|
spasticColon posted:I myself think it's going to take another revolutionary leap in GPU tech in order for 4K gaming to become feasible. Maybe something similar to the jump to the Unified shader model? All that Nvidia and AMD have been doing for the last 10+ years is building on the Unified Shader Model technology so they probably need another revolutionary leap beyond that. That's crazy. I swear I remember reading at one point people moaning that there was nothing but console ports out and nothing was really coming down the pipeline to make PC hardware sweat except for 1 or 2 exceptions. And now we're back in the swing of things where even the new hardware is having issues keeping up.
|
#
?
Jul 19, 2016 16:34
|
|
|
|
| # ? May 16, 2024 00:45 |
|
|
So the realization just hit that we have aftermarket 1060's out on the market, some even for MSRP. The RX 480 looks like it still won't see aftermarket cards for another 3 weeks or so. There hasn't even been any significant stock of reference RX 480's here in the US since launch on top of no aftermarket card. What the gently caress are AMD and its partners doing.
|
|
#
?
Jul 19, 2016 16:37
|
|
|
Seamonster posted:I don't think its as much of a leap as it is bits and pieces coming together. DX12/Vulkan have shown the potential for efficiency gains when devs decide to give a gently caress and the 14/16nm process node will only mature. It may take bigass end-stage 14/16nm chips on HBM2 chugging 300W to get there but 4K can totally be A Thing without a super revolutionary leap. 4k is four times more pixels to render over the expected 1080p target. That is a revolutionary leap, full stop, when the expected performance gains of a new GPU gen have been 50% at best for two decades.
|
|
#
?
Jul 19, 2016 16:37
|
|
|
Beautiful Ninja posted:So the realization just hit that we have aftermarket 1060's out on the market, some even for MSRP. The RX 480 looks like it still won't see aftermarket cards for another 3 weeks or so. There hasn't even been any significant stock of reference RX 480's here in the US since launch on top of no aftermarket card. I've been dead set on a 480 for a while. Just pulled the trigger on a 1060. AMD just seem determined to push themselves into irrelevancy.
|
|
#
?
Jul 19, 2016 16:40
|
|
|
BangersInMyKnickers posted:4k is four times more pixels to render over the expected 1080p target. That is a revolutionary leap, full stop, when the expected performance gains of a new GPU gen have been 50% at best for two decades. Well, if dx12/Vulkan allow for true multi GPU without stutter or other downsides, then 4k gaming could start happening via parallel gpu's.
|
|
#
?
Jul 19, 2016 16:41
|
|
|
I think people are being a bit hyperbolic about 4k. Yes, to run 4k above 100hz is a ways off. However, 4k is perfectly playable today on AAA titles. Doom maxed out at 4k is really really smooth. RotTR is also very playable at 4k with just a few things turned down a notch. Yes, both may dip below 60fps at times, but they rarely go below 50 and are goddamn loving beautiful.
|
|
#
?
Jul 19, 2016 16:44
|
|
|
VR needs to stay above 100fps at all times at like 8k resolution before it'll be halfway decent, so no, people aren't actually being hyperbolic. I mean, they are, but about the wrong thing. 
|
|
#
?
Jul 19, 2016 16:47
|
|
|
Looking at Newegg for SnG, I see a bunch of 1060's at $250 with AIB like the Asus Strix for $300 which seems like a rather ok deal for a just released card you can actually get your hands on today. It is evil of Nvidia to disable SLI on the 1060 though, as putting to X60's together was always a good way to get at or above the X80 card sped at the $450-500 range depending on era. (2 $200 560Ti's > 580, etc). Sure SLI can have it's quirks, but when it does work, usually in games that you actually need it, it did work rather well, even when I was hacking a X48 Intel System to run SLI native. It was of course better at it then the 780/790i systems of the time lol. One thing about SLI. I remember reading up on 3DFX and their VooDoo chips right before they got bought out. The Voodoo 5 6000 had 4 of those VESA-100 chips and technically, with the tech the chips brought, they could stack as many of them as you could mount on a board and connect together (which sounded like a nightmare above 4 and had issues all over with AGP being what it was). But when everything worked, they scaled rather well. What is keeping Nvidia from using some of that old research and doing something similar (or is that what they are doing with the Compute Link with Tesla?) and instead of being SLI when they have a dual GPU board? Or is this sort of tech similar to what we have now with the Shader Cores and we have come so far now that all the magic that was T&L is just older than dirt tech news these days? EdEddnEddy fucked around with this message at 16:57 on Jul 19, 2016 |
|
#
?
