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

[tviceman]

Why would Nvidia show pre-launch benchmarks? They are (by a very large margin) the market leader, and showing pre-launch benchmarks essentially only hurts their own sales. AMD, on the other hand, have nothing to lose if they've got a winner. Sadly, It appears that they do not.

I said the exact same thing about six weeks ago on here and was lambasted for having a double standard. It's very obvious AMD has a dud on their hands and it's why they're not showing anything other than blind comparison's against Nvidia's 3rd best card.

And further, everyone who thinks Freesync permanently locks people into AMD cards are blind. The adaptive refresh of freesync ability may be limited to AMD cards, but the monitor will still work on ANY video card and if AMD doesn't deliver a product consumers are happy with, then the consumer either not buy a video card or will buy an Nvidia card on the hopes that future AMD cards will get better by the time they upgrade again.

 

[IllogicalGlory]

Have you seen the amount of nonsense that idiot spreads all over the internet? Something like this? Or this? Or this? Or this? Or this? Or this? Or this? LOL? Sure? Wow. hmm... mkeeey. Ok. And let's not forget this.

And a moment of rare honesty.

Stop evangelizing people that tell you what you want to hear.

Great post. We should be showing it to people every time some fantasy rumor of their is making the rounds.

 

[Cloudfire777]

Hype hype hype

https://twitter.com/AMD_UK/status/887653978049318914

 

[beginner99]

Hype hype hype

https://twitter.com/AMD_UK/status/887653978049318914

Yeah it was hardly possible to not beat a 1080 and consume double the power...

If there isn't a magic surprise with Vega launch, like it comes close to 1080ti + price, then this is just one of many idiotic teasers in a long chain.

 

[wanderica]

Hype hype hype

https://twitter.com/AMD_UK/status/887653978049318914

Interesting, but I'm not sure there's anything to take from that. "Make the impossible possible." "Aim for the stars." "Dare to dream." Be a pioneer with Frontier Edition, etc., seems to be a theme of sorts with Vega. In other words, I suspect it's just marketing speak for "Come see AMD at Siggraph."

Unless . . .

The "impossible" part is talking price. 1080 performance at an impossible price would certainly perk my ears after all the poor numbers we've seen from FE. Ryzen price / performance ended up in a far better place than expected considering its performance and where the competition prices their HEDT parts. Maybe it's a new theme for AMD this round. Or maybe that's just the optimist in me talking.

 

[Veradun]

Why would Nvidia show pre-launch benchmarks? They are (by a very large margin) the market leader, and showing pre-launch benchmarks essentially only hurts their own sales. AMD, on the other hand, have nothing to lose if they've got a winner. Sadly, It appears that they do not.

Intel, as market leader on CPUs by a really larger margin than nVidia is on GPUs, is known to have shown pre-launch benchmarks for ages now. Is your point that Osbourne Effect only applies to GPUs?

I said the exact same thing about six weeks ago on here and was lambasted for having a double standard. It's very obvious AMD has a dud on their hands and it's why they're not showing anything other than blind comparison's against Nvidia's 3rd best card.

And further, everyone who thinks Freesync permanently locks people into AMD cards are blind. The adaptive refresh of freesync ability may be limited to AMD cards, but the monitor will still work on ANY video card and if AMD doesn't deliver a product consumers are happy with, then the consumer either not buy a video card or will buy an Nvidia card on the hopes that future AMD cards will get better by the time they upgrade again.

Aaaaand, AMD has never shown GPU pre-launch benchmarks, at least since I can remember (meaning "from the beginning").

I'm not telling you Vega is surely a star. I'm telling you that you can't infer product quality and/or performance from the fact they didn't give you all the details prior to launch. Giving details is the whole point of having a launch.

 

[Muhammed]

Hype hype hype

https://twitter.com/AMD_UK/status/887653978049318914

They are not hyping anything, they said the same thing a month ago, mentioning Ryzen, Epyc, and ThreadRipper. That's their impossible for SIGGRAPH.
http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=threads/vega-navi-rumors-updated.2486940/page-189#post-38991933

 

[exquisitechar]

Great post. We should be showing it to people every time some fantasy rumor of their is making the rounds.

Fottemberg posted on SemiAccurate forums that the "AMD is sandbagging" thing is just what he personally believes, he has nothing to back it up with, in case anyone was wondering. Honestly, "sandbagging" would make zero sense in this case.

 

[Tee9000]

You do realise he might be accurate about the die size,since AMD actually released a picture of the bare RR chip next to a Epyc/Threadripper chip??

