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

[Elixer]

Did you actually read the issues though?

Odd question, but yes.

It is live reloading drivers without a reboot. That's something that afaik hasn't been done, and NVidia doesn't even let you use different drivers on their pro/gaming cards.

What? Where did you come up with this?
You can't hide drivers, and if you looked in the Vega package, there aren't multiple dll's that are enabled via a "toggle".
AMD is doing codepath switching, as in If (VEGA) do this...type of code.
There are both 32 & 64 bit versions of all the primary libs (openGL, vulkan, mantle, DX). There isn't anything "extra".

Most of the rest are related to sleep / resume situations and don't happen during normal usage.

The biggest one that you missed is the 10 bit display issues.
Again, this is marketed to the Pro market, where it is very common to have 10 bit displays.
There is 0 excuse to release a card if it has 10-bit display issues.

The TDR's shouldn't be there at all for anything Pro related. We aren't talking about consumer drivers here, where they can get away with that.

 

[Kenmitch]

Vega RX condemned by Vega FE's launch drivers! Dang internet is getting crazy!

I'm with the waiting to see Vega RX's launch before passing judgement. Seems like just a few days ago the norm was to wait for the official launch of a product and it's reviews before the bashing.

 

[QualityTime]

Perfect. Also, PCB shots:
https://www.pcper.com/news/Graphics-Cards/Radeon-Vega-Frontier-Edition-GPU-and-PCB-Exposed

Empty connector pads and all those test points.
It's a pre-production card you'd seed to developers/companies. Explains the pro/game driver switching. Doesn't explain why ship it retail. Not something you'd rely on for production or gaming.

What the hell did they put in there to make that GPU 560mm²?! Did they throw density off a cliff to hit those clock speeds?

Thats the exciting question.

 

[vissarix]

So basically Amd were saying PoorVolta as a hint of upcoming Vega performance and in the end is barely faster then a Gtx980Ti?

That gotta be the biggest Fail in the history..

 

[deathBOB]

If FE is not representative of RX, all AMD needs to do is issue a statement saying that game performance will improve with FX/updated drivers. That would be entirely reasonable. Drivers are hard, shit happens, and consumers wouldn't hold it against AMD.

That they are completely silent is telling.

 

[Cloudfire777]

From reddit, not sure how reliable, but here it is:

https://www.reddit.com/r/Amd/comments/6kdwea/vega_fe_doesnt_seem_to_be_doing_tiled/djld98t/

This is so funny.

The usual excuse anytime, which is pretty much everytime, AMD launch a profuct:
Wait for drivers. It will fix everything.
DX12 will make it shine.
Developers just need to use a special code in the games.

Yeah, lets buy a product based on a wet fantasy that a magical driver will fix everything.

I guess buying and supporting an inferior product this is the only thing you cling on to and twist reality to make it seem like you did the right choice.

Eventually the card will be good.
Maybe
But probably not

 

[Cloudfire777]

So basically Amd were saying PoorVolta as a hint of upcoming Vega performance and in the end is barely faster then a Gtx980Ti?

That gotta be the biggest Fail in the history..

15% slower than GTX 1080. 14 months later.
But atleast it got HBM2.

Remember all the marketing slides from AMD when HBM was new? Their mojo was that it offer more bandwidth/watt.

Guess they got all worked up and forgot the prime contributor to power consumption:
The GPU cores.
Its like they moved on to memory because they couldnt keep up vs Nvidia on improving the power consumption on the cores

FE Vega consumes 100W more than GTX 1080 yet is 15% slower.
What on earth are they doing?

 

[Malogeek]

Vega RX condemned by Vega FE's launch drivers! Dang internet is getting crazy!

Vega silicon condemned by FE's launch. You're implying they're different.

 

[crashtech]

Perhaps it really is a coding nightmare to get all the extra hardware functionality just right. That would explain a lot.

 

[jpiniero]

Perhaps it really is a coding nightmare to get all the extra hardware functionality just right. That would explain a lot.

Yeah. I would expect 10-20% more performance once they get everything straightened out. How long is that going to take though?

 

[crashtech]

How long is that going to take though?

If you ask me, it seems like forever. I just post in here when I get too crazy from waiting.

 

[Malogeek]

Yeah. I would expect 10-20% more performance once they get everything straightened out. How long is that going to take though?

AMD's FineWine technology can span years! It's great for long term ownership though

 

[Mopetar]

AMD's FineWine technology can span years! It's great for long term ownership though

It wouldn't completely surprise me if in two years if the full RX Vega were faster than the 1080 Ti. But if the performance isn't significantly better by launch, I can't see too many people picking on up.

 

[.vodka]

FE Vega consumes 100W more than GTX 1080 yet is 15% slower.
What on earth are they doing?

