|
Is it possible to enable DSR for only some games? It seems to disappear from the options in Nvidia CP when I pick a specific app. Also, what's the consensus on MFAA? Always enable it when you can use MSAA in a game? Or not?
|
#
?
Jan 19, 2016 21:22
|
|
|
|
| # ? Apr 29, 2024 12:51 |
|
|
DSR allows you to choose higher "virtual" resolutions in games. If you enable 4x (you should for integer scaling) it'll give you the option to set 2560x1440 and 5120x2880 in games (or whatever your monitors native resolution is and 4x that) and only the 5120x2880 resolution is the DSR one.
|
|
#
?
Jan 19, 2016 22:26
|
|
|
The_Franz posted:Fullscreen windows still have to go through the desktop compositor, which usually means there is an extra copy of the entire screen-sized window taking place. On a 1920x1080 screen at 60Hz that's roughly 500 megs of data per-second that needs to be copied around. That's no longer really true. If the swapchain is FLIP_SEQUENTIAL, then as of Windows 8, there's no additional copy, just the composition. As of Windows 8.1, if your window covers the whole screen and is top-most and is FLIP_SEQUENTIAL, then the OS flips directly to your swapchain and avoids both copy and composition. Even if it did have to copy, though, a 500MB buffer is about 2ms to copy on a 970, and it would each time be in parallel with the generation of the next frame. Even if it were totally serialized, that takes you from 60.2 to 60.1 fps.
|
|
#
?
Jan 19, 2016 23:41
|
|
|
The_Franz posted:Fullscreen windows still have to go through the desktop compositor, which usually means there is an extra copy of the entire screen-sized window taking place. On a 1920x1080 screen at 60Hz that's roughly 500 megs of data per-second that needs to be copied around. Windows has done the optimization since 8 (and gotten better at it in 8.1 and 10), but the program/game needs to also be using the D3D flip modes introduced in 8 to trigger it, so it's fairly uncommon. Windows 10 can even use some GPU features or another new mode to pull it off with not-quite-fullscreen windows, such as when a notification is being displayed. Edit: As part of the D3D12 info, Microsoft made a video about all the different presentation modes, and what frame delay/max fps limits they have https://www.youtube.com/watch?v=E3wTajGZOsA Kobata fucked around with this message at 00:15 on Jan 20, 2016 |
|
#
?
Jan 19, 2016 23:52
|
|
|
Update on my Zotac GTX 970 (regular 2 fan edition): A new PSU fixed the computer resetting in games problem and even the coil whine problem.
|
|
#
?
Jan 20, 2016 13:43
|
|
|
PirateBob posted:Update on my Zotac GTX 970 (regular 2 fan edition): A new PSU fixed the computer resetting in games problem and even the coil whine problem. Good to hear  Wish my issue was the PSU 
|
|
#
?
Jan 20, 2016 16:12
|
|
|
PirateBob posted:Update on my Zotac GTX 970 (regular 2 fan edition): A new PSU fixed the computer resetting in games problem and even the coil whine problem. Did it reset in all games or just a select one?
|
|
#
?
Jan 20, 2016 18:32
|
|
|
Is there a GTX 970 shortage going on? I'm pricing them out, and everything is $20-30 more than back in November, and EVGA doesn't even list it in their B-stock.
|
|
#
?
Jan 20, 2016 22:11
|
|
|
Stanley Pain posted:Good to hear It still might be. People assume that a PSU issue means not enough total wattage or not enough rail amperage but it could also be fluctuations in voltage levels, i.e. ripple. If the circuitry is there to handle high amounts of current but voltages are all over the place, your components will not be cool with that either. If you've got a 1000W PSU that can indeed handle 1000W of load but the 12V rail is drooping to 11.1v under that load then you don't have a very good unit.
|
|
#
?
Jan 20, 2016 22:20
|
|
|
B-Mac posted:Did it reset in all games or just a select one? It reset in all games unless I ran adaptive half-refresh vsync (30 fps). And yeah, what Seamonster said. These 970's can be finicky bitches and there are many bad PSUs out there, despite impressive wattage numbers.
|
|
#
?
Jan 20, 2016 22:37
|
|
|
Mutation posted:How do you sell something as expensive as a Titan X without putting a target on yourself for scammers? Sell to Goons. They're more trustworthy because they have been blessed by the Rowdy Trout.
|
|
#
?