Someone actually superimposed a picture of the chip without the heatspreader on top:

http://i.imgur.com/fb4kFYB.png

You do also need to stop getting so worked up about rumours - all the websites are doing it nowadays as they re-publish rumours to gain page hits and some like WCCFTech despite being inaccurate a lot of times,now actually gets invited to more events by companies and they get hardware to review.

If anything by continuously talking about them you give them more exposure which means they make more out of it.

In the end nothing is final until the product is launched and that is how it has been for decades,and rumours are just that,rumours.

What are you saying? He might be accurate in size? We have leaked AMD internal documents that prove he's statement is full of shait. Because he might be right in one statement we have to eat the pile of BS that comes with it? Come on.

I'm ok with rumours, but making stuff up or saying things as if you have some inside info that is pure BS, that spreads with the speed of light all over the internet, is another thing. And I'm looking at $850 TR. Him and rumour sites are making a living citing his fantasies.

 

[Veradun]

I'm ok with rumours, but making stuff up or saying things as if you have some inside info that is pure BS, that spreads with the speed of light all over the internet, is another thing. And I'm looking at $850 TR.

Forgive me, you are looking at what?

 

[Tee9000]

Forgive me, you are looking at what?

This.

Most tech press took his words as gospel, as leaked info, and ran with it. He intoxicated most of the internet with his lies/ his own opinions camouflaged as leaks.

 

[Veradun]

This.

Most tech press took his words as gospel, as leaked info, and ran with it. He intoxicated most of the internet with his lies/ his own opinions camouflaged as leaks.

Oh that, that's still something entirely possible, given they have a socket and a motherboard lineup for just 2 SKUs :>

 

[Carfax83]

I rarely look into this thread, but from the last few pages it seems the worst case scenario for Vega has manifested. Roughly similar performance as a GTX 1080, but far less energy efficient. I've always said that AMD will never regain the performance crown, as long as the performance per watt delta between them and NVidia is so massive. Unfortunately it doesn't seem that Vega will make any serious inroads into closing that gap.

The only thing I can say is that it might be possible that AMD will continue to optimize their drivers and somehow get a large performance boost several months after Vega launches. This has happened before, namely with the 7970 but with other GCN cards as well to a lesser extent. So it might be possible for Vega to creep up on the 1080 Ti in the next six months to a year, though most likely it will never match it.

This is the exact reason why I never really buy AMD products. Endless delays, hype and terrible execution. Nobody wants to buy a GPU that is significantly slower at launch than it should be, and wait months or even years until the drivers finally manage to do it justice.

It also kind of forebodes that NVidia might be on to something by using hybrid scheduling which not only saves on power, but might very well increase performance compared to using strictly hardware based scheduling like AMD. Using conventional logic, one would likely assume that a strictly hardware based scheduler would always be better than a hybrid software based one due to being local to the GPU, but when you think about it, modern CPUs are far more powerful and have much greater out of order processing capability than any hardware scheduler which is comparatively far more simplistic in terms of internal logic.

When you look at it from that perspective, it doesn't seem foolish at all to use the CPU to schedule and optimize the instruction stream to the GPU rather than an internal hardware scheduler. The biggest disadvantage to using software scheduling would be increased latency due to the CPU being off die, but GPU workloads aren't particularly susceptible to latency as CPU workloads.

 

[Tweak155]

Fottemberg posted on SemiAccurate forums that the "AMD is sandbagging" thing is just what he personally believes, he has nothing to back it up with, in case anyone was wondering. Honestly, "sandbagging" would make zero sense in this case.

Which is precisely the reason to do it, muwahahaha... hahaha... ha? Yeah I don't know. I do hope RX Vega pulls SOMETHING off, even if it's a free bumper sticker.

 

[nathanddrews]

I think we already know what the "impossible" is: SIGGRAPH is an association of professional graphics designers. They are going to show off how for a similar pricer Vega FE is better than GTX Titan for select professional applications, the same benchmarks we've already seen. Then they'll show us all sorts of new Machine Learning/AI/TensorFlow/OpenCL stuff using Vega Instinct and how much better it is than GP100 for less money or something along those lines... It's going to be about professional value across the spectrum of AMD products from EPYC to ThreadRipper to Ryzen to Vega FE. I mean, that's nothing to scoff at - AMD is serving up unseen value in the face of Intel and Titan in this regard.

I'd love to be proven wrong and see RX Vega launched for $349, beating 1080Ti, but that's not likely... (sarcasm BTW)

 

[tviceman]

Intel, as market leader on CPUs by a really larger margin than nVidia is on GPUs, is known to have shown pre-launch benchmarks for ages now. Is your point that Osbourne Effect only applies to GPUs?