Not to mention the die size, 564mm². What the hell?

Nothing makes any sense.

 

[ZipSpeed]

Raja is either sweating bullets right now, or he's laughing at all of us. Guess we'll find out in a month. Looks like AMD's direction with their CPU and GPUs is heading toward server and professional. Unfortunately for us gamers, AMD is chasing larger profits which is not unreasonable I suppose.

 

[Bacon1]

Got these off reddit:

Look to be the same pcper numbers but with the prices attached

Looks like it does well for a Semi-Pro card

 

[raghu78]

AMD could have easily avoided all this unnecessary speculation by launching Vega Radeon FE along with Radeon RX Vega. I am surprised that Vega die size is 564 sq mm. I remember reading somewhere that Raja Koduri stated Vega die size is slightly below 500 sq mm. If these measurements are correct he was way off the mark. I still think there are more pieces to this Vega puzzle which will unfold over the next month and half.

 

[jpiniero]

Looks like it does well for a Semi-Pro card

I think AMD would have been better off just leaving Vega FE in the Radeon Pro WX lineup, because that's what it is really. AMD's just heavily discounting the benefit of the pro drivers versus nVidia. Keep in mind the P5000 is more or less the Pro 1080.

 

[Bacon1]

I think AMD would have been better off just leaving Vega FE in the Radeon Pro WX lineup, because that's what it is really. AMD's just heavily discounting the benefit of the pro drivers versus nVidia. Keep in mind the P5000 is more or less the Pro 1080.

Well sounds like they have a Pro WX9100 coming out soon which will have the pure "Pro" drivers so who knows.... I've never understood the point of the Titans since they are always awful price/perf but people seem to love them. AMD wanted some of that massive profit margin before releasing the high end RX ("Ti").

 

[nathanddrews]

I think Ryan's assessment is correct in that if RX Vega comes to market with 1080 performance for $399 - that's going to be disruptive. Granted, that's assuming some enhancements can improve its performance and reduce its power consumption. Otherwise people will complain about "teh watz".

 

[Glo.]

What would be the point, that's Polaris performance territory right?

With current drivers - maybe.

As I have said, we do not know if the features that are supposed to bring the performance increase are working with Vega.

On the other hand you have compute benchmarks, which show even increase per clock, compared to Fiji. It is the same situation as is with Ryzen. In Engineering and scientific workloads, the CPU was on par with Broadwell-E chips. Where it lacked was gaming, because the software was never designed with it in mind.

Same thing appears to be with Vega.

Again: per clock, Vega appears to have higher compute throughput than Fiji. Which in itself indicates progress of this architecture.

Also, I see that shills have waken up again when there is upcoming AMD launch. Its good sign actually, that the product in the end will be very good. It was with Ryzen and one specific suspect which is apparent in this thread, was also apparent in Ryzen threads, downplaying Ryzen CPUs.

Other two are also here.

 

[senseamp]

If FE is not representative of RX, all AMD needs to do is issue a statement saying that game performance will improve with FX/updated drivers. That would be entirely reasonable. Drivers are hard, shit happens, and consumers wouldn't hold it against AMD.

That they are completely silent is telling.

They are catatonic. Millions of dollars worth of reputational damage being done to their brand, and no response. They should stop playing games.

 

[crisium]

So basically Amd were saying PoorVolta as a hint of upcoming Vega performance and in the end is barely faster then a Gtx980Ti?

If they don't improve performance tremendously, they are going to have serious trouble with GV106 / GTX 2060 which based on history will at least be comparable to GTX 1080.

 

[Bacon1]

I think Ryan's assessment is correct in that if RX Vega comes to market with 1080 performance for $399 - that's going to be disruptive. Granted, that's assuming some enhancements can improve its performance and reduce its power consumption. Otherwise people will complain about "teh watz".

Hah at this point they can launch at $600+ since mining is going nuts... I gave up hope for a well priced Vega weeks ago after the mining craze. Miners are even buying FE @ $1000 <.<

 

[Paratus]

AMD could have easily avoided all this unnecessary speculation by launching Vega Radeon FE along with Radeon RX Vega. I am surprised that Vega die size is 564 sq mm. I remember reading somewhere that Raja Koduri stated Vega die size is slightly below 500 sq mm. If these measurements are correct he was way off the mark. I still think there are more pieces to this Vega puzzle which will unfold over the next month and half.

If we go with the RX580 transistor density of 24.6M/mm^2 that puts Vega FE around 13.9B transistors or 60% more than Fury X

If the mediocre gaming performance holds the same for the RX Vega then what the hell are they doing with the extra 5B transistors?

I'd like AT to look into that.

I've got time before I buy a new card but if this performance holds it is disappointing in an HD 2900 way.