Vega GPU benchmarks emerge, revealing several GPU specifications

[jimma47]

IMO AMD doesn't have an answer to the GTX 1080Ti or Titan XP but they do have an answer to the GTX1080 and below.

Plus anyway i wouldn't be buying any AMD GPU's IMDO they run way to hot.

My 980s run hot. 79/80 degrees on load the room is like an oven. They're a lot cooler than my 570 was though, and the 260 before that. It's evolution just because a card a few generations ago was rubbish doesn't mean they all are.
Until we see Vega in the open with actual benchmarks and temps then it's all just speculation.

I'd consider AMD solely for the multi screen flexibility. Nvidia surround is a PITA unless you have exactly the same res and sync polarity on each monitor.

 

[AngryGoldfish]

Either way, what I have said stands. If Vega offers optimized cards that perform on the pair with competition for less money it is a big win. Previous Radeon cards were furnaces that just gobbled power, and they were not good value for the money, because for a little more cash you got much better product from the green team.

I don't think that's going to change dramatically with Vega; Vega 10 will still likely be behind Nvidia in temperature and power consumption. That's not a problem, though, unless they force a AIO on us again like they did with the Fury X. If they give a fully unlocked chip on a card like an ASUS Strix with reasonable temperatures at reasonable fan speeds in the same way the R9 Fury ASUS Strix had reasonable temperatures at reasonable fan speeds then that's a winner. No, it won't be as cool as the ASUS Strix 1080 and will likely draw more power, but unlike some people here I don't consider that a deal breaker. A card drawing 270W versus 240W is not that big of a deal. A card at 75°C versus 72°C is not that big of a deal. I do imagine the GTX 1080 will draw considerably less power than Vega 10 and be cooler, but it's really not that big of a deal unless you live in Japan or want to Crossfire on a 750W PSU using air cooling.

 

[NeverBackDown]

Behind in temperature? Wow..
AMD is quite happy in the mid low 70s the entire time. That's impressive. Nvidia cards boost to 80 on all there cards. Why is this an argument? It doesn't even matter anymore outside of reference coolers.
Sure the low end Nvidia stuff is cool, but then so is AMD, sure it may not be as cool, but who cares? Anything below 80 is fine and dandy.

 

[Doomslayer]

My 980s run hot. 79/80 degrees on load the room is like an oven. They're a lot cooler than my 570 was though, and the 260 before that. It's evolution just because a card a few generations ago was rubbish doesn't mean they all are.
Until we see Vega in the open with actual benchmarks and temps then it's all just speculation.

I'd consider AMD solely for the multi screen flexibility. Nvidia surround is a PITA unless you have exactly the same res and sync polarity on each monitor.

I still wouldn't own another AMD GPU even if AMD payed me :lol:

IMO some of the blame has to be put on the "Case" your using, That's why i stop using Corsair cases really poor air flow compared to other case makers.

Friends of mine in RL (yes i do have friends in RL) i know it's hard to believe lol, There both AMD fan boy's they even told me no way in the world they would ever use AMD GPU's they said they just run too hot :lol:

 

[jimma47]

I still wouldn't own another AMD GPU even if AMD payed me :lol:

IMO some of the blame has to be put on the "Case" your using, That's why i stop using Corsair cases really poor air flow compared to other case makers.

Friends of mine in RL (yes i do have friends in RL) i know it's hard to believe lol, There both AMD fan boy's they even told me no way in the world they would ever use AMD GPU's they said they just run too hot :lol:

I did choose this case for style over function but temps were about the same in my old 500R which had more room and better airflow.

I'll just buy an air conditioner

 

[NickOmega]

This is purely speculation but I think game performance will be around the GTX 1080, *maybe* slightly ahead and still behind the Ti, however I think it will be priced in between the 1070/80.
With that in mind I can see Vega having an anti-climactic release and getting an undeserved negative reception because it's not a game changer.
It probably will run hot, on reference at least because AMD's GPU coolers aren't especially effective, but I don't think anything will run as hot as EVGA's ACX cooled Nvidia 10XX series were.... :lol:

 

[AngryGoldfish]

This is purely speculation but I think game performance will be around the GTX 1080, *maybe* slightly ahead and still behind the Ti, however I think it will be priced in between the 1070/80.
With that in mind I can see Vega having an anti-climactic release and getting an undeserved negative reception because it's not a game changer.
It probably will run hot, on reference at least because AMD's GPU coolers aren't especially effective, but I don't think anything will run as hot as EVGA's ACX cooled Nvidia 10XX series were.... :lol:

There is a rather large gaping hole between the 1080 and 1080ti though. That's what I think AMD will aiming for—or at least should be. That's what the Fury X was aimed at. If it weren't for the dastardly 980ti it would have been above a 980 but below a Titan X in both price and performance and ruled that segment. Now that the 1080ti is pretty much a Titan XP, if the GTX 1080 equates to 100% at $500 and the 1080ti equates to 130% at $700, Vega 10 might come in at 115% at $550-600. That's exactly where it needs to be in my opinion.

 

[AlienALX]

There is a rather large gaping hole between the 1080 and 1080ti though.

