Support NeoGAF

[DieH@rd]

Posted? Non-FE [Gaming] Vega results pulled from 3DMark 11 online database
http://wccftech.com/amd-rx-vega-ben...nfirmed-1630mhz-8gb-hbm2-484gbs-of-bandwidth/

Today's fresh testing scores:

 

[RenegadeXV8]

More charts at source -
Vega FE vs. Fury X Clock-for-Clock: "IPC" Tested

This makes things even more confusing. Clock for clock, practically 0% improvement in gaming but compute results are 1.5-2x better across the board.

 

[BobbeMalle]

Posted? Non-FE [Gaming] Vega results pulled from 3DMark 11 online database

Can't really say if i'm happy or not about this, it really depends on the price tag.

 

[opticalmace]

Can't really say if i'm happy or not about this, it really depends on the price tag.

Yeah, pretty much.

 

[Hesemonni]

With the rumoured HBM2 shortages it doesn't really matter what these go for, you ain't going to find them anywhere anyway.

 

[dgrdsv]

AMD Radeon RX Vega 3DMark11 Performance

Well, 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.

Wouldn't that imply a decrease in IPC? Although the performance we're seeing is probably not representative of the Gaming variant of Vega. So it's probably too early to come to that conclusion.

Not really a decrease but more of a lack of notable increase. This isn't really surprising considering that their focus with Vega was on pushing for higher clocks and in chip design this is usually something which is an opposite of an IPC increase.

This makes things even more confusing. Clock for clock, practically 0% improvement in gaming but compute results are 1.5-2x better across the board.

It's not compute (SPECviewperf is professional rendering) and it's not confusing at all considering that Vega FE drivers are a symbiosis of Pro and Gaming drivers while Fury X's drivers are just pure gaming.

 

[opticalmace]

I'm a bit out of touch, what are regular selling prices for the 1070 and 1080? I looked at pc part picker but it didn't seem right (1070 almost as much as a 1080?).

I'm guessing this means RX Vega has to be what, at most $500 to be competitive?

My last AMD card was a 6950, which you could flash to unlock extra shaders to basically make it a 6970. That was a great card.

 

[Delusibeta]

I'm a bit out of touch, what are regular selling prices for the 1070 and 1080? I looked at pc part picker but it didn't seem right (1070 almost as much as a 1080?).

I'm guessing this means RX Vega has to be what, at most $500 to be competitive?

My last AMD card was a 6950, which you could flash to unlock extra shaders to basically make it a 6970. That was a great card.

The RRP of the Founder's Edition 1070 is currently $400 and the Founder's Edition 1080 is $550. However, due to the ongoing cryptocurrency bubble, 1070s are currently in very high demand, hence the high prices. Additionally, while 1080s have not been raided by the bitcoiners, the fact that it's basically the only graphics card left in stock that's stronger than a 1050Ti means that it too has seen a minor price increase. (Pity AMD in this regard: supplies of the 480 has been wiped out, and 580s and 570s have been evaporating from shelves, almost all of which have been taken for mining).

But yeah, if what we've seen from Vega Frontier Edition is representative of the gaming performance of the RX Vega, then $500 is the absolute price ceiling AMD can get away with, and ideally it probably should be no more than $450.

 

[x3sphere]

I'm a bit out of touch, what are regular selling prices for the 1070 and 1080? I looked at pc part picker but it didn't seem right (1070 almost as much as a 1080?).

I'm guessing this means RX Vega has to be what, at most $500 to be competitive?

My last AMD card was a 6950, which you could flash to unlock extra shaders to basically make it a 6970. That was a great card.

Hopefully no more than $500 but honestly going by the 3Dmark results posted above I think they could get away with $550. If initial stock is going to be low it's going to sell out regardless of whether $500 or $550.

Pricing at 500-550 would make it easy for Nvidia to retaliate with a price drop too, but I still think the chances are slim of that happening, partly due to the crypto boom

 

[thelastword]

This last week had me shaking my head, so many people taking frontier edition benchmarks and running with it, yet the thing was never poised as a gaming card.....

As Raja confirmed, the gaming cards will be faster..., yet, this isn't even the top end Vega card we're seeing in this benchmark.

Stock gaming RX Vega (8GB) beats GTX 1080, we still have to see the 16GB variation and see if they push the clocks higher on that one or if they decide to go with water-cooling on one of the SKU's....One thing is sure, 1630 on the core is quite impressive and we can expect Zotac, XFX, MSI to push these much further with their cooling solutions.....

I expect the 8GB Vega to take over the 1070+1080 market, a vega card which outperforms the GTX 1080 at a GTX 1070 price-point is already a win. If Vega's 16GB card can increase the clocks on core and memory and offer a better cooling solution over Vega 8GB, it will easily outdo 1080ti, especially in real world benchmarks..

 

[MadJackChurchill]

This last week had me shaking my head, so many people taking frontier edition benchmarks and running with it, yet the thing was never poised as a gaming card.....

