Will AMD release any high performance GPU's in Q2 or Q3?

[Kagthul]

Yeah, i just watched the GN breakdown of Vega FE.

... we're looking at another generation where AMD has no competition at the high end.

RX Vega isn't going to be different silicon. So.. its going to be about as good as a 1080.

And then about 2-4 months after it comes out, Volta is going to launch, and its going to be about as good as a 2060.

come on guys. You did some good work with Ryzen.

Get the damn lead out on the GPU side.

I'd like to have options when i rebuild with Coffee Lake here, rather than just defaulting to "well i guess ill grab whatever the best nVidia card is"

[Triggered Fridgekin]

I have a suspicion that RX Vega is going to be synonymous with paper launch which will only further exacerbate the issues surrounding this card. It also reminds me a lot of the R600 series which was mired in delays, performance and power issues while carrying a somewhat affordable price tag.

If Nvidia is making any headroom on a consumer-grade Volta for Q1 2018 release then AMD is going to get buried behind two generations if they push for the high-end market again. Heck, even their 1060 refresh might be enough to hold them over in the mid-range market for some time especially if 570s and 580s continue to be elusive in certain markets.

It's kind of funny in a way. For a while it was the GPU market carrying AMD with the CPU market dragging behind but now it could very well be the opposite.

[Gray_Matter]

Not only that, but the whole concept of AMD being a good buy from a price performance perspective gets undercut by the PSU and case cooling requirements that are likely to be slapped onto these cards. The RX 480 was successful, because it has extremely low heat/power draw and was an effective upgrade for people with less than top of the line hardware who were looking for an upgrade that could fit with pretty much any case or PSU. I think these cards are going to come with a 750W+ recommended PSU, which most people looking to buy at that price range probably don't have.

That's why I said a lower clocked one. I am guessing that we are seeing the power requirements dramatically higher than they should be because AMD are purposely pushing the cards higher than they planned. That means that clocking down by something like 10% might end up dropping the power requirements by 40%, as an example. Will have to wait and see, though. Any additional power requirements should of course be included with the cost. At least, in my case they will, because I am certainly not going to pay an extra $100 for a new PSU to get a card of similar performance. For me it's all about bang for buck. I am not in the 1080+ market and I have time to see what happens.

[Tiberria]

That's why I said a lower clocked one. I am guessing that we are seeing the power requirements dramatically higher than they should be because AMD are purposely pushing the cards higher than they planned. That means that clocking down by something like 10% might end up dropping the power requirements by 40%, as an example. Will have to wait and see, though. Any additional power requirements should of course be included with the cost. At least, in my case they will, because I am certainly not going to pay an extra $100 for a new PSU to get a card of similar performance. For me it's all about bang for buck. I am not in the 1080+ market and I have time to see what happens.

But, if they clock them down, don't they then risk pushing them from being between a 1070 and 1080 to at or below 1070 performance? If they do that, and even if they get the power requirements <200 W, they still probably have to undercut 1070 pricing numbers, because the driver weakness relative to Nvidia (real or imagined) means that most people would choose Nvidia if everything was equal.

I suspect that they deliberately have the Frontier clocked as high as possible to avoid the bad press/look that they would get if it was only at 1070 performance. After all, that would make Vega about the same relative performance as the RX 480 was when it first came out, and the RX GPUs were already considered a disappointment, because they meant AMD was abandoning the top of the GPU hierarchy. It would mean they've made no real progress.

[Zenny]

AMD reference cards and early drivers have always had issues, I think with proper AIB cards and better drivers the Vega will at least match the 1080 and maybe even exceed it in some DX12 titles.

That being said, the card has a much higher power draw then it's direct competition, and is coming absurdly late to the market. With a bit of smart shopping getting a 1080 for well under the $499 reference point (~$450) is rather easy, so this card needs to launch at around $400 to make a splash, at that price point it has great price/performance.

Problem is, Volta is around the corner and the 2060 may well match/be close to Vega performance at a even much lower price point. As well as being much cheaper to manufacture.

You could say wait for Navi, but Nvidia is not sitting on it's laurels either, I'm sure the next-gen architecture after Volta is on schedule as well.

[Linkedblade]

Well we have the announcement of release date at SIGGRAPH. So we shall see. The first release usually has some hell to go through with drivers. But the second release, as amd always does this, will probably on par with the 1080ti.

