AMD Radeon RX Vega 64 and 56 Reviews [*UPDATED* Aug 28]

[Headfoot]

Some of you guys are hilarious with your expectations of this new magical driver.

When in the history of GPU's has there ever been a driver release that boosted performance over 20%?(other than when nvidia Multithreaded their drivers a few years ago)

Of course it never having happened before doesnt mean it cant, but it does mean it is extremely unlikely.

I dont think anyone (reasonable, at least) is expecting a 20% across the board thing. I do suspect that the new geometry capability is powerful, or they wouldn't have spent so much time on it and talking about it. It seems like the primary design change for Vega beyond clockspeed and HBCC. So I think we could see a variable increase in performance, maybe 10-15% at best, from that. Depending on how bottlenecked something is by geometery - probably a good number of games, knowing that the current geometry increase in Vega from Fiji is just clock speed (and Fiji was already geometry bottlenecked in 2015). Maybe another 5% from fixing AVFS and sorting out the throttling. So maybe you'd see a total of 20% in the best case scenario, but probably closer to 5-15% averaged. Which isn't going to change the relative standing of Vega but it would definitely make it more competitive.

I'm still disappointed, because what us enthusiasts really would have liked was enough old-school hard geometry resources so that it was even with 1080 Ti, with the new primitive capability being the icing 'Wow' factor. ROP/Geometry unit count remaining the same since Hawaii is plain out disappointing.

 

[krumme]

Some of you guys are hilarious with your expectations of this new magical driver.

When in the history of GPU's has there ever been a driver release that boosted performance over 20%?(other than when nvidia Multithreaded their drivers a few years ago)

Of course it never having happened before doesnt mean it cant, but it does mean it is extremely unlikely.

When in the history of gpu have we seen a product where most of the new features is not enabled?

What vega brings now is faster frequency. Heck even the avfs is beta.

I dont think we get it all working optimal in a single magic driver. But ngg and proper efficiency from the hbm? Yeaa we need a poll.
Whats your take on those two combined?
1-2% naa
15% ? You have slammed Glo for nearly 2 months now. Lets see the numbers !

 

[maddie]

Some of you guys are hilarious with your expectations of this new magical driver.

When in the history of GPU's has there ever been a driver release that boosted performance over 20%?(other than when nvidia Multithreaded their drivers a few years ago)

Of course it never having happened before doesnt mean it cant, but it does mean it is extremely unlikely.

If one has no knowledge about a situation, then historical precedents will be the best data available. In this case, we know for a fact that several performance enhancing features are not operating at present. If made to work, they can make a large immediate improvement.

What's the magic with that?

 

[Despoiler]

Some of you guys are hilarious with your expectations of this new magical driver.

When in the history of GPU's has there ever been a driver release that boosted performance over 20%?(other than when nvidia Multithreaded their drivers a few years ago)

Of course it never having happened before doesnt mean it cant, but it does mean it is extremely unlikely.

Wait so big driver improvements can't happen except for when they can? I realize you hedged with a likelihood statement, but still.... Something that has been bugging me is this whole "Poor Volta" advert. You have to question why would AMD put out such a bold claim when the perf we are looking at cannot compete. It's not like AMD doesn't know what Pascal's performance is and can assume that Volta will be faster. So why make the claim? One explanation is drug induced levels of fantasy from AMD's GPU marketing team. I don't buy that. Typically marketing generates its talking points from the technical specs and things the technical side of the house is telling them. How would they have heard anything about competing with Volta if the engineers are telling them the chip is targeted at Pascal performance? It doesn't make a lot of sense to me. There could be another explanation. I think the more likely angle is that AMD released a chip that targeted Pascal perf right now with the intent that with all features enabled or able to be envoked would target Volta. We know several of these key features are disabled or only leveraged in certain apps. Clearly Vega isn't showing all of its cards so to speak. We also have Raja making statements that the software side of Vega has been complex. We also have people like zlatan saying NGG fast path has a possibility to release late Q4(pure speculation on his part, but the path doesn't currently exist is the point). Anyways my point is that if there is any time that drivers could provide a big boost in perf, it would be this scenario. Major features turned off at launch is not typical. We should expect not typical increases in performance.

 

[Mopetar]

Some of you guys are hilarious with your expectations of this new magical driver.