Aaaaand, AMD has never shown GPU pre-launch benchmarks, at least since I can remember (meaning "from the beginning").

I'm not telling you Vega is surely a star. I'm telling you that you can't infer product quality and/or performance from the fact they didn't give you all the details prior to launch. Giving details is the whole point of having a launch.

If it smells like a duck, quacks like a duck, and looks like a duck then the guy who is telling you that it might not be a duck is full of it.

 

[Despoiler]

"Showing off the impossible" links to the PRO Radeon site not the gaming Radeon site. People that think they are talking about RX Vega are the reason why impossible hype exists. Then they blame AMD for their own mental gymnastics.

 

[R81Z3N1]

I have to agree on the general assumption that consumer Volta is a ways off, another aspect is just look at the supply and demand of current cards. Most cards are way above suggested retail price, unless you want a basic card, even those are pushing the limits.

Really no incentive for rushing out new cards, unless your AMD it seems that as long as demand out stripes supply they are selling everything they produce I don't think that is an environment for releasing new products. I do hope the best for Vega, and if the price is right and performance is good I could see folks picking up cards.

In a way this might be the best time for AMD, release a product that is semi-competitive due to over inflated prices on general gpu's and due to the shortage of product sale everything due to constraints and profit all around. If things are really that bad give more orders for products that folks want ie more 570, 580 chips. People might buy Vega, as that for the price is the only card they can find, even if it turns out to be a little piggy in the power department.

 

[Muhammed]

We all have seen this coming:

1-In 2016, Raja talked about the uselessness of Ultra settings, and how the High settings are more important as they are more "optimized".
2-RX Vega Shown slightly faster than 1080 in Doom @Ultra settings (not Nightmare)
3-AMD showing Titan X with BF1 Ryzen demos, instead of RX Vega
4-Consistent lack of fps counters in several consecutive Vega demos.
5-Lisa Su shivering voice when talking about RX Vega
6-Lack of fps counters in TitanXp vs Vega FE gaming comparisons
7-Vega FE gaming numbers are underwhelming, between 1070 and 1080, when stable @1600MHz, it trade blows with the 1080.
8-Several AMD officials stating drivers for FE are not gimped, and are not Fiji drivers.
9-RX Vega compared to the 1080 in a demo without fps counters again.

It doesn't get much clearer than that. AMD thought Vega is competitive with the 1080, they were happy showing fps numbers for the card back then, after 1080TI launched they've changed tactics, and chose to hide their numbers. Now Vega FE comes with a lower performance target than 1080, because of it's throttling and downclocking, it needs water cooling to stay competitive, and AMD wasn't shy about announcing water cooling versions of Vega, this is FuryX 2.0 all over again, but this time it's much worse.

 

[USER8000]

What are you saying? He might be accurate in size? We have leaked AMD internal documents that prove he's statement is full of shait. Because he might be right in one statement we have to eat the pile of BS that comes with it? Come on.

I'm ok with rumours, but making stuff up or saying things as if you have some inside info that is pure BS, that spreads with the speed of light all over the internet, is another thing. And I'm looking at $850 TR. Him and rumour sites are making a living citing his fantasies.

But in your annoyance you seem to have not even bothered to look at the picture - you said the 210mm2 die estimate was a lie,and yet that picture shows it is probably close to that level. The fact of the matter you seem utterly new to rumour sites - they have been around for at least a decade. Sometimes they are right sometimes they are wrong. If you look at the chap's history he has actually mentioned some things which came to fruition - one of them was that AMD was working on a GPU over 500MM2,which nobody believed. It turned out to be Fiji.

But at the same time he has just obviously made stuff up since it was a slow news day.

Its a rumour site - take everything with a pinch of salt.

You are just raging at some chap on a rumour site,and the hilarious thing is he is getting more and more exposure because of your forum vendetta. There have been loads of rumour and clickbait sites such as WCCFtech which make up stuff too,and they are not the first or the last in over a decade. The more you moan at them the more exposure they get,and it increases the chance of them getting hardware to review and being invited to events.

Look at WCCFTech - people kept bringing them up despite their crap track record(remember the "beast mode" RX480 rubbish) and now they are apparently being invited to events and getting hardware sent to them,whilst making money from all the clicks.

Does WCCFTech have the level of journalist integrity that Anand had when he started this website when he was a teenager - most likely not.

You only have yourself to blame in believing ALL of it and annoying yourself. It was like with that semiaccurate site,sometimes it could be right and other times it was utterly wrong - people like you were moaning at the owner and it kept the site in the public eye,especially when it was quite apparent he was posting in such away to make sure he annoyed people so they would keep mentioning him.