[Gray_Matter]

But, if they clock them down, don't they then risk pushing them from being between a 1070 and 1080 to at or below 1070 performance? If they do that, and even if they get the power requirements <200 W, they still probably have to undercut 1070 pricing numbers, because the driver weakness relative to Nvidia (real or imagined) means that most people would choose Nvidia if everything was equal.

I suspect that they deliberately have the Frontier clocked as high as possible to avoid the bad press/look that they would get if it was only at 1070 performance. After all, that would make Vega about the same relative performance as the RX 480 was when it first came out, and the RX GPUs were already considered a disappointment, because they meant AMD was abandoning the top of the GPU hierarchy. It would mean they've made no real progress.

Right now they are between the 1070 and 1080 with immature gaming drivers and a Prosumer card. It's very likely that they will end up somewhere between the 1080 and the 1080-ti with more mature drivers. As long as they price the card in line with the performance and slightly under the Nvidia equivalence then it should still be fine. They need volumes to drop the manufacturing cost for future and put themselves in a position where they can get better binning. The apple deal helps but they will have to provide an affordable option to push numbers. I can't guess how consumers will choose but big clients, like Apple, will certainly be willing to sign on the bottom line for a 10% discount. Those deals will be what AMD is looking for initially.

I think AMD tried the same thing with Vega that they did with Ryzen. A single core for all markets. In the case of Ryzen it was servers and workstations. In the case of Vega it's professional and gaming. I would take the present numbers with a pinch of salt, though. I think they still have some serious driver issues and it's not a card that's made 100% for gaming. I don't think that we are going to see a 50% speed jump but it will probably be 15%-20%. That's certainly enough to launch with, provided they have the right pricing. Don't work out the performance of the lower clocked variants based on the present gaming benchmarks. Base it on the present ones + 20%. That will give you a card that sitting at about 1070 + 10% with a reasonable power draw.

If they play well enough in the professional market and middle the gaming market then it will be enough for them to keep pushing into the future. The market for 1080ti cards really isn't very big. It doesn't look great but it won't hurt their bottom line as much as not having a good professional card will.

I do think that they are going to struggle at the top end at least until they release their 14+ Vega but probably until Navi.

[Thunderball]

Right now they are between the 1070 and 1080 with immature gaming drivers and a Prosumer card. It's very likely that they will end up somewhere between the 1080 and the 1080-ti with more mature drivers.

Right now they are 7-10% behind stock GTX 1080 Founders, and considering how powerful even the stock PCB is I have a lot of doubts that AIB partners can do anything do ramp up the performance here. Also, consumer variants are unlikely to use the full specs of the current Vega silicon. In the end there is no way it can beat 1080, I would be happy if it consistently beats 1070.

They need volumes to drop the manufacturing cost for future and put themselves in a position where they can get better binning. The apple deal helps but they will have to provide an affordable option to push numbers. I can't guess how consumers will choose but big clients, like Apple, will certainly be willing to sign on the bottom line for a 10% discount. Those deals will be what AMD is looking for initially.

Not going to happen. The die itself is huge, cards also use HBM, so no drops here. Apple and console market related products dont have anything to do with what they are offering for the desktop, those are custom designs, people who ordered those control everything: development, pricing and production.

I think they still have some serious driver issues and it's not a card that's made 100% for gaming. I don't think that we are going to see a 50% speed jump but it will probably be 15%-20%. That's certainly enough to launch with, provided they have the right pricing. Don't work out the performance of the lower clocked variants based on the present gaming benchmarks. Base it on the present ones + 20%. That will give you a card that sitting at about 1070 + 10% with a reasonable power draw.

People have been saying the same shit about Ryzen gaming performance, yet updates rolled out and literally nothing changed.

[Lathais]

Picking up a ruler and doing it, incredibly accurate. Which one of them do you prefer to believe in? Raja Koduri himself or PCP/GN? There is a packaging around the die to the HBM2 stacks. Measuring things you don't know how to measure only gives you inaccurate numbers. They did a good job taking into consideration what they could do, but that doesn't randomly change the actual size of the die.

To be honest the rest isn't even worth replying to, we still don't know whether the GPU is working with all the features or not. Just like the die measurements were all off, even if those rasterisation tests were all pretty convincing they could be wrong too. We don't know enough about the product to make conclusions yet, yes it could be a huge wasted effort with improvements that don't really translate into real world benefits in gaming scenarios, but it could also be that the software wasn't ready and they played safe to avoid functionality problems.