When in the history of GPU's has there ever been a driver release that boosted performance over 20%?(other than when nvidia Multithreaded their drivers a few years ago)

Of course it never having happened before doesnt mean it cant, but it does mean it is extremely unlikely.

Even if it were possible I don't think we'll see it. It's far more likely to see ~20% in total performance over several years. Getting a 5% general performance improvement (obviously some games will be better, but I just mean across the board average) this year would go a long way towards improving Vega's value. It would make Vega 56 far more compelling than the 1070 assuming both continue to have similar prices above MSRP, and would make Vega 64 at least worth the consideration for more people.

 

[Rifter]

When in the history of gpu have we seen a product where most of the new features is not enabled?

What vega brings now is faster frequency. Heck even the avfs is beta.

I dont think we get it all working optimal in a single magic driver. But ngg and proper efficiency from the hbm? Yeaa we need a poll.
Whats your take on those two combined?
1-2% naa
15% ? You have slammed Glo for nearly 2 months now. Lets see the numbers !

The poster above my post mentioned 50%....

My take is this will be less than 15% from these disabled features in most games, possibly more in ones that AMD specifically optimizes for but not more than 15% on average across all games. Otherwise they would never have released with them disabled in the first place.

The reviewers who did benchmark the disabled memory feature found it actually performed worse or the same, no one has showed it with more of a increase than a few percentage points which are within the margin of error.

 

[EXCellR8]

well... looks like won't be settling for a ref 56 until the better 64's are out in the fall some time. sold out across the board to people who will likely pout about how their 1070 is superior and better value. oh well, not for trying but this party's a hot mess.

 

[Rifter]

Wait so big driver improvements can't happen except for when they can? I realize you hedged with a likelihood statement, but still.... Something that has been bugging me is this whole "Poor Volta" advert. You have to question why would AMD put out such a bold claim when the perf we are looking at cannot compete. It's not like AMD doesn't know what Pascal's performance is and can assume that Volta will be faster. So why make the claim? One explanation is drug induced levels of fantasy from AMD's GPU marketing team. I don't buy that. Typically marketing generates its talking points from the technical specs and things the technical side of the house is telling them. How would they have heard anything about competing with Volta if the engineers are telling them the chip is targeted at Pascal performance? It doesn't make a lot of sense to me. There could be another explanation. I think the more likely angle is that AMD released a chip that targeted Pascal perf right now with the intent that with all features enabled or able to be envoked would target Volta. We know several of these key features are disabled or only leveraged in certain apps. Clearly Vega isn't showing all of its cards so to speak. We also have Raja making statements that the software side of Vega has been complex. We also have people like zlatan saying NGG fast path has a possibility to release late Q4(pure speculation on his part, but the path doesn't currently exist is the point). Anyways my point is that if there is any time that drivers could provide a big boost in perf, it would be this scenario. Major features turned off at launch is not typical. We should expect not typical increases in performance.

AMD's "poor Volta" marketing is just pure Marketing BS, and its outright dishonest. Trying to string along any fanboys left on the fence, or people who dislike Nvidia enough to possibly buy AMD even though the facts show Nvidia has clearly won this round of GPU's.

Vega isnt fast or efficient enough to say poor Pascal let alone poor volta. Very dishonest on AMD's part, they should not have lowered themselves to that level of marketing.

 

[maddie]

The poster above my post mentioned 50%....

My take is this will be less than 15% from these disabled features in most games, possibly more in ones that AMD specifically optimizes for but not more than 15% on average across all games. Otherwise they would never have released with them disabled in the first place.

The reviewers who did benchmark the disabled memory feature found it actually performed worse or the same, no one has showed it with more of a increase than a few percentage points which are within the margin of error.

Is that your prime reason for thinking as you do? Cards are still selling. Do you give up those sales? Trying to understand.

 

[Rifter]

Is that your prime reason for thinking as you do? Cards are still selling. Do you give up those sales? Trying to understand.

Giving up a few weeks of sales to not have this disaster of a launch would have been a good idea yeah, if they can fix this in software soon. You only get one first impression. No one is going to re-review these cards until volta launches no matter how AMD tries to spin it.

 

[Janooo]

The poster above my post mentioned 50%....