Jul 19, 2016 16:53
|
|
|
Truga posted:VR needs to stay above 100fps at all times at like 8k resolution before it'll be halfway decent, so no, people aren't actually being hyperbolic. Well, yeah, for VR. Though I expect a lot of rendering tricks to end up helping along the way. 4k monitor gaming owns right now though.
|
|
#
?
Jul 19, 2016 16:54
|
|
|
Lungboy posted:First 1060s have started appearing: mini version here and standard version here.
|
|
#
?
Jul 19, 2016 16:56
|
|
lllllllllllllllllll posted:Please excuse the dumb question, but why are there two versions? Some people just want the best cooling possible, or want to look of a dual fan cooler, but really at the power draw of the 1060 the single fan cooler is going to be plenty.
|
|
|
#
?
Jul 19, 2016 16:58
|
|
|
lllllllllllllllllll posted:Please excuse the dumb question, but why are there two versions? The reference 1060 board is actually a lot shorter than the cooler, so swapping the Reference cooler for a short stubby one is actually the PCB's real size. Similar to the 480's it looks. It is up to the Aftermarket guys whether they want to run a super cooler on it extending the size, or run something smaller for the ITX crowd.
|
|
#
?
Jul 19, 2016 16:59
|
|
|
spasticColon posted:If AMD was actually competitive the GPU market wouldn't be such a mess. Even at their best I don't think I've ever seen AMD over 40% market share, and 30% is pretty typical for them. nvidia has always had mind and marketshare momentum. I remember during the geforce FX days they were still winning sales numbers and that was nvidia at their worst (or near to it).
|
|
#
?
Jul 19, 2016 17:05
|
|
|
One thing of note though is how that darn 390X seems to keep up in a lot of the benchmarks, even surpassing the 480 and 1060 in some. (on Guru3d.com at least). I do wonder if with time the 480 can be tuned to improvement to last a similar length of time. Hmm.
|
|
#
?
Jul 19, 2016 17:13
|
|
|
spasticColon posted:I myself think it's going to take another revolutionary leap in GPU tech in order for 4K gaming to become feasible. Maybe something similar to the jump to the Unified shader model? All that Nvidia and AMD have been doing for the last 10+ years is building on the Unified Shader Model technology so they probably need another revolutionary leap beyond that. Tremendously faster memory (HBM) is going to be necessary, but you're ultimately still going to run into compute/shading limitations that are getting harder and harder to improve on (certainly not without making chips a lot more expensive). On the plus side, various types of "smart upscaling" as discussed above are getting better and hopefully will help fill the gap. Hieronymous Alloy posted:Well, if dx12/Vulkan allow for true multi GPU without stutter or other downsides, then 4k gaming could start happening via parallel gpu's. Unfortunately it has tremendous downsides, because there is a limit to how any things you can parallelize within a frame and how much latency you can tolerate for that. Improving inter-chip bandwidth will help substantially and might be a revolution of the type you are describing. bull3964 posted:I think people are being a bit hyperbolic about 4k. Yeah, but Doom has a LOT of adaptive stuff in there to maintain that framerate (which isn't necessarily a bad thing so much as just an example of the type of compromises mentioned above). Anyways, personally I'm much more excited about 1440p-ish resolutions at > 200 Hz refresh rates than 4k.
|
|
#
?
Jul 19, 2016 17:20
|
|
|
EdEddnEddy posted:One thing of note though is how that darn 390X seems to keep up in a lot of the benchmarks, even surpassing the 480 and 1060 in some. (on Guru3d.com at least). I doubt there is any way of knowing of course, but what you're betting on here is how well (ie, how badly) AMD implemented its drivers and how much they will improve on them in the future. Any true *innovative* improvements both sides will accomplish for the most part. AMD just went through its driver revolution... It appears to me the 480 gets a much smaller DX12 boost due to reasons someone explained to me and I promptly forgot. There aren't a lot of other cards on the table here unfortunately except praying that the actual chips themselves get produced at the source. Produced better* in the sense that 150 watts on nvidia with the new node nets you a 1070... penus penus penus fucked around with this message at 17:37 on Jul 19, 2016 |
|
#
?
