I'm thinking about upgrading, but it seems the performance gained to money spend ratio might not be worth shelling out for a 1080.
I am contemplating getting an HTC Vive, too bad they didn't really test VR devices with the card. But I believe NVIDIA mentioned that the 10xx cards are all highly optimised for VR, so that might pull me over the edge and get a 1080.
I think a lot of people are in this camp right now. I think when they release of Vive-optimized drivers and related tests, then we'll see the real results.
Unsurprisingly so, yet the price point is way too high to go for an upgrade. I feel like upgrading when the 11XX (they are coming, right?) series come in is a much better idea.
AMD Polaris comes out tomorrow, which should meet VR minimum specs and will be a lot cheaper.
If it doesn't meet VR minimum specs then AMD has really missed a golden opportunity that nVidia handed them. The prestige is in the high end cards, the money is in the mainstream & mobile ones.
VR's steep requirements plus the really nice performance bump we're getting from finally getting off of 28nm means that this is probably going to be one of the most vigorous upgrade cycles for GPUs that we've seen in a while.
I'm not going to be able to warm up to AMD until they get their OpenGL drivers' performance up to par with NVIDIA. It's such a big deal breaker for me, as someone who actively uses OpenGL for both applications and development.
AMD is never going to get their OGL driver up to par with Nvidia's. That's why they created Mantle and are enthusiastic about Vulkan/DX12. It gets them off the hook to make and maintain those drivers.
Whoever decided to save a buck back at ATI and not hire a massive driver development team a couple decades ago did irreparable harm.
Interesting that they're starting with a mainstream part. I hope this becomes a trend since the mainstream part is the only one I'm ever really interested in every generation (as my gaming needs are rather modest, and currently served adequately by a GTX 750 TI).
I was planning on waiting for a mainstream version of the 1070/1080 from NVIDIA to use in a Thunderbolt graphics dock with a Thunderbolt 3 compatible ultrabook, which is a setup that, I assume, will be far more limited by the ultra low voltage CPU in the ultrabook than any mainstream graphics card from this next generation anyways. But it looks like Polaris might fit the bill and release sooner.
Have you had any experience using a graphics dock? I was talking about it the other day and while it seems like a nice idea I've yet to encounter anyone who's actually used one and can speak to how good compatibility is.
My Surface Book is effectively that (the better GPU is in the detachable half) and compatibility is perfect (once had an issue where a game used the wrong card, detached and reattached and it was all fine, drivers have updated since then) - admittedly that's a system where everything is first-party and designed to work that way.
Up to this point graphics docks have either been hacked together or proprietary one-offs, so they're not widely used. Standardized TB3 graphics docks are coming very soon though.
Nvidia has done a bunch of product release cycles where this was not the case lately. Kinda like Intel's depressingly bad "4-7%" single core performance increases, maybe a little bit better.
This is one I might finally want to consider upgrading to, from my GTX 670...
I wouldn't take PassMark too seriously. It's a dated synthetic benchmark that isn't controlled for other system components or driver versions, since the results are coming from many different users at different times.
I've got out of the game of upgrading PCs years ago, so have no clue.
To me I'd think the average user with his drivers and different components might get a better impression what to expect if he buys the card from the first link, than your link? (which would more express what to could be expected at maximum?)
Not really. The Passmark is useful to get a broad comparison, but real benchmarks will be closer to what users see in practice. Some results show that clearly. E.g. the 780 Ti is in the passmark before the GTX 970, but in games with modern drivers the GTX 970 is better. But well, they are close in both. More severe is the GTX 960 – the passmark sees that card above a R9 380X. That's completely unrealistic, the GTX 960 is slower than a R9 380, without the X.
Most user run games and not artificial benchmarks, thus the performance measured in games is closer to what they will get in practice.
I would love to buy a couple 1080s for my new rig. Too bad it seems impossible to get any. The only ones for sale are from scalpers. Any tips on how to actually acquire them?
Still, the reference version is more expensive and clocked lower than the vendor versions will be. Also, at least the GTX 1080 is limited by the temperature it reaches with the reference cooler. SLI is really the one usage where the reference version is well suited, but otherwise it is good to wait.
I was going to go with 2 EVGA GTX 1080s [0] in SLI with 4 4k monitors [1]. I currently have no video cards at all, and will be installing using on-board video. The computer will be used for both development (therefore 4 monitors) and gaming (powerful video cards), and without any video cards, I can't really use it for either. I want to get pascal chips with GDDR5X, but can't wait months for the next wave of cards to be released. What would you recommend?
When going SLI the lower clock of the reference card won't matter, they'd be fast enough – to put it mildly – for every game anyway. And the reference cooler might really be a good thing in such a setup. On the other hand, there will be aftermarket cards with coolers like this as well – I think I even saw a MSI card which combined such a cooler with a second liquid cooler? Yes, the Sea Hawk: https://msi.com/Graphics-card/GeForce-GTX-1080-SEA-HAWK.html... – of course idle speculation without reviews, but something like this could be cool for you.
