value to the "wait for DX12 games" propaganda that bolstered Fury back in 2015.

EDIT:

Day late and a dollar short. AMD needs to pull a rabbit out of its hat to avoid continuing to wallow in its inadequacy.

At least they'll still sell a million to miners, for whatever that's worth to a graphics card company.

If AMD were smart, they've have switch to focus on miners back in 2014. Gamers will be upset, but when you have money coming in to fund better R&D, you can win gamers back later (maybe).

But for a good portion of the last 5 years AMD needed MONEY. They had a surefire way to make that money but they didn't even bother.
"Sure we'll sell you this GPU for $200. What, you're just going to resell it for $500? We we still got our cut."

 

[linkgoron]

The way this card has been released, the secrecy, no benchmarks etc... I'm not expecting anything but disappointing gaming results.

 

[jpiniero]

If AMD were smart, they've have switch to focus on miners back in 2014. Gamers will be upset, but when you have money coming in to fund better R&D, you can win gamers back later (maybe).

GPU mining was pretty much dead until like a month or two ago.

 

[raghu78]

If Vega 10 is on par with an overclocked 1080 its a disaster much worse than Fury X and right up there with Bulldozer among AMD's worst failures.

 

[railven]

GPU mining was pretty much dead until like a month or two ago.

Not in 2014, where AMD could have made good money. THEN continue to make money after the bubble burst by selling their cards at what they normally would have sold them. They'd have sold a lot less cards @ 2xMSRP but when the bubble burst, they'd also have less used cards cannibalizing their sales.

Rinse repeat now, at least then you come out with someone in the end.