We simply don't know yet, the gaming cards aren't here. What we have here is a prosumer card that aims to do both, and currently isn't exactly doing either that well.

If that's the actual gaming performance, they'll just price it accordingly in order to still be able to sell a product. And it would be a sad failure.

If it isn't, we should see the actual performance once the gaming cards are released. The damage done by releasing gaming cards with bad gaming drivers is much more problematic than releasing a prosumer card with bad gaming drivers, so I hope they won't do this.

Again, let's wait and see.

Exactly. We have seen some MASSIVE improvements in gaming performance, even months after cards are released, by driver updates from AMD. If the gaming drivers are not optimized properly yet, which often does take AMD more time, then it makes total sense for them to have released the prosumer version and hold off on releasing the gaming version until the drivers are ready. This is something AMD has been really bad at in the past. Just look at what the Crimson drivers did for them. Went from being behind the competitors to ahead, though people still refuse to accept that. Perhaps they learned from that and are not releasing the card until the drivers are actually ready this time. Only time will really tell us.

- - - Updated - - -

How are you coming to that conclusion, drivers are not putting this card to 1080 levels. The card is closer to 1070 in the majority of game tests from both pcper and GN.

Because, drivers are not yet optimized for gaming. I once again refer you to the Crimson Driver release and the pretty massive improvement it gave to their cards.

[Fascinate]

Sure maybe a few games will get a slight uptick, but this card is slower than a 1080, drivers ain't changing that.

[Deleted]

Sure maybe a few games will get a slight uptick, but this card is slower than a 1080, drivers ain't changing that.

I am not even referring to drivers, old games use old rendering techniques, newer games use more advance techniques in the rendering pipeline which can take advantage of newer hardware features.

This is why as I said before, newer games just because they are newer perform well on AMD hardware, granted still well on Nvidia hardware but old games has Nvidia on top.

Mean, take a look at Titanfall 2 and compare it to GTA 5, GTA 5 is a fucking old game right now yet still used, if you want to use an old game for benchmarking purposes, use Crysis 1.

Newer games are more indicative at showing current hardware performance, however I can only see the Vega line only compete with the 1080, however its not a terrible thing out right though AMD does at some point need to put a card out that competes with the ti series.

Some people want at the least 1080 performance from an AMD card so they can use higher settings on high refresh rate freesync monitors, if its priced properly then its not a huge issue, people buy the performance they need from hardware really.

Also take a look at this;
https://www.techspot.com/review/1329...y-geforce-980/

The performance gap in modern titles from the 980ti and Fury X has diminished, the 980ti had a bigger gap then this when both were released and most of the games tested are NEW.

[Shakadam]

Sure maybe a few games will get a slight uptick, but this card is slower than a 1080, drivers ain't changing that.

As of right now the Vega FE is slower at the same clock speed than a Polaris chip or even a Fury chip. That's a pretty clear sign that the gaming drivers for Vega aren't ready yet or they could have simply scaled up the Polaris chip for better performance. Vega was created because of some design issues with the Fury chip which made AMD not opt for a simple die shrink, higher clock, and optimizations of the Fiji chip.

There's much speculation about what's going on, but the most sensible I've read is that Vega FE currently uses some kind of slightly modified Fiji gaming driver, because the gaming performance is a bit similar to a simply higher clocked Fury chip. The driver simply isn't taking advantage of the new features of the Vega chip.

This thread on Reddit is a good read if you wanna know a little of what's going on: https://www.reddit.com/r/Amd/comment...e_doing_tiled/

In short, there is some reason why RX Vega for gaming has a later release. I predict that there's going to be a fairly dramatic difference in performance when the "real" RX Vega is released, with a proper gaming driver designed for the chip.

[Lathais]

Sure maybe a few games will get a slight uptick, but this card is slower than a 1080, drivers ain't changing that.

The Fury X, on release with Catalyst drivers performed worse than a 980ti. After Crimson drivers, it performed better than better than a 980ti in newer titles, worse in older. Not sure why drivers in addition to other changes made to the card for gaming purposes. can't do the same thing here and make it better than a 1080 while it is currently worse. It's not like we have not seen drivers do this for AMD cards in the past. Guess we'll have to wait and see.