Jul 19, 2016 17:29
|
|
|
Truga posted:VR needs to stay above 100fps at all times at like 8k resolution before it'll be halfway decent, so no, people aren't actually being hyperbolic. Yes but that's actually going to be easy to do with eye tracking and foveated rendering, which will almost certainly be a standard feature before we see 8k OLED VR panels. https://www.youtube.com/watch?v=Qq09BTmjzRs I really do feel like there's going to be an end-game for GPUs. At comfortable viewing distance, there's strong diminishing returns past 4k. You can't differentiate the pixels unless you're way up close to a large monitor, but that would strain your eyes and your neck. VR Headsets have lenses that focus to infinity, so then you can use all of the PPI more than a normal monitor or TV. But your eyes can only notice high definition in the narrow center of your FOV, which means that a decade from now we could actually have both 16k OLED panels and a GPU to drive the fuckers selectively at the points you're looking at.
|
|
#
?
Jul 19, 2016 17:32
|
|
|
So when can we expect the next batch of 1060 cards? August? September?
|
|
#
?
Jul 19, 2016 17:34
|
|
|
THE DOG HOUSE posted:
Vulkan contains a ton of code from Mantle, AMDs discarded API. that's bound to play a factor, at least to begin with.
|
|
#
?
Jul 19, 2016 17:37
|
|
|
EdEddnEddy posted:Looking at Newegg for SnG, I see a bunch of 1060's at $250 with AIB like the Asus Strix for $300 which seems like a rather ok deal for a just released card you can actually get your hands on today. What is this witchcraft? I can't find a gtx 1060, reference or AIB, in stock anywhere. Also, do rx 480s even exist beyond blurred photos in the woods? I've never had to buy a new graphics card before. Is there any chance of actually getting either of these cards before mid August/September? Enigma fucked around with this message at 17:44 on Jul 19, 2016 |
|
#
?
Jul 19, 2016 17:42
|
|
|
Zero VGS posted:Yes but that's actually going to be easy to do with eye tracking and foveated rendering, which will almost certainly be a standard feature before we see 8k OLED VR panels. I know, I did say it's still hyperbolic :p Zero VGS posted:I really do feel like there's going to be an end-game for GPUs. At comfortable viewing distance, there's strong diminishing returns past 4k. IIRC the resolution required for "unnoticeable" natural vision at monitor range (~0.5m) is like 300dpi, beyond which point there's not enough nerves to differentiate the pixels, so I wanna say anything beyond 8k monitors won't be very useful, but antialiasing is pushing even that number down a bit so maybe even 6k will be "good enough"? .
|
|
#
?
Jul 19, 2016 17:44
|
|
|
EdEddnEddy posted:One thing of note though is how that darn 390X seems to keep up in a lot of the benchmarks, even surpassing the 480 and 1060 in some. (on Guru3d.com at least). The 290/290X/390/390X all have similar or higher theoretical single-precision compute performance, a higher pixel fill rate, and more memory bandwidth than the RX 480. I'm not that shocked that they're in the same ballpark, they've got way more transistors and bandwidth to throw at the problem and the RX 480 isn't clocked anywhere near something like the GTX 1060.
|
|
#
?
Jul 19, 2016 17:44
|
|
|
Zero VGS posted:I really do feel like there's going to be an end-game for GPUs. At comfortable viewing distance, there's strong diminishing returns past 4k. You can't differentiate the pixels unless you're way up close to a large monitor, but that would strain your eyes and your neck. Ultrawide is a thing. Surround gaming is a thing. People will learn not to rubberneck constantly. As someone who works with multiple monitors open it's a godsend, and xthetenth never stops jerking off about how awesome his X34 is for gaming. 5K 120hz ultrawide gaming is not a thing yet. And even then people love to crank up the supersampling and all that poo poo, so we will need to be effectively driving some 4x or 8x multiple of that. Every time a newer larger resolution comes out people say "oh, 1080p is so dense that you don't need supersampling because you can't see the aliasing" and then they inevitably turn it back on when GPU performance catches up. We have a ways to go. And by then we will probably come up with something even newer and even more wasteful of GPU horsepower. Paul MaudDib fucked around with this message at 17:50 on Jul 19, 2016 |
|
#
?