I don't think the aftermarket cards are month away. A german vendor was listing some versions yesterday as 3-10 days away. I think that could be accurate, the listings on the vendors sites are all ready.
I would recommend to do it stepwise. One card with two displays, look how it performs, then add one or two displays more and maybe the second 1080, if needed – I don't know which game one would play on more than one 4K displays, and to play on one 4K display one 1080 would be fine.
Don't think I have to explain here that this is still a pretty crazy setup, do I? ;)
The 4 monitors aren't for gaming, they're for keeping code, multiple ssh terminals, file managers, query results, browsers for research and documentation, and all that other stuff up at the same time while working. It's gotten to a point where 2 monitors is simply not enough when there are so many things I want to view side-by-side.
I also figure it'll be a good chance to learn how to offload my data processing to the GPUs with CUDA, and then I'll be able to use as much power as the cards can possibly give, even if it's overkill for any particular game.
I saw the listings for nonreference cards up, but couldn't find indications on availability. Thanks for the advice!
This is impressive for the money. AMD is in a real fix, the 1070 is the new 970 and the 1080 is the 980Ti all over again. These 1070s are going to fly off shelves. I just don't see people getting excited over the Polaris chips as no one cares about performance they already had. People really want the 1080 but 1070 offers the bang for buck yet still top-tier performance.
That said, being as I work all day at a computer and spend my freetime at one, I'm looking to change up my lifestyle. I probably won't be building more desktop gaming machines at this point. I'm looking at Intel NUCs for desktop machines or a gaming laptop. I'm interested in VR gaming using a laptop to go tether-free with a backpack harness.
I'm looking forward to seeing the 1080M. A laptop with that would be a great match for cord free SteamVR.
28 comments
[ 3.2 ms ] story [ 74.8 ms ] threadI am contemplating getting an HTC Vive, too bad they didn't really test VR devices with the card. But I believe NVIDIA mentioned that the 10xx cards are all highly optimised for VR, so that might pull me over the edge and get a 1080.
If it doesn't meet VR minimum specs then AMD has really missed a golden opportunity that nVidia handed them. The prestige is in the high end cards, the money is in the mainstream & mobile ones.
VR's steep requirements plus the really nice performance bump we're getting from finally getting off of 28nm means that this is probably going to be one of the most vigorous upgrade cycles for GPUs that we've seen in a while.
Whoever decided to save a buck back at ATI and not hire a massive driver development team a couple decades ago did irreparable harm.
I was planning on waiting for a mainstream version of the 1070/1080 from NVIDIA to use in a Thunderbolt graphics dock with a Thunderbolt 3 compatible ultrabook, which is a setup that, I assume, will be far more limited by the ultra low voltage CPU in the ultrabook than any mainstream graphics card from this next generation anyways. But it looks like Polaris might fit the bill and release sooner.
https://youtu.be/vGFRoWj0YOw?t=181
Compatibility wise they're supposed to work with any laptop that has Thunderbolt 3, supports the Intel eGPU UEFI extension, and is running Windows 10.
Congratulations?
This is one I might finally want to consider upgrading to, from my GTX 670...
Here's an aggregate of 15 real world benchmarks on a consistent test rig: http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080/2...
To me I'd think the average user with his drivers and different components might get a better impression what to expect if he buys the card from the first link, than your link? (which would more express what to could be expected at maximum?)
Most user run games and not artificial benchmarks, thus the performance measured in games is closer to what they will get in practice.
[0] http://www.evga.com/Products/Product.aspx?pn=08G-P4-6180-KR
[1] http://www.amazon.com/dp/B00PC9HFO8/?tag=pcpapi-20
I don't think the aftermarket cards are month away. A german vendor was listing some versions yesterday as 3-10 days away. I think that could be accurate, the listings on the vendors sites are all ready.
I would recommend to do it stepwise. One card with two displays, look how it performs, then add one or two displays more and maybe the second 1080, if needed – I don't know which game one would play on more than one 4K displays, and to play on one 4K display one 1080 would be fine.
Don't think I have to explain here that this is still a pretty crazy setup, do I? ;)
I also figure it'll be a good chance to learn how to offload my data processing to the GPUs with CUDA, and then I'll be able to use as much power as the cards can possibly give, even if it's overkill for any particular game.
I saw the listings for nonreference cards up, but couldn't find indications on availability. Thanks for the advice!
That said, being as I work all day at a computer and spend my freetime at one, I'm looking to change up my lifestyle. I probably won't be building more desktop gaming machines at this point. I'm looking at Intel NUCs for desktop machines or a gaming laptop. I'm interested in VR gaming using a laptop to go tether-free with a backpack harness.
I'm looking forward to seeing the 1080M. A laptop with that would be a great match for cord free SteamVR.