My take is this will be less than 15% from these disabled features in most games, possibly more in ones that AMD specifically optimizes for but not more than 15% on average across all games. Otherwise they would never have released with them disabled in the first place.

The reviewers who did benchmark the disabled memory feature found it actually performed worse or the same, no one has showed it with more of a increase than a few percentage points which are within the margin of error.

I have a limited knowledge, I admit, but what I read about NGG makes me believe that we are not going to see a "regular" 15% increase from drivers that AMD was providing even without disable features in their HW before.

This appears to be a completely different scenario, a much bigger change.
The improvement estimates are a fun game, entertainment. I stand by mine >50%.

 

[moonbogg]

I think after the dust (and prices) settle down, Vega will be seen as a reasonable AMD choice, the 56 model in particular. It looks bad right now because of disappointment VS the competition, availability, delayed release etc etc. But, it is what it is. A capable enough gaming card for 1440p. Hell, plug the card in and it will work. Whatever. Wanted to see better from AMD but whatever man.

 

[Rifter]

If one has no knowledge about a situation, then historical precedents will be the best data available. In this case, we know for a fact that several performance enhancing features are not operating at present. If made to work, they can make a large immediate improvement.

What's the magic with that?

Please provide independent benchmarks showing these performance enhancing features actually do what AMD says they do.

 

[Konan]

If one has no knowledge about a situation, then historical precedents will be the best data available. In this case, we know for a fact that several performance enhancing features are not operating at present. If made to work, they can make a large immediate improvement.

What's the magic with that?

Hi Maddie, do you mean if a switch was flicked (a driver update) and things are "turned on" does that mean real world performance, say in games?
Or do you mean from a capability perspective? As in, the features are there ready for someone to code to and make use of, so over time, like a 3 year game development these features can be implemented?

Immediate capability improvement or future capability improvement that requires additional development work?

 

[LTC8K6]

I have a limited knowledge, I admit, but what I read about NGG makes me believe that we are not going to see a "regular" 15% increase from drivers that AMD was providing even without disable features in their HW before.

This appears to be a completely different scenario, a much bigger change.
The improvement estimates are a fun game, entertainment. I stand by mine >50%.

You probably won't sell very many new cards with +50% performance coming for the old cards...
It also kinda' means that your new cards have to be 70 or 80 percent faster than the old cards at launch.

 

[raghu78]

Giving up a few weeks of sales to not have this disaster of a launch would have been a good idea yeah, if they can fix this in software soon. You only get one first impression. No one is going to re-review these cards until volta launches no matter how AMD tries to spin it.

Thats the big IF. Since AMD have launched with drivers which are far from well optimized and we have Raja stating that developing drivers for this architecture is hard it seems the software part of Vega could take upto 4-6 months to get right. AMD have a major release every year in Dec so that would be the first time window for AMD to try and meet. If not then it would be atleast in time for Volta launch.

 

[EXCellR8]

hmmm... called a former contact from previous job and he's a manager at the local Best Buy--well local-ish. said there have been a few returns of vega 64 and that they haven't been restocking them due to internal refunding/RMA complications. there's a slim chance I might be able to grab one though, so I'm keeping tabs on them but I was told it was more likely that the cards were going to be returned to the "warehouse for processing."

all i know is best buy has been sold out since I started poking around a couple of weeks ago.

 

[Rifter]

hmmm... called a former contact from previous job and he's a manager at the local Best Buy--well local-ish. said there have been a few returns of vega 64 and that they haven't been restocking them due to internal refunding/RMA complications. there's a slim chance I might be able to grab one though, so I'm keeping tabs on them but I was told it was more likely that the cards were going to be returned to the "warehouse for processing."

all i know is best buy has been sold out since I started poking around a couple of weeks ago.

How many units have they moved? are they out of stock due to receiving a token amount of cards or are they out of stock and actually moving a real number of cards?

 

[Janooo]

Please provide independent benchmarks showing these performance enhancing features actually do what AMD says they do.

Your request does not make any sense.
He said, the features are not enabled. Who can test something that is not enabled?
Then he continued... "if made to work..."
All of us are guessing what NGG might bring in the future. If AMD does not make it happen, I am not excluding that possibility, then it's going to be one big fat zero.