Jul 19, 2016 17:47
|
|
|
When I was watching the Digital Foundry 1080p comparison of the 1060 against the 480, 970, and 980 this morning I noticed some really big periodic frame time dips in the 480 graph on The Division and The Witcher 3 especially. Am I right in assuming that this may result in perceived hitchiness in some games even though overall framerate remains high? Or is this something that the average person will not even notice since the speeds involved are so fast?
|
|
#
?
Jul 19, 2016 18:00
|
|
spasticColon posted:So when can we expect the next batch of 1060 cards? August? September? Considering they do weekly shipments I would guess in a week. 
|
|
|
#
?
Jul 19, 2016 18:00
|
|
|
spasticColon posted:I'm still hoping a fool's hope I can snag an evga 1060 for $250-$260 but it's looking like I'll have to wait for the next batch of cards to come in stock and who the hell knows when that will be. I'm thinking at least a month but until then it's price-gouged slim pickings. one of my friends got a gigabyte 1060 for $234 with some promo code he found...
|
|
#
?
Jul 19, 2016 18:03
|
|
|
I'm getting antsy about 1440p performance and wondering if I should wait for the RX490 or jump on a Fury X firesale. I don't intend to go beyond 1440p 144hz for quite awhile, and I'm a little concerned Vega 10 will only promise 15-20% improvement over the Fury X (while consuming like 180-190W of course) and costing upwards of 600$.
|
|
#
?
Jul 19, 2016 18:03
|
|
|
drat, that Sapphire Fury was on sale for $299 ($275 if you paid with Paypal) and I woke up to it being sold out before I had even finished reading the post about it. 
|
|
#
?
Jul 19, 2016 18:04
|
|
|
Ok, who hijacked Faustian's account?
|
|
#
?
Jul 19, 2016 18:05
|
|
|
owl_pellet posted:When I was watching the Digital Foundry 1080p comparison of the 1060 against the 480, 970, and 980 this morning I noticed some really big periodic frame time dips in the 480 graph on The Division and The Witcher 3 especially. Am I right in assuming that this may result in perceived hitchiness in some games even though overall framerate remains high? Or is this something that the average person will not even notice since the speeds involved are so fast? Frametime dips aren't a problem, that means a frame was really easy and the GPU knocked it out really quick. Frametime spikes/framerate dips are a problem, that's microstutter and you will perceive that. All cards do it somewhat, some frames are just really tough, but if it's happening regularly you have a problem.
|
|
#
?
Jul 19, 2016 18:10
|
|
|
spasticColon posted:So when can we expect the next batch of 1060 cards? August? September? Stores get these things in stock daily. Just look at the history on nowinstock, stuff is in and out all the time.
|
|
#
?
Jul 19, 2016 18:11
|
|
|
Beautiful Ninja posted:So the realization just hit that we have aftermarket 1060's out on the market, some even for MSRP. The RX 480 looks like it still won't see aftermarket cards for another 3 weeks or so. There hasn't even been any significant stock of reference RX 480's here in the US since launch on top of no aftermarket card.
|
|
#
?
Jul 19, 2016 18:12
|
|
|
spasticColon posted:I myself think it's going to take another revolutionary leap in GPU tech in order for 4K gaming to become feasible. Maybe something similar to the jump to the Unified shader model? All that Nvidia and AMD have been doing for the last 10+ years is building on the Unified Shader Model technology so they probably need another revolutionary leap beyond that. What do you mean by feasible? It's already possible with 1080 SLI and the Titan P/1080 ti will almost certainly be able to handle it with a single GPU with anything but the most demanding titles. Affordable 4k on the other hand will take a lot longer.
|
|
#
?