Jan 20, 2016 23:11
|
|
|
Any reason why I shouldn't sell my SLI 970s and switch over to a 980ti right now? SLI is a lot easier than when I did it with 8800GTs back in the day, but I've found myself mostly playing games that don't support SLI at all. My second card probably hasn't done anything except render my desktop background and youtube videos on the other monitor in the past couple weeks. To top it off, I've managed to start exceeding 4GBs of VRAM usage in X-Plane 10 thanks to mods, and while the 512MB of slow RAM on a 970 isn't really that noticeable, running out of VRAM completely sure as hell is. I've gotta a buddy who's in the market for one who would pay me close to what I paid originally, and after SA-Mart/Ebaying the other it wouldn't be that much for the difference for a new 980ti. I just haven't been keeping up with updates in the GPU market since I got the 970s, and I wanted to make sure that I'm not about to have to resell this new 980ti before it depreciates anymore because the next lineup gets announced a month later. And is there any reason for me to switch away from MSI? I've really appreciated how quiet their cooler's are since I've switched to them, not to mention I quiet like the look of the card.
|
|
#
?
Jan 21, 2016 07:17
|
|
|
This gen , you have every reason to do exactly that. A 980ti is virtually 970 sli, and practically speaking even superior in many ways
|
|
#
?
Jan 21, 2016 07:49
|
|
|
T800 posted:Any reason why I shouldn't sell my SLI 970s and switch over to a 980ti right now? you don't want to wait 6-9 months to see what the next gen will bring? you can always sell your 970s then.
|
|
#
?
Jan 21, 2016 07:59
|
|
|
Seamonster posted:It still might be. People assume that a PSU issue means not enough total wattage or not enough rail amperage but it could also be fluctuations in voltage levels, i.e. ripple. If the circuitry is there to handle high amounts of current but voltages are all over the place, your components will not be cool with that either. If you've got a 1000W PSU that can indeed handle 1000W of load but the 12V rail is drooping to 11.1v under that load then you don't have a very good unit. future ghost fucked around with this message at 08:34 on Jan 21, 2016 |
|
#
?
Jan 21, 2016 08:30
|
|
|
T800 posted:Any reason why I shouldn't sell my SLI 970s and switch over to a 980ti right now? The Lord Bude posted:you don't want to wait 6-9 months to see what the next gen will bring? you can always sell your 970s then. Still, we're looking at getting the top model Pascal first when it releases. If you're not looking to spend the $$$$ to early adopt you could be without a video card if you sell before Pascal hits. SlayVus fucked around with this message at 08:35 on Jan 21, 2016 |
|
#
?
Jan 21, 2016 08:31
|
|
|
SlayVus posted:Still, we're looking at getting the top model Pascal first when it releases. If you're not looking to spend the $$$$ to early adopt you could be without a video card if you sell before Pascal hits. I'm not sure where this is coming from? Last I heard little Pascal is what will hit the market first because big Pascal is going to get gobbled up in the HPC sector for some time as Teslas and Quadros. Nvidia has a commanding position in the consumer sector, they can stand to potentially lose a few sales there to make sure AMD doesn't edge them out in HPC, as well as let supply reach the point that prices for Big Pascal become competitive. If they try to do both they'll end up in a reverse 980ti: Fury X situation where the new Fury might offer similar or better performance at a lower price point while Nvidia tries to provide for both markets. To do what you say would require immensely good yields on GP100. IMHO, everything points to either Nvidia releasing second and later, but more will be clarified in April.
|
|
#
?
Jan 21, 2016 08:57
|
|
|
SlayVus posted:Personal suggestion, sell your cards before retail release if the reviews for Pascal were like Maxwell. 2nd market for 600/700 series saw a 50% reduction in price because of Maxwell. Better to lose some money instead of a whole video card amount of money. Are you suggesting sell now and not have a GPU until Pascal hits? I would urge against that, Pascal isn't releasing for months and it would not be surprising to not see cards priced effectively until the summer.
|
|
#
?
Jan 21, 2016 13:12
|
|
|
i'm pretty sure the implication is "sell those cards and get a 980ti instead for the time being"
|
|
#
?
Jan 21, 2016 13:14
|
|
|
FaustianQ posted:I'm not sure where this is coming from? Last I heard little Pascal is what will hit the market first because big Pascal is going to get gobbled up in the HPC sector for some time as Teslas and Quadros. Nvidia has a commanding position in the consumer sector, they can stand to potentially lose a few sales there to make sure AMD doesn't edge them out in HPC, as well as let supply reach the point that prices for Big Pascal become competitive. If they try to do both they'll end up in a reverse 980ti: Fury X situation where the new Fury might offer similar or better performance at a lower price point while Nvidia tries to provide for both markets. To do what you say would require immensely good yields on GP100. The WCCFTech rumour back in October was that we'd get the 1080ti in April along with a professional card that would have a max 16GB HBM2. Then in Q3 there would be the low end GDDR5x cards and a new Titan with yet another pro card that could have up to 32GB HBM2. Their reasoning was the expected shortage of HBM2 and nvidia rushing to get a VR Max settings card out before the Oculus Rift. It would also allow them to counter any move AMD made between April-October. Turns out the chips they claimed were GP100 with a fake Maxwell code on the shipping container were probably 960Ti mobile gpu's for laptops due in April aiming for minimum settings VR. Nobody has reported seeing any sign of Pascal chips being ready for Q2 so far, which sucks. Except for my friend who claims he bought $800 worth of AMD shares right before the Polaris rumours and Samsung deal were announced.
|
|
#
?