 

[tential]

I think after the dust (and prices) settle down, Vega will be seen as a reasonable AMD choice, the 56 model in particular. It looks bad right now because of disappointment VS the competition, availability, delayed release etc etc. But, it is what it is. A capable enough gaming card for 1440p. Hell, plug the card in and it will work. Whatever. Wanted to see better from AMD but whatever man.

It looks bad now because it isn't good.
It won't get better later because Volta will be coming closer and closer providing an ACTUAL generational leap in performance.

Agreed, I put 15% as a not too ridiculous number. A possibility of a larger improvement exists depending on specifics.

The IWD also looks to have a lot of potential. Haven't seen any mention of this. From the same whitepaper.

"Some draw calls also include many small instances (i.e.,
they render many similar versions of a simple object). If an
instance does not include enough primitives to fill a
wavefront of 64 threads, then it cannot take full advantage
of the GPU’s parallel processing capability, and some
proportion of the GPU's capacity goes unused. The IWD
can mitigate this effect by packing multiple small instances
into a single wavefront, providing a substantial boost to
utilization."

We can all agree that AMD has severly lagged Nvidia in utilizing their shaders efficiently.

When will ya'll be done posting about potential large performance gains for Vega? I mean, it gets tiring hearing about the potential huge performance gain that will make Vega a great chip if we only wait. Wait for AMD is something we hear nonstop it's not enjoyable in the slightest.

 

[maddie]

Hi Maddie, do you mean if a switch was flicked (a driver update) and things are "turned on" does that mean real world performance, say in games?
Or do you mean from a capability perspective? As in, the features are there ready for someone to code to and make use of, so over time, like a 3 year game development these features can be implemented?

Immediate capability improvement or future capability improvement that requires additional development work?

An example of what a driver upgrade could accomplish.

Regarding Vega's Primitive Shaders & NGG Fast path from RTG Game Engineering Rys Sommefeldt.

Rys Sommefeldt:
"The driver takes care of it for you. Just render as normal."
Petrus Laine:
"What about the NGG Fast Path culling? To my understanding it needs primitive shaders, do the devs have control to use it instead of native?"
Rys Sommefeldt:
"No control, it's all transparent"
CatMerc:
"Wait so primitive shaders just work automatically from a developer's point of view? It's done in driver?"
Rys Sommefeldt:
"Yup!"

edit:
Some of the new features can be implemented in the drivers, so an enabled driver will give a performance increase for those features. Other features, rapid packed math, for example, will need coding changes.

 

[Janooo]

You probably won't sell very many new cards with +50% performance coming for the old cards...
It also kinda' means that your new cards have to be 70 or 80 percent faster than the old cards at launch.

AMD needs new features working flawlessly to compete with Volta. Their next release is far away.
As it stands, Fury to Vega performance increase is a "joke". Yes, 70%-80% faster for Navi when compared to launched Vega.
Navi would have those 50% already in so it would be only 20%-30% compared to what Vega should have been at launch.

 

[maddie]

It looks bad now because it isn't good.
It won't get better later because Volta will be coming closer and closer providing an ACTUAL generational leap in performance.



When will ya'll be done posting about potential large performance gains for Vega? I mean, it gets tiring hearing about the potential huge performance gain that will make Vega a great chip if we only wait. Wait for AMD is something we hear nonstop it's not enjoyable in the slightest.

I'm not telling anyone to wait. Only discussing some exciting new features.

Did AMD screw up the release? yes. Does that automatically mean nothing is forthcoming? No.

If tiring, just ignore. Not forcing anyone to read.

 

[tential]

I'm not telling anyone to wait. Only discussing some exciting new features.

Did AMD screw up the release? yes. Does that automatically mean nothing is forthcoming? No.

If tiring, just ignore. Not forcing anyone to read.

It's not this, it's that you're saying there are features that amd just needs to flip a switch for. Or that you make sound easy to have happen. If these were simply things they had to enable in drivers they would have done so.

Amd says these features are on and then we have forum members saying to wait for the large performance gain once the feature is actually on....

 

[deathBOB]

If the features you all are discussing 1) provided improved performance and 2) were being enabled by an imminent driver release, AMD would have said as much at launch.

I'm betting that the effects on performance are uneven or it will be a while before we get fully functional drivers.