Jul 19, 2016 18:13
|
|
|
FaustianQ posted:I'm getting antsy about 1440p performance and wondering if I should wait for the RX490 or jump on a Fury X firesale. I don't intend to go beyond 1440p 144hz for quite awhile, and I'm a little concerned Vega 10 will only promise 15-20% improvement over the Fury X (while consuming like 180-190W of course) and costing upwards of 600$. Gonkish posted:drat, that Sapphire Fury was on sale for $299 ($275 if you paid with Paypal) and I woke up to it being sold out before I had even finished reading the post about it. The catch is the 4 GB of VRAM, as always. That's really only just enough for 1440p right now, and because they can't just throw everything in VRAM there's some microstutter in games that bump up against the limit as stuff pages in and out. You also have to ask yourself where VRAM usage is headed 3 years down the road - AMD is optimizing VRAM utilization in DX11 right now, but will they continue to do that down the road? And will DX12/Vulkan game programmers be as diligent about it? Otherwise the Fury for $275 was a really fantastic deal and you'd be much better off with that than the 4 GB RX480 for sure.
|
|
#
?
Jul 19, 2016 18:15
|
|
|
So how does this 1060 compare to the 1070? Currently using a 780 which can at least run everything i'm playing on ok settings, so not in an urgent need to upgrade, but trying to decide whether to just get a 1060 now or wait until the holiday season and get a 1070 or 80 then.
Evrart Claire fucked around with this message at 18:22 on Jul 19, 2016 |
|
|
#
?
Jul 19, 2016 18:17
|
|
|
spasticColon posted:I myself think it's going to take another revolutionary leap in GPU tech in order for 4K gaming to become feasible. Maybe something similar to the jump to the Unified shader model? All that Nvidia and AMD have been doing for the last 10+ years is building on the Unified Shader Model technology so they probably need another revolutionary leap beyond that. Are you only considering PC here? I'm not meaning to start a stupid fanboy war here, but within a year the 4K consoles are going to be out, and you had better believe MS and Sony are going to browbeat their developers into hitting that 4K native res at all costs, whether it's framerate or avoiding expensive post-processing or what. 4K gaming is perfectly feasible right now, at 40-60 FPS!
|
|
#
?
Jul 19, 2016 18:17
|
|
|
Paul MaudDib posted:Frametime dips aren't a problem, that means a frame was really easy and the GPU knocked it out really quick. Yeah sorry, there was a dip in the graph but the graph was scaled inversely such that a dip was bad meaning a spike in time. It wasn't happening constantly in the videos, but for some of the games there were frametime problems for the 480 where there was none for the 970/980/1060.
|
|
#
?
Jul 19, 2016 18:19
|
|
|
lDDQD posted:Ok, who hijacked Faustian's account? I think you're confused about correctly prognosticating and lamenting the new node release as actually hating on AMD. My posting history in this thread is literally hopefulAMDcustomer.txt.
|
|
#
?
Jul 19, 2016 18:19
|
|
|
Paul MaudDib posted:Ultrawide is a thing. Surround gaming is a thing. People will learn not to rubberneck constantly. As someone who works with multiple monitors open it's a godsend, and xthetenth never stops jerking off about how awesome his X34 is for gaming. Super sampling just really inefficient but slightly more accurate anti aliasing though. The only good use I've seen of it is to make text more legible in VR where your'e resolution limited. Anti-aliasing itself is becoming less needed as resolutions go up too. I have an ultrawide and I've done triple monitor gaming as well. 4K VR will straight-up obsolete multi-monitor junk. Why spend a bajillion dollars on 3+ 4K monitors and SLI GPUs when you'll be able to do the same with a single headset?
|
|
#
?
Jul 19, 2016 18:28
|
|
|
|
| # ? May 16, 2024 00:45 |
|
|
Twerk from Home posted:Are you only considering PC here? I'm not meaning to start a stupid fanboy war here, but within a year the 4K consoles are going to be out, and you had better believe MS and Sony are going to browbeat their developers into hitting that 4K native res at all costs, whether it's framerate or avoiding expensive post-processing or what. Economically feasible. And the "new" consoles are going to do some kind of fancy up-scaling to meet the 4K target. It's gonna be a while yet before a $250 video card or a $400 console can run games natively at 4K with all the graphical whistles and bells turned on. Edit: Honestly I just don't understand the 4K gaming circlejerk. Movies and TV do look better on 4K sets but for gaming I'm satisfied with 1080p high/ultra settings at 60fps. spasticColon fucked around with this message at 18:34 on Jul 19, 2016 |
|
#
?
Jul 19, 2016 18:29
|
|