There actually isn't.

http://www.gamersnexus.net/hwreviews/2830-nvidia-gtx-1080-ti-fe-review-and-game-benchmarks/page-6

As for the “Founders Edition” card specifically, it’s still a reference model. Its name sounds fancier and its design looks better, but the thing doesn’t cool well for the GPU core. FET cooling is impressive and competitive, to be fair, but the core is where we encounter limitations on clock-rate as imposed by thermals. This is part of why we couldn’t replicate nVidia’s boasted 2038MHz boost with <80C thermals – you’d need a screaming loud fan to do that, and we don’t consider that usable. That’s potentially misleading marketing and isn’t new to a Pascal launch, with the GTX 1080 being similarly demonstrated during last year’s Epic Games demo. We also had trouble to replicate those findings under usable conditions.

OK so when we take that information and look around? I've seen people on OCUK saying that they upgraded from a 1080 that overclocked very well and at 1440p are only geting 10-15% more out of the TI.

So as I said all along (well, it was pretty obvious really tbh) the 1080ti needs water of some sort. If you want to extract the real 20-30% you need water. And an AIO will void your warranty (as will real water cooling) and you can expect to add at least another £100 to that in the best scenario (an AIO unit or a block for your existing loop).

So that £700 is now £800. That is nearly double what the 1080 costs and it's only 25-30% faster at best at 4k only.

So that's not really a gaping hole. I mean it is in price terms.. Some one (I think it was Paul of Paul's hardware) recently pointed out that bang for buck the 1080 is the far better deal per FPS and it was pretty pointless going from a 1080 to a TI any way, because it's just wasted power.

If you were coming from a 980ti it was far better, but tbh? if people have not upgraded from their 980ti to a 1080 there would be good reason for that. From what I have seen? all of the people buying 1080tis are Nvidia fans, and every one I have seen apart from one on here (that dude who bought it for his missus) has gone from a 1080 to the TI.

Which is just proof beyond doubt that they will buy whatever it is Nvidia are selling. The higher you make the price the less people you will have who will actually pay it. So Nvidia are doing it the short and quick way. Charge top dollar, sell up, move on. Instead of doing what AMD are with the 400 series and selling more for smaller profits.

I have a Titan X and tbh? I've not had a single twitch toward buying a 1080ti. Not a single one. I'm just not paying those prices.

 

[AngryGoldfish]

There actually isn't.

http://www.gamersnexus.net/hwreviews/2830-nvidia-gtx-1080-ti-fe-review-and-game-benchmarks/page-6

As for the “Founders Edition” card specifically, it’s still a reference model. Its name sounds fancier and its design looks better, but the thing doesn’t cool well for the GPU core. FET cooling is impressive and competitive, to be fair, but the core is where we encounter limitations on clock-rate as imposed by thermals. This is part of why we couldn’t replicate nVidia’s boasted 2038MHz boost with <80C thermals – you’d need a screaming loud fan to do that, and we don’t consider that usable. That’s potentially misleading marketing and isn’t new to a Pascal launch, with the GTX 1080 being similarly demonstrated during last year’s Epic Games demo. We also had trouble to replicate those findings under usable conditions.

OK so when we take that information and look around? I've seen people on OCUK saying that they upgraded from a 1080 that overclocked very well and at 1440p are only geting 10-15% more out of the TI.

So as I said all along (well, it was pretty obvious really tbh) the 1080ti needs water of some sort. If you want to extract the real 20-30% you need water. And an AIO will void your warranty (as will real water cooling) and you can expect to add at least another £100 to that in the best scenario (an AIO unit or a block for your existing loop).

So that £700 is now £800. That is nearly double what the 1080 costs and it's only 25-30% faster at best at 4k only.

So that's not really a gaping hole. I mean it is in price terms.. Some one (I think it was Paul of Paul's hardware) recently pointed out that bang for buck the 1080 is the far better deal per FPS and it was pretty pointless going from a 1080 to a TI any way, because it's just wasted power.

If you were coming from a 980ti it was far better, but tbh? if people have not upgraded from their 980ti to a 1080 there would be good reason for that. From what I have seen? all of the people buying 1080tis are Nvidia fans, and every one I have seen apart from one on here (that dude who bought it for his missus) has gone from a 1080 to the TI.

Which is just proof beyond doubt that they will buy whatever it is Nvidia are selling. The higher you make the price the less people you will have who will actually pay it. So Nvidia are doing it the short and quick way. Charge top dollar, sell up, move on. Instead of doing what AMD are with the 400 series and selling more for smaller profits.

I have a Titan X and tbh? I've not had a single twitch toward buying a 1080ti. Not a single one. I'm just not paying those prices.

I just wrote out a huge explanation of my theory, but I pushed the wrong button and I lost it all. The frustrating thing is I don't know what button it was so I can't unbind it.

Basically, the 1080ti is 20-25% faster than the 1080. Benchmarks show this. That's a gaping hole in the performance market that Vega can snuggle into. In fact it's marginally bigger than the difference between the 980ti and the 1080 at 21% versus 22% (according to TechPowerUp). Pricing is not as relevant. In fact I wasn't even initially going to mention price, but it is relevant so I added it.