As Raja confirmed, the gaming cards will be faster..., yet, this isn't even the top end Vega card we're seeing in this benchmark.

Stock gaming RX Vega (8GB) beats GTX 1080, we still have to see the 16GB variation and see if they push the clocks higher on that one or if they decide to go with water-cooling on one of the SKU's....One thing is sure, 1630 on the core is quite impressive and we can expect Zotac, XFX, MSI to push these much further with their cooling solutions.....

I expect the 8GB Vega to take over the 1070+1080 market, a vega card which outperforms the GTX 1080 at a GTX 1070 price-point is already a win. If Vega's 16GB card can increase the clocks on core and memory and offer a better cooling solution over Vega 8GB, it will easily outdo 1080ti, especially in real world benchmarks..

I don't see people who have used Pascal for over a year buying around the same performance again for an additional outlay. They'd just stick with where they are. This will help the AMD base, but that's diminishing daily.

 

[JaseC]

I expect the 8GB Vega to take over the 1070+1080 market, a vega card which outperforms the GTX 1080 at a GTX 1070 price-point is already a win. If Vega's 16GB card can increase the clocks on core and memory and offer a better cooling solution over Vega 8GB, it will easily outdo 1080ti, especially in real world benchmarks..

You're putting a lot of stake in a PR comment. Vega is too power-hungry as-is; a 1080 Ti equivalent isn't happening this gen, which will put AMD in a very tough spot if the rumours of Volta being moved forward to Q3 happen to pan out. All evidence points to Vega hovering around the 1080 and AMD having a winner on its hands will depend entirely on how aggressive it can be with pricing -- and even then the window of opportunity may be short.

 

[x3sphere]

This last week had me shaking my head, so many people taking frontier edition benchmarks and running with it, yet the thing was never poised as a gaming card.....

As Raja confirmed, the gaming cards will be faster..., yet, this isn't even the top end Vega card we're seeing in this benchmark.

Stock gaming RX Vega (8GB) beats GTX 1080, we still have to see the 16GB variation and see if they push the clocks higher on that one or if they decide to go with water-cooling on one of the SKU's....One thing is sure, 1630 on the core is quite impressive and we can expect Zotac, XFX, MSI to push these much further with their cooling solutions.....

I expect the 8GB Vega to take over the 1070+1080 market, a vega card which outperforms the GTX 1080 at a GTX 1070 price-point is already a win. If Vega's 16GB card can increase the clocks on core and memory and offer a better cooling solution over Vega 8GB, it will easily outdo 1080ti, especially in real world benchmarks..

are they even releasing a 16GB consumer card? Haven't seen any information on that. And that's a big if as far as increasing the clocks, looking at these 3dmark results there's still a large gap from the 1080 Ti. I do not see Vega matching the 1080 Ti unless it overclocks as well as Maxwell or the HD7970 did. Fury did not have a ton of overclocking headroom so my expectations in that regard aren't very high, but we will see.

 

[thelastword]

I don't see people who have used Pascal for over a year buying around the same performance again for an additional outlay. They'd just stick with where they are. This will help the AMD base, but that's diminishing daily.

I don't think that's the target market, but rather persons who were interested in GTX 1070/1080 or even Ti, but were waiting to see what AMD would bring on a perf to price ratio. There are already tweets out there that the price to perf ratio on vega will be incredibly good.....

You're putting a lot of stake in a PR comment. Vega is too power-hungry as-is; a 1080 Ti equivalent isn't happening this gen, which will put AMD in a very tough spot if the rumours of Volta being moved forward to Q3 happen to pan out. All evidence points to Vega hovering around the 1080 and AMD having a winner on its hands will depend entirely on how aggressive it can be with pricing -- and even then the window of opportunity may be short.

What's the big deal with 250W TDP against 300W with a larger chip and more next gen features.....?

are they even releasing a 16GB consumer card? Haven't seen any information on that. And that's a big if as far as increasing the clocks, looking at these 3dmark results there's still a large gap from the 1080 Ti. I do not see Vega matching the 1080 Ti unless it overclocks as well as Maxwell or the HD7970 did. Fury did not have a ton of overclocking headroom so my expectations in that regard aren't very high, but we will see.

Pro Vega has 16GB HBM2, Frontier Edition has 16GB of HBM2, with 1080Ti at 11GB, there's no way AMD only offers a 8GB gaming card on the high end, not with all this talk of 4k textures et al...

 

[Unknown Soldier]

This makes things even more confusing. Clock for clock, practically 0% improvement in gaming but compute results are 1.5-2x better across the board.

They have focused heavily on compute because they think they'll get a piece of that AI deep learning self driving car pie. Which they won't. They have literally zero software competence and that's what separates Nvidia and Google from the likes of Intel, AMD, and Qualcomm.

 

[x3sphere]

Pro Vega has 16GB HBM2, Frontier Edition has 16GB of HBM2, with 1080Ti at 11GB, there's no way AMD only offers a 8GB gaming card on the high end, not with all this talk of 4k textures et al...