These sites live or die on exposure and page hits.

Best thing is to ignore all rumour sites if you want accuracy or proper reporting.

Its like reading the National Enquirer and getting annoyed when at least half of it is not true. In the end you enabled them,and they are making the money,not you.

 

[Glo.]

Have you seen the amount of nonsense that idiot spreads all over the internet? Something like this? Or this? Or this? Or this? Or this? Or this? Or this? LOL? Sure? Wow. hmm... mkeeey. Ok. And let's not forget this.

And a moment of rare honesty.

Stop evangelizing people that tell you what you want to hear.

Why would products not being released today show that BitsAndChips are not reliable?

What if most of this will turn into reality?

 

[IllogicalGlory]

Why would products not being released today show that BitsAndChips are not reliable?

What if most of this will turn into reality?

I'm sure price cuts on the 6900K, 3GHz Pascal and Broadwell-E on 14nm+ are right around the corner. Also 32 core Skylake Xeons. Oh wait that was just his "hope", he didn't have a source. I wonder if that applies to every single one of those?

Threadripper 16 core is $999, not $850 as well.

It's just him making things up.

 

[Erithan13]

Polaris has been out for over a year, and Vega is just about to be launched (and lord help us it could be a flaky paper launch at that). That is not something I would have guessed when I splashed out on a freesync monitor last August. So it's pretty much indisputable that something has gone seriously, seriously wrong with Vega, and the only question is how much it could be salvaged. In the case of launching a year later than the competition and the FE power/performance issues, the question is whether it can even be salvaged at all. I wonder if AMD have considered just canning RX Vega completely and taking the hit as overall less damaging than what could be the next Bulldozer.

I was watching this earlier which has a neat summary of the timeline since late last year. The thing that stood out was how relatively open AMD were being about Vega until just after the infamous 'Poor Volta' ad. It's been what, 6 months of effective radio silence since then? Only a very sketchy FE Vega in the meantime and the RX launch getting pushed back as far as possible.

On paper, Vega is a monster. Just the sheer size of the chip, number of SPs, improvements from Fiji, 1600-1700mhz clocks etc, mean it should be stomping the 1080 and giving the Titan a run for the money. If I'm being optimistic then I honestly don't think Poor Volta was hubris. AMD have surprised us before and I think they could have had a 1080/1080ti killer brewing for an early Q1 launch. At the very least we should have gotten something halfway decent and maybe somewhat competitive against NV.

And you know the thing that frustrates me the most? We're probably never going to know what happened to poor Vega. Oh, if this is as bad as it's looking, we'll have a full Bulldozer style postmortem going on with people decompiling the drivers bit by bit and doing god knows what to the hardware to try and figure out how on earth they regressed from Fiji. I look forward to the endless theorycrafting and the same questions being asked as Bulldozer faced: Was this doomed before it even left the drawing board? Good design screwed over by a bad process? Hardware good but the drivers proved too much to handle? Did they get 99% of the way there and something unexpected came up and completely derailed them and they could never recover? Is Raja a cackling demon drinking our salty tears or is he currently cowering under a table smoking his last cigar before Lisa Su comes along to drown him in a barrel of FineWine(tm)?

 

[guskline]

We are so close to the RX Vega launch we might as well wait to see what happens. I think all realistic hopes of an impressive gpu launch are pretty well gone.

I hope for AMD/RTGs sake I'm wrong.

 

[Elixer]

On paper, Vega is a monster. Just the sheer size of the chip, number of SPs, improvements from Fiji, 1600-1700mhz clocks etc, mean it should be stomping the 1080 and giving the Titan a run for the money. If I'm being optimistic then I honestly don't think Poor Volta was hubris. AMD have surprised us before and I think they could have had a 1080/1080ti killer brewing for an early Q1 launch. At the very least we should have gotten something halfway decent and maybe somewhat competitive against NV.

That is the whole conundrum here.

That is one big chip, on a smaller node, yet, it is barely beating Fury on some workloads.
As a prosumer card, it also is losing on workloads it shouldn't.

Worst case is, that the first iteration of Vega's plumbing is fubared, and they had to fuse it off, and write workarounds in the driver which makes it really, really slow for certain workloads. This makes the most sense right now considering what we have seen.

If that is the case, and since it takes so long to do a re-spin, the next iteration would be back from the fabs by the end of August. So, if all of the sudden they release a Vega Ti a few months from now, we know why.

That is all conjecture though.

Best case is, the drivers will bring a 30% performance boost to some things.

10 days to go, and I think this whole thread has pretty much covered every single possible explanation there could be for why Vega turned out the way it turned out.
Can't think of anything that anyone missed.