Regarding water cooling, the 1080ti will not likely benefit from water that much (not compared to the 980ti) as I imagine it will still be locked to 2.1Ghz like the 1080 was. Considering FE 1080ti's can reach the low-mid 2000Mhz, I don't see how adding a water block is going to give you that much of a performance boost. Pascal doesn't scale like Maxwell. It seemingly hits its ceiling at around 2.1Gz. Pro benchers discussed this when Pascal was initially released.

 

[AlienALX]

I tell you I've never been so anxious about a release.

If it's 1070-11/and 1080-10 (why is the faster card the one with the lower number?!) then that's cool, because AMD currently don't have anything that fast or with that much VRAM. But HBM is worrying me really really badly.

Firstly AMD will never "get" what Nvidia do for a graphics card. I'm under no illusions here, it's never going to happen no matter how good their card is.

Secondly HBM automatically means "Very expensive". So they are kinda hobbling themselves before they even step out of the gate.

It's kinda like Fury X, right. AMD release a card slower than Nvidia's with less VRAM yet charge the same amount for it. I sat and watched stock levels on OCUK and seriously, nobody bought them. And even the ones who wanted to could not get them at first so just bought a 980ti. And why wouldn't you? you know? being completely honest and all. Why would you pay 980ti money for a card with an anchor attached that can't be easily inserted/removed that's slower than the Nvidia card and has less VRAM?

It worries me when AMD try and think about the future too much. They've been doing it for years and there are only so many resounding failures any company can take.

Async on the 7970 made it far slower than it could have been.
290 hot AF. See also - 290x.
Fury X? the less said the better really. Yeah, I know I own one but my mate was so displeased with it that he sold it to me for £420 four days after launch so I couldn't really say no...

Polaris.. A step in the right direction. But Vega could be another four steps backwards. And that worries me, especially with all of the problems and bad press Ryzen is getting. Today I have seen four people who have bought a Kaby I5 or I7, and that simply should not be happening when they could have got a CPU that performs like a 5960x or better for the same money.

But alas, it is happening, and all because AMD have somewhat messed up yet another launch.

I dunno, I love AMD but god give me strength !

 

[AngryGoldfish]

I tell you I've never been so anxious about a release.

If it's 1070-11/and 1080-10 (why is the faster card the one with the lower number?!) then that's cool, because AMD currently don't have anything that fast or with that much VRAM. But HBM is worrying me really really badly.

Firstly AMD will never "get" what Nvidia do for a graphics card. I'm under no illusions here, it's never going to happen no matter how good their card is.

Secondly HBM automatically means "Very expensive". So they are kinda hobbling themselves before they even step out of the gate.

It's kinda like Fury X, right. AMD release a card slower than Nvidia's with less VRAM yet charge the same amount for it. I sat and watched stock levels on OCUK and seriously, nobody bought them. And even the ones who wanted to could not get them at first so just bought a 980ti. And why wouldn't you? you know? being completely honest and all. Why would you pay 980ti money for a card with an anchor attached that can't be easily inserted/removed that's slower than the Nvidia card and has less VRAM?

It worries me when AMD try and think about the future too much. They've been doing it for years and there are only so many resounding failures any company can take.

Async on the 7970 made it far slower than it could have been.
290 hot AF. See also - 290x.
Fury X? the less said the better really. Yeah, I know I own one but my mate was so displeased with it that he sold it to me for £420 four days after launch so I couldn't really say no...

Polaris.. A step in the right direction. But Vega could be another four steps backwards. And that worries me, especially with all of the problems and bad press Ryzen is getting. Today I have seen four people who have bought a Kaby I5 or I7, and that simply should not be happening when they could have got a CPU that performs like a 5960x or better for the same money.

But alas, it is happening, and all because AMD have somewhat messed up yet another launch.

I dunno, I love AMD but god give me strength !

All it has to be is less expensive than the Fury X. Even $50 less is good enough as long as the performance is proportionate. HBM2 is supposed to be cheaper than HBM1 and hopefully they won't force the price up by using another annoying AIO. If they can get it to $550 for the reference design and $600 for a flagship AIB partner card like the ASUS Strix then I think they could be successful. It'll be considerably cheaper than the $750-800 for the same 1080ti GPU whilst only being 15% slower and growing in maturity with every driver update and every new DX12/Vulkan title. AMD know that the Fury range let everyone down. I think they will do everything they can to prevent that from happening again. They are arguably relying on Vega more than they were relying on Fiji. Vega is an entirely new architecture that will need to stand for the next two years in a wide range of sectors. Fiji was just one GPU for the high-end market that had already been ruled by their competitor for months and even years.

Despite the resounding failure of Fury, their stocks are always increasing. Even amateur analysts are predicting huge things for AMD. They are promoting this 'Master Plan' that involves steady progress to the top citing Lisa Su's intense vision and resolution to remove debt and regain global market share in all relevant sectors. I think they can do this in part with Vega. Ryzen wasn't a resounding success, but it was a massive step in the right direction and shows huge promise. Polaris was better than Fiji; I think Vega will be too.