I suppose if there's demand, then they will do it. But I'd probably just go with the 8GB as by the time VRAM requirements have gone up significantly there will be better cards. Don't even know of anything out there right now where 8GB is a limitation.

 

[Neizel]

I don't see people who have used Pascal for over a year buying around the same performance again for an additional outlay. They'd just stick with where they are. This will help the AMD base, but that's diminishing daily.

The common sense indicates that no one with a card that performs the same as the one you have already should buy the new one.

This Vega cards are for those with GTX 970 and 290X/390X levels of performance that have been waiting for an upgrade.

 

[StrongBlackVine]

So would this be considered a true 4K GPU? The 1080 isn't and this only about 10-15 percent faster.

 

[MadJackChurchill]

The common sense indicates that no one with a card that performs the same as the one you have already should buy the new one.

This Vega cards are for those with GTX 970 and 290X/390X levels of performance that have been waiting for an upgrade.

True. They've waited for a year longer than they should have, though. Hope the market's still there.

 

[Neizel]

True. They've waited for a year longer than they should have, though. Hope the market's still there.

That's right,and I'm included in that waiting club (290X here) so I'm very interested about what AMD will release 😃

 

[JaseC]

What's the big deal with 250W TDP against 300W with a larger chip and more next gen features.....?

300W is quite literally the limit of a PCI-E 3.0 slot + one eight-pin connector + one six-pin connector, and even the FEs throttle. AMD could release the consumer flavour with two eight-pin connectors, but, again, heat is an issue as-is. Efficiency simply isn't Vega's strongsuit.

For comparison, the 1080 is 180W and 1080 Ti 250W.

 

[dgrdsv]

This last week had me shaking my head, so many people taking frontier edition benchmarks and running with it, yet the thing was never poised as a gaming card.....

As Raja confirmed, the gaming cards will be faster..., yet, this isn't even the top end Vega card we're seeing in this benchmark.

Stock gaming RX Vega (8GB) beats GTX 1080, we still have to see the 16GB variation and see if they push the clocks higher on that one or if they decide to go with water-cooling on one of the SKU's....One thing is sure, 1630 on the core is quite impressive and we can expect Zotac, XFX, MSI to push these much further with their cooling solutions.....

I expect the 8GB Vega to take over the 1070+1080 market, a vega card which outperforms the GTX 1080 at a GTX 1070 price-point is already a win. If Vega's 16GB card can increase the clocks on core and memory and offer a better cooling solution over Vega 8GB, it will easily outdo 1080ti, especially in real world benchmarks..

No, it doesn't, not according to whatever benchmarks we have for now.

A Vega card which outperforms 1080 by some 10% while consuming 75% more power won't do anything to 1080's market and so far we don't know anything on Vega based alternative to 1070. It's also absolutely clear that Vega 10 will never be able to outdo 1080Ti, especially in those real world benchmarks.

 

[icecold1983]

This last week had me shaking my head, so many people taking frontier edition benchmarks and running with it, yet the thing was never poised as a gaming card.....

As Raja confirmed, the gaming cards will be faster..., yet, this isn't even the top end Vega card we're seeing in this benchmark.

Stock gaming RX Vega (8GB) beats GTX 1080, we still have to see the 16GB variation and see if they push the clocks higher on that one or if they decide to go with water-cooling on one of the SKU's....One thing is sure, 1630 on the core is quite impressive and we can expect Zotac, XFX, MSI to push these much further with their cooling solutions.....

I expect the 8GB Vega to take over the 1070+1080 market, a vega card which outperforms the GTX 1080 at a GTX 1070 price-point is already a win. If Vega's 16GB card can increase the clocks on core and memory and offer a better cooling solution over Vega 8GB, it will easily outdo 1080ti, especially in real world benchmarks..

No to almost all of this

 

[GameAddict411]

This last week had me shaking my head, so many people taking frontier edition benchmarks and running with it, yet the thing was never poised as a gaming card.....

As Raja confirmed, the gaming cards will be faster..., yet, this isn't even the top end Vega card we're seeing in this benchmark.

Stock gaming RX Vega (8GB) beats GTX 1080, we still have to see the 16GB variation and see if they push the clocks higher on that one or if they decide to go with water-cooling on one of the SKU's....One thing is sure, 1630 on the core is quite impressive and we can expect Zotac, XFX, MSI to push these much further with their cooling solutions.....

I expect the 8GB Vega to take over the 1070+1080 market, a vega card which outperforms the GTX 1080 at a GTX 1070 price-point is already a win. If Vega's 16GB card can increase the clocks on core and memory and offer a better cooling solution over Vega 8GB, it will easily outdo 1080ti, especially in real world benchmarks..

I hope you are not serious, because otherwise you are just delusional. We already have the 16 GB variant and that's the Founder edition. I doubt increasing the VRAM would suddenly increase performance either. The power consumption on the card is very high on stock clocks already. I doubt the overclocking potential of the card is that high even with water cooling. The GTX 1080 runs cooler, consumes significantly less power, and has more overclocking potential. Considering that Nvidia is already preparing their next gen cards, they won't have a chance in a few months from now.