[Life-Binder]

https://videocardz.com/70777/amd-rad...11-performance

I didn’t really want to post this because I think it’s still too early, but since the highest score started to float around the web I think it’s worth to clear some misunderstanding. The highest score the 687F:C1 has achieved is an overclocked chip. 3DMark11 does not recognize unreleased overclocked graphics cards very well. The good news is that this puts RX Vega above overclocked GTX 1070, bad news, it might still be slower than overclocked GTX 1080. I guess time will verify those results.

an improvement over FE Vega, its faster than a stock 1080 and about on par or slightly behind an OCed 1080

although the 1630 Mhz is probably already close to Vegas max OC, judging from FE OCing

[Lathais]

https://videocardz.com/70777/amd-rad...11-performance






an improvement over FE Vega, its faster than a stock 1080 and about on par or slightly behind an OCed 1080

although the 1630 Mhz is probably already close to Vegas max OC, judging from FE OCing

and this is still the prosumer card with early drivers?

[Life-Binder]

this is the RX Vega

RX Vega = gaming card

you can even see the exact same device ID they used in their Doom/Battlefront RX Vega demos, as well as 8GB VRAM (so its not FE with its 16GB)



drivers = well they have only a little over 3 weeks left to make RX drivers unless Vega isnt actually launching at the end of July

[Tiberria]

The Fury X, on release with Catalyst drivers performed worse than a 980ti. After Crimson drivers, it performed better than better than a 980ti in newer titles, worse in older. Not sure why drivers in addition to other changes made to the card for gaming purposes. can't do the same thing here and make it better than a 1080 while it is currently worse. It's not like we have not seen drivers do this for AMD cards in the past. Guess we'll have to wait and see.

However, the Fury X was a huge failure in terms of sales numbers because it's not good enough to have drivers that eventually boost the card's performance 6 months to a year down the road on future titles. The people who wanted 980 Ti level performance would have already bought a 980 Ti by the time the driver situation was straightened out, and by the time it was under control, the 1080 was out anyway, obsoleting both cards in terms of future purchases. And, people don't just care about future AAA title performance; people still play current and past games. Think of how many people still play games like GTA5 and Witcher 3 for example. The lifespan of graphics cards - at least if you want to play at the AAA level is realistically 2-3 years, and if you are only optimizing drivers for future games, your card is probably close to obsolete by the time those games come out. Plus, about 95% of hardware reviews are written within the first month a product is launched, so anyone looking for a new card. The 480 had many of the same failures. It performs better than the 970 - months to years after release. However, it came out so much later than the 970 that people wanting that performance range had mostly already bought 970s, and the drivers at launch were not as good, putting performance at or below 970 levels (again - resulting in mediocre review coverage).

I think that the bottom line is, even if there are huge theoretical future gains from drivers and new games, AMD has pretty much already failed with Vega from a gaming perspective. Or, at least, they will have failed if they do not get these driver based gains in place in the next few weeks before the gaming card launches and more widespread reviews are available - not several months in the future. It's simply not good enough to release hardware and not have the driver and software infrastructure in place to make it work to its potential for several months to years in the future. They need to get their shit together and have it available at launch. Otherwise, their only hope is to drastically undercut Nvidia in terms of price to performance to get some more of the mid range market.

[Fascinate]

https://videocardz.com/70777/amd-rad...11-performance






an improvement over FE Vega, its faster than a stock 1080 and about on par or slightly behind an OCed 1080

although the 1630 Mhz is probably already close to Vegas max OC, judging from FE OCing

Eh those nvidia scores are wrong:
http://www.3dmark.com/compare/3dm11/...dm11/12256169#

Stock 1080 3d mark 11 score 24,867:
http://www.guru3d.com/articles-pages...review,27.html

[Lathais]

Eh those nvidia scores are wrong:
http://www.3dmark.com/compare/3dm11/...dm11/12256169#

Stock 1080 3d mark 11 score 24,867:
http://www.guru3d.com/articles-pages...review,27.html

In the test he linked they were using a 6700K. In the test you linked they were using a 5960X. So I would not expect the scores to be the same.

[Fascinate]

In the test he linked they were using a 6700K. In the test you linked they were using a 5960X. So I would not expect the scores to be the same.

Look again....first one has a 6700k. 1080's dont score so low in 3dmark 11.

The vega score is correct from life binders link, the nvidia scores arent.

This isnt hard guys, just google gtx 1080 3d mark 11 performance, and you will see those nvidia scores are wrong.

- - - Updated - - -

Ignore the overall score, graphics score on 1080 is around ~35k. In the one life binder linked it was only 27k.