 

[Firenze1]

So would this be considered a true 4K GPU? The 1080 isn't and this only about 10-15 percent faster.

Raja said at ces that 60 fps 4k isnt there yet

 

[Unknown Soldier]

I hope you are not serious, because otherwise you are just delusional. We already have the 16 GB variant and that's the Founder edition. I doubt increasing the VRAM would suddenly increase performance either. The power consumption on the card is very high on stock clocks already. I doubt the overclocking potential of the card is that high even with water cooling. The GTX 1080 runs cooler, consumes significantly less power, and has more overclocking potential. Considering that Nvidia is already preparing their next gen cards, they won't have a chance in a few months from now.

The story goes that Nvidia has had the 11ghz VRAM 1080's ready to go for months now. They'll pull those things out when Vega is released to crush Vega before it even has a chance to live. If you're looking to sell a 1080 for whatever reason like upgrading to 1080 Ti you should do so ASAP because regular 1080's will be hitting fire sale status when the 11ghz VRAM 1080's hit the streets.

 

[icecold1983]

The story goes that Nvidia has had the 11ghz VRAM 1080's ready to go for months now. They'll pull those things out when Vega is released to crush Vega before it even has a chance to live. If you're looking to sell a 1080 for whatever reason like upgrading to 1080 Ti you should do so ASAP because regular 1080's will be hitting fire sale status when the 11ghz VRAM 1080's hit the streets.

They were released a while back. Only a few% faster. Nvidia gpus are rarely bandwidth bound

 

[thelastword]

I suppose if there's demand, then they will do it. But I'd probably just go with the 8GB as by the time VRAM requirements have gone up significantly there will be better cards. Don't even know of anything out there right now where 8GB is a limitation.

8GB of HBM2 will be plenty and will be good for many persons, notwithstanding, hbm2 runs at twice the data-rate of GDDR5.

300W is quite literally the limit of a PCI-E 3.0 slot + one eight-pin connector + one six-pin connector, and even the FEs throttle. AMD could release the consumer flavour with two eight-pin connectors, but, again, heat is an issue as-is. Efficiency simply isn't Vega's strongsuit.

For comparison, the 1080 is 180W and 1080 Ti 250W.

Again, the gaming cards and pro cards are different. People said the same thing about the Polaris cards with reference to wattage, but they didn't burn down any houses and put persons into abject poverty because they could not pay their electricity bill.

Max TDP is being blown out of proportion, yet I'm enthused because of the newer architecture and real world results...coupled with a great price to perf ratio. I think all of that combined will exceed any qualms raised due to a few extra watts....

No, it doesn't, not according to whatever benchmarks we have for now.

A Vega card which outperforms 1080 by some 10% while consuming 75% more power won't do anything to 1080's market and so far we don't know anything on Vega based alternative to 1070. It's also absolutely clear that Vega 10 will never be able to outdo 1080Ti, especially in those real world benchmarks.

It is faster actually, stock Vega is faster over stock GTX 1080. Whether it uses 75% more power or not is besides the point, what will be more important is at what price range it beats GTX1080, the lower it is over the price of a GTX1080, the sweeter the deal will be for consumers.

No to almost all of this

Ok....

I hope you are not serious, because otherwise you are just delusional. We already have the 16 GB variant and that's the Founder edition. I doubt increasing the VRAM would suddenly increase performance either. The power consumption on the card is very high on stock clocks already. I doubt the overclocking potential of the card is that high even with water cooling. The GTX 1080 runs cooler, consumes significantly less power, and has more overclocking potential. Considering that Nvidia is already preparing their next gen cards, they won't have a chance in a few months from now.

Answer me this, do you believe AMD will offer only one high end sku after all this wait for Vega? So much talk about power consumption and overclocking capability, yet the gaming vega has not even launched yet. Real world gaming performance is what's important, at a great price? even more so.....

We will get to know more soon, all we have so far is one GPU entry for gaming Vega, I'm sure things will improve and I'm pretty sure that we are not aware of all Vega gaming sku's at this point.

 

[icecold1983]

8GB of HBM2 will be plenty and will be good for many persons, notwithstanding, hbm2 runs at twice the data-rate of GDDR5.

Again, the gaming cards and pro cards are different. People said the same thing about the Polaris cards with reference to wattage, but they didn't burn down any houses and put persons into abject poverty because they could not pay their electricity bill.

Max TDP is being blown out of proportion, yet I'm enthused because of the newer architecture and real world results...coupled with a great price to perf ratio. I think all of that combined will exceed any qualms raised due to a few extra watts....

It is faster actually, stock Vega is faster over stock GTX 1080. Whether it uses 75% more power or not is besides the point, what will be more important is at what price range it beats GTX1080, the lower it is over the price of a GTX1080, the sweeter the deal will be for consumers.

Ok....

Answer me this, do you believe AMD will offer only one high end sku after all this wait for Vega? So much talk about power consumption and overclocking capability, yet the gaming vega has not even launched yet. Real world gaming performance is what's important, at a great price? even more so.....

We will get to know more soon, all we have so far is one GPU entry for gaming Vega, I'm sure things will improve and I'm pretty sure that we are not aware of all Vega gaming sku's at this point.

yeah man, the gaming version of vega is going to be 50+% faster than the FE and easily outdo the 1080ti. youre smack dab on the money here

 

[Firenze1]

Vega feature set vs pascal

 

[Locuza]

From GamerNexus we have four gaming results where we can compare Vega @ Default (unknown clock rate per app) vs. Vega @ 1050 Mhz.
http://www.gamersnexus.net/guides/2977-vega-fe-vs-fury-x-at-same-clocks-ipc

I looked at the 4K results for the best scaling without (relevant) driver overhead:

Doom @ Stock is only 12% faster, than on 1050 Mhz, for linear scaling that would mean only 1176 Mhz.
Ghost Recon: Wildlans 37%, ~ 1440 Mhz.
Ashes of the Singularity 21%, 1270 Mhz
Sniper Elite 4, 4,5%, 1097 Mhz.

Pcper wrote down that on their parcour Vega is clocking between 1300 Mhz (+24% and 1500 Mhz (+43%) but in general around 1440 Mhz (+37%).

In general, the clock speeds on the Vega Frontier Edition appear to hover around the 1440 MHz mark, well under the 1600 MHz rated clock speed of the card from specifications. The jumps hit inside the 1300 and 1500 MHz marks though the granularity of the shifts are quite large, indicating that maybe the boost/turbo capability of the Vega architecture remains more limited than what we see on the GeForce side of things.

The Radeon Vega Frontier Edition 16GB Air Cooled Review | Clock Speeds and Power Consumption

So it seems that at least on two games Vega is extremely bandwidth starved because under 1200 Mhz seems unlikely for Vega.
Or something else is very off.

 

[JaseC]

Again, the gaming cards and pro cards are different. People said the same thing about the Polaris cards with reference to wattage, but they didn't burn down any houses and put persons into abject poverty because they could not pay their electricity bill.

Max TDP is being blown out of proportion, yet I'm enthused because of the newer architecture and real world results...coupled with a great price to perf ratio. I think all of that combined will exceed any qualms raised due to a few extra watts....

Polaris debuted with the 4xx series, which tops out at the 480, which itself is equivalent to the 1060 and merely 150W.

If you'd rather place your faith in AMD's claim and disregard the FE, then that's your prerogative. There may be very well be a 16GB consumer SKU, but I think you're setting yourself up for disappointment by expecting it to go toe-to-tie with the 1080 Ti. All real-world data points to an aggressively-priced answer to the 1080 being the best case scenario this gen.

 

[thelastword]

yeah man, the gaming version of vega is going to be 50+% faster than the FE and easily outdo the 1080ti. youre smack dab on the money here

With higher clocks than 1630 and better cooling solutions I don't see why not...In any case, the takeaway here is that 8GB Vega is better at stock over a 1080 in 3Dmark...What will be most interesting is real world results though.....With all the new features that AMD has implemented to save bandwidth and some of it's new arch strengths, real world results will be quite interesting come July 31st...

Polaris debuted with the 4xx series, which tops out at the 480, which itself is equivalent to the 1060 and merely 150W.

If you'd rather place your faith in AMD's claim and disregard the FE, then that's your prerogative. There may be very well be a 16GB consumer SKU, but I think you're setting yourself up for disappointment by expecting it to go toe-to-tie with the 1080 Ti. All real-world data points to an aggressively-priced answer to the 1080 being the best case scenario this gen.

The 16GB going toe to toe with the 1080ti is pretty much what I can deduce from all the information I've seen....It has already leaked that there are 3 high end sku's for Vega; 687F:C1, 687F:C2 and 687F:C3, so having only 1 8GB vega does not make sense. I do think it's necessary for AMD to target the Ti/XP and having higher clocks with a better cooling solution on one of the cards will achieve that. Perhaps the arch improvements of Vega will be enough to give comparable real world results to the 1080TI, but we shall see how they go about it...

Looking at all the leaked info, it just makes sense that AMD target the 3 high end cards on the market.....1080Ti, 1080 and 1070.

 

[icecold1983]

With higher clocks than 1630 and better cooling solutions I don't see why not...In any case, the takeaway here is that 8GB Vega is better at stock over a 1080 in 3Dmark...What will be most interesting is real world results though.....With all the new features that AMD has implemented to save bandwidth and some of it's new arch strengths, real world results will be quite interesting come July 31st...

The 16GB going toe to toe with the 1080ti is pretty much what I can deduce from all the information I've seen....It has already leaked that there are 3 high end sku's for Vega; 687F:C1, 687F:C2 and 687F:C3, so having only 1 8GB vega does not make sense. I do think it's necessary for AMD to target the Ti/XP and having higher clocks with a better cooling solution on one of the cards will achieve that. Perhaps the arch improvements of Vega will be enough to give comparable real world results to the 1080TI, but we shall see how they go about it...

Looking at all the leaked info, it just makes sense that AMD target the 3 high end cards on the market.....1080Ti, 1080 and 1070.

What clockspeeds you you expect that it will easily outdo a 1080ti?

 

[JaseC]

The 16GB going toe to toe with the 1080ti is pretty much what I can deduce from all the information I've seen....It has already leaked that there are 3 high end sku's for Vega; 687F:C1, 687F:C2 and 687F:C3, so having only 1 8GB vega does not make sense. I do think it's necessary for AMD to target the Ti/XP and having higher clocks with a better cooling solution on one of the cards will achieve that. Perhaps the arch improvements of Vega will be enough to give comparable real world results to the 1080TI, but we shall see how they go about it....

Well, we know the C3 is an 8GB card with 64 compute units and 4096 stream processors, which strongly suggests it's either an 8GB version of the FE or the consumer equivalent. It fared much better than the 1080 in the SiSoft compute benchmark that outed its existence but still landed notably behind the 1080 Ti.

Also, I didn't mean to imply there'll be just one consumer SKU. I'm just disagreeing that AMD has the 1080 Ti in its sights.

 

[PFD]

What clockspeeds you you expect that it will easily outdo a 1080ti?

I needs to be at least 2Ghz, going by what's leaked so far

In other words, ain't gonna happen

 

[icecold1983]

I needs to be at least 2Ghz, going by what's leaked so far

In other words, ain't gonna happen

I know im just trying to see him further explain his fantasy

 

[thelastword]

What clockspeeds you you expect that it will easily outdo a 1080ti?

A much higher clockspeed may not even be necessary as Vega is already ahead of the T.I in raw flop count. in essence, they already have the hardware, so with maturity of drivers and real world results, we should see some interesting numbers come the end of July....

We should keep in mind that the final results and performance of these cards are not known yet and things would have naturally improved with updated revisions.

In any case, it should also be noted that if AMD is achieving 1630 on their in-house cards, other vendors may be able to reach much higher speeds on the core and memory.

Well, we know the C3 is an 8GB card with 64 compute units and 4096 stream processors, which strongly suggests it's either an 8GB version of the FE or the consumer equivalent. It fared much better than the 1080 in the SiSoft compute benchmark that outed its existence but still landed notably behind the 1080 Ti.

Also, I didn't mean to imply there'll be just one consumer SKU. I'm just disagreeing that AMD has the 1080 Ti in its sights.

When the Ti first came on the scene it was said that it was roughly 35% faster than a GTX 1080, but here we have an 8GB Vega C3 that 's said to be 35% faster than GTX 1080....in your article...

We should also realize that the article you linked was published since march 2017, so what does that say? At the time of the article Vega C1 was at 1200Mhz on core and 700MHz on memory, C3 was always faster than C1 which at the time was 35% ahead of GTX 1080 in Sandra....

We have now progressed to early July where leaks of a C1 has seen clocks increase to 1630 MHz on the core and 945 on the memory, yet C3 is the strongest card in the bunch, so what are the improvements to C3 since then?

The latest benches now show that the C1 is 15% stronger over GTX 1080 in 3DMark (though a different test), however, if we agree that C3 is stronger, then it should be a no-brainer that one sku is aiming for the 1080Ti....All the tests AMD has shown so far is with the C1, but at 1200MHZ core and 700MHz memory, yet, we saw that running Doom at 4k @ 66fps and battlefront at 4k @ 70fps average, so real world results on all skus will be much stronger now with the increased clocks, better drivers et al.......

I think these early tests are more positive than we think they are, even in your article, it is said that real world results of the C3 back in march had the vega matching the 1080ti in real world results in DOOM and battlefront, despite the 1080ti having a 25% lead over Vega in Sandra, but again your article also said that Sisoft Sandra read the vega as only having a 344Mhz core and only 16kb of L2 cache, so maybe those Sisoft scores are not that accurate for the unreleased hardware right now, taking into account the real world results as a comparison...

I'll say this, if gaming Vega at 1200MHz core and 700MHz memory can run DOOM and Battlefront at 4k 60fps since march, then it bodes well for real world results at 1630Mhz and 945MHz speeds or even higher, depending on the different AMD sku's or the manufacturer cards at higher speeds over 1630MHZ....Again, real world results will be most interesting come July 31st. I imagine even those 3D benchmarks will receive an uptick come the release of final hardware.....

 

[icecold1983]

A much higher clockspeed may not even be necessary as Vega is already ahead of the T.I in raw flop count. in essence, they already have the hardware, so with maturity of drivers and real world results, we should see some interesting numbers come the end of July....

We should keep in mind that the final results and performance of these cards are not known yet and things would have naturally improved with updated revisions.

In any case, it should also be noted that if AMD is achieving 1630 on their in-house cards, other vendors may be able to reach much higher speeds on the core and memory.

When the Ti first came on the scene it was said that it was roughly 35% faster than a GTX 1080, but here we have an 8GB Vega C3 that 's said to be 35% faster than GTX 1080....in your article...

We should also realize that the article you linked was published since march 2017, so what does that say? At the time of the article Vega C1 was at 1200Mhz on core and 700MHz on memory, C3 was always faster than C1 which at the time was 35% ahead of GTX 1080 in Sandra....

We have now progressed to early July where leaks of a C1 has seen clocks increase to 1630 MHz on the core and 945 on the memory, yet C3 is the strongest card in the bunch, so what are the improvements to C3 since then?

The latest benches now show that the C1 is 15% stronger over GTX 1080 in 3DMark (though a different test), however, if we agree that C3 is stronger, then it should be a no-brainer that one sku is aiming for the 1080Ti....All the tests AMD has shown so far is with the C1, but at 1200MHZ core and 700MHz memory, yet, we saw that running Doom at 4k @ 66fps and battlefront at 4k @ 70fps average, so real world results on all skus will be much stronger now with the increased clocks, better drivers et al.......

I think these early tests are more positive than we think they are, even in your article, it is said that real world results of the C3 back in march had the vega matching the 1080ti in real world results in DOOM and battlefront, despite the 1080ti having a 25% lead over Vega in Sandra, but again your article also said that Sisoft Sandra read the vega as only having a 344Mhz core and only 16kb of L2 cache, so maybe those Sisoft scores are not that accurate for the unreleased hardware right now, taking into account the real world results as a comparison...

I'll say this, if gaming Vega at 1200MHz core and 700MHz memory can run DOOM and Battlefront at 4k 60fps since march, then it bodes well for real world results at 1630Mhz and 945MHz speeds or even higher, depending on the different AMD sku's or the manufacturer cards at higher speeds over 1630MHZ....Again, real world results will be most interesting come July 31st. I imagine even those 3D benchmarks will receive an uptick come the release of final hardware.....

1630 at almost 400 watts and ~10% faster than a stock 1080

 

[JaseC]

When the Ti first came on the scene it was said that it was roughly 35% faster than a GTX 1080, but here we have an 8GB Vega C3 that 's said to be 35% faster than GTX 1080....in your article...

Again, it was a compute-specific benchmark.

We should also realize that the article you linked was published since march 2017, so what does that say?

That it was published in March 2017 and is recent enough to be relevant to the discussion.

We have now progressed to early July where leaks of a C1 has seen clocks increase to 1630 MHz on the core and 945 on the memory, yet C3 is the strongest card in the bunch, so what are the improvements to C3 since then?

The latest benches now show that the C1 is 15% stronger over GTX 1080 in 3DMark (though a different test), however, if we agree that C3 is stronger, then it should be a no-brainer that one sku is aiming for the 1080Ti....All the tests AMD has shown so far is with the C1, but at 1200MHZ core and 700MHz memory, yet, we saw that running Doom at 4k @ 66fps and battlefront at 4k @ 70fps average, so real world results on all skus will be much stronger now with the increased clocks, better drivers et al.......

The latest 3D Mark benchmarks actually put the C1 just barely ahead of the 1080 (see dr_rus' post below DieHard's). Also, we don't know how C1-C3 compare in terms of clockspeeds: SiSoft didn't detect the C3's properly and there's no 3D Mark GPU data for the C3 or C2. We do know, however, that the C1 and C3 have the same core configuration, thereby allowing us to deduce that the C2 has the same core configuration, which suggests that they're not three separate SKUs but rather different engineering samples of the one. (As before, I'm not saying there'll be just the one consumer SKU.)

I think these early tests are more positive than we think they are, even in your article, it is said that real world results of the C3 back in march had the vega matching the 1080ti in real world results in DOOM and battlefront, despite the 1080ti having a 25% lead over Vega in Sandra, but again your article also said that Sisoft Sandra read the vega as only having a 344Mhz core and only 16kb of L2 cache, so maybe those Sisoft scores are not that accurate for the unreleased hardware right now, taking into account the real world results as a comparison...

Well, "not far shy" means it still trailed behind to some degree, and I think it's fairly obvious that the SiSoft benchmark wasn't actually ran with a 344MHz core and 16KB L2 cache. That was a detection error.

 

[michaelius]

With higher clocks than 1630 and better cooling solutions I don't see why not...In any case, the takeaway here is that 8GB Vega is better at stock over a 1080 in 3Dmark...What will be most interesting is real world results though.....With all the new features that AMD has implemented to save bandwidth and some of it's new arch strengths, real world results will be quite interesting come July 31st...

Higher clock speeds? It already needs power limit increased from 300W to even hit 1600.

1630 is also probably only theoretical maximum.

 

[llien]

The latest 3DMark scores put Vega exactly where I'd expect it to be, in between 1080 and 1080Ti.

Considering the fact, that factory OCed 1080's go for 700 Euro Vega still has a chance to be competitive. (even ignoring FreeSync/FineWine)

Not really a decrease but more of a lack of notable increase. This isn't really surprising considering that their focus with Vega was on pushing for higher clocks and in chip design this is usually something which is an opposite of an IPC increase.

Actually, higher IPC was explicitly promised.

Higher IPC, Tiling, & More
http://www.anandtech.com/show/11002/the-amd-vega-gpu-architecture-teaser

Tiling is nowhere to be seen either.

 

[Lagspike_exe]

Vega has a chance, but AMD will be taking a margin hit to compete with NVidia.

I would buy it over 1080 purely due to Free Sync/Fine Wine if I was in the market for a GPU.

 

[joesiv]

AMD Radeon RX Vega 3DMark11 Performance

We have to be careful with the above chart... It was assumed by videocardz.com that there has been a great increase in performance "over time", due to optimizations to get this scale of results... but if you look at the best result and worst result you'll find that they're running the SAME driver version, the only difference is the platform it's being ran on (the low result is Ryzen 1800, the top if i7 6700)

The story of performance getting better over time is great, and I was excited about the prospects, but the chart is no indication of it.

I think we might just have to wait until closer to launch.

 

[Marmelade]

A much higher clockspeed may not even be necessary as Vega is already ahead of the T.I in raw flop count. in essence, they already have the hardware, so with maturity of drivers and real world results, we should see some interesting numbers come the end of July....

That doesn't mean much though, just look at the Fury X (8.6TF) and GTX 1080 (close to 9TF) and the gap in real world performance between the two

 

[icecold1983]

crysis 3 gameplay performance @ 1440p

 

[LelouchZero]

crysis 3 gameplay performance @ 1440p

Hehe I was about to post this.

From the description:

Crysis 3 Pc gameplay at 1440p very high settings. I'm using an AMD Vega Frontier Edition in game mode and a Kaby Lake 7700K in this fps frame rate performance benchmark. The clock speed doesn't stay at 1600 but it does if I max out the power limit but this is out of the box performance with a higher fan speed. I'll be doing a few benchmarks with this card until AMD RX Vega comes along.

In the video the clock speed is heavily fluctuating from around 1440-1600MHz, the performance is mostly around 38-45 fps at 1440p very high settings with 2X MSAA.

 

[dgrdsv]

Actually, higher IPC was explicitly promised.

Higher IPC, Tiling, & More
http://www.anandtech.com/show/11002/the-amd-vega-gpu-architecture-teaser

Tiling is nowhere to be seen either.

And this is why you shouldn't listen to such promises. AMD's PR/marketing on both Polaris and Vega has been abysmal so far. "Poor Volta" anyone?

As for tiling rasterization - the sole fact that the tool which used a byproduct of such rasterization on Paxwell to show its existence doesn't show the same result on Vega doesn't mean that it's not there and not working as AMD's implementation may not show itself under the same load profile for example. Here we'll have to wait on technical details.

We have to be careful with the above chart... It was assumed by videocardz.com that there has been a great increase in performance "over time", due to optimizations to get this scale of results... but if you look at the best result and worst result you'll find that they're running the SAME driver version, the only difference is the platform it's being ran on (the low result is Ryzen 1800, the top if i7 6700)

The story of performance getting better over time is great, and I was excited about the prospects, but the chart is no indication of it.

I think we might just have to wait until closer to launch.

The chart basically compiles all the results from 3DM database we have so far into one place, it doesn't say anything about results getting better over time.

 

[joesiv]

The chart basically compiles all the results from 3DM database we have so far into one place, it doesn't say anything about results getting better over time.

If you follow the link to the creators of the chart, they say the following about the chart:

Speaking of 3DMark11 scores, compared to some older benchmark results (~3 months old) the performance has increased by around 15% (quite impressive).

I was just warning people not to make the same mistake as the creators of the chart, and also not to take the screenshot of the chart out of context. If you follow the link, you can click to the actual 3dmark results, including driver version and system configuration. The numbered RX Vega results don't appear to be sorted by time, which some might assume (since they are numbered).

 

[LelouchZero]

Doom Vulkan AMD Vega Frontier Edition Vs GTX 1080 TI Vs GTX 1080 4K Frame Rate Comparison

Cards Used -

AMD Vega Frontier Edition Fan Maxed
Asus GTX 1080 TI Founders Edition Custom Fan Curve MSI
Asus GTX 1080 Founders Edition Custom Fan Curve MSI

The fan speed is around 90% on the Nvidia cards at over 60 degrees.

CPU Kabe Lake 7700K

AMD's ReLive software doesn't record with Vulkan so I had to use VSR and an Elgato HD60 to record. The fps is the same as if I was in native 4K as I tested before hand. Also I can not control the fan speed of the Vega card with MSI afterburner. It was a lot hotter in the day when I did the 1080 benchmark which is why the temps are higher.

The Vega Frontier has many clock speed fluctuations however appears to offer performance similar to the GTX 1080, with the GTX 1080 leading in some scenarios, and the VEGA FE in others.