Jan 21, 2016 14:11
|
|
|
Seamonster posted:It still might be. People assume that a PSU issue means not enough total wattage or not enough rail amperage but it could also be fluctuations in voltage levels, i.e. ripple. If the circuitry is there to handle high amounts of current but voltages are all over the place, your components will not be cool with that either. If you've got a 1000W PSU that can indeed handle 1000W of load but the 12V rail is drooping to 11.1v under that load then you don't have a very good unit. Yeah, I'm 100% sure the PSU is fine. It's a Corsair AX1200i and both the software and my own tests show that all the rails performing well within spec. I should probably just RMA the cards and try my like with some refurbs.
|
|
#
?
Jan 21, 2016 14:12
|
|
|
THE DOG HOUSE posted:This gen , you have every reason to do exactly that. A 980ti is virtually 970 sli, and practically speaking even superior in many ways Like the aforementioned VRAM and working on every game issues. Definitely do it! Also I really like MSI, I had a 580 Lightning and loved it. I think they have the best cooling design and their construction quality is high, which is good because I have heard their customer support is terrible.
|
|
#
?
Jan 21, 2016 14:37
|
|
|
Verizian posted:The WCCFTech rumour back in October was that we'd get the 1080ti in April along with a professional card that would have a max 16GB HBM2. Then in Q3 there would be the low end GDDR5x cards and a new Titan with yet another pro card that could have up to 32GB HBM2. Their reasoning was the expected shortage of HBM2 and nvidia rushing to get a VR Max settings card out before the Oculus Rift. It would also allow them to counter any move AMD made between April-October. To be fair, I'm pretty bullish on AMD too, but I don't play the stock market, or I might have dropped a thou on AMD shares as well. I'm pretty much on a hairtrigger to buy if Polaris/Zen pans out, though.
|
|
#
?
Jan 21, 2016 15:35
|
|
|
The only consideration in favor of 970 SLI is the small bit less than a 980 Ti it's worth and the two cards potentially devaluing less (and tbh I bet there'll be a good used market for the Ti if you want to flip it). Otherwise it's the same or greater performance without the annoyance. Price out your best guess on how much you'd pay to shift from the 970s to the Ti, and ask if you want to pay that much for a big jump in performance in some of your games, smoother frames where that doesn't happen, and no memory issues. Personally the answer's yes. I very much doubt big Pascal will be relevant until nearly a year after Polaris and small Pascal drop. It's like big Kepler. It'll exist, but you'll know if you're in the market and that market is "will pay thousands for top HPC performance", and with that eating good dice, it'll probably be a year till the yields get good enough to sell it for a grand, unless they go the salvage core route. AMD stock is a bet on Polaris and Zen, and I'm hoping they're going to do well and actually compete respectively, so I'm a bit tempted. It'll be interesting to see if they're doing decently when their debt matures. Heck of a gamble though. xthetenth fucked around with this message at 15:52 on Jan 21, 2016 |
|
#
?
Jan 21, 2016 15:45
|
|
|
there's no reason not to get off of that SLI setup on the performance front as well a 970-SLI setup uses up to 380w before overclocks, whereas the 980ti caps out at 250w as advertised, so you'll have significantly more headroom if you want to overclock a 980ti, and that has more or less the same frequency cap and better (lower) power climb than smaller maxwell chips edit: glhuhgalghhuhg Anime Schoolgirl fucked around with this message at 16:27 on Jan 21, 2016 |
|
#
?
Jan 21, 2016 16:21
|
|
|
Someone should give me a 970! :bigtran:
|
|
#
?
Jan 21, 2016 16:26
|
|
|
http://wccftech.com/nvidia-flagship-pascal-gpu-2h-2016/ finally at least a rumor for a release time period. Although its nothing new, everybody pretty much guessed second half of this year anyway edit: pretty exciting year probably. two actual releases probably around the same time penus penus penus fucked around with this message at 21:34 on Jan 22, 2016 |
|
#
?
Jan 22, 2016 21:16
|
|
|
I can't seem to find which HDMI version my R9 290 has. Can my R9 290 support 1440p @ 60hz over HDMI? Or do I have to use a displayport cable? My monitor only has one displayport input but two HDMI inputs. I have my keyboard/mouse hooked up to the USB hub on my monitor and it'd be nice to only switch out the USB wire when switching from my MBP to my Desktop PC.
|
|
#
?
Jan 22, 2016 21:41
|
|
|
THE DOG HOUSE posted:http://wccftech.com/nvidia-flagship-pascal-gpu-2h-2016/ finally at least a rumor for a release time period. Although its nothing new, everybody pretty much guessed second half of this year anyway Broadwell-E is also coming out around the same time. Look forward to consumer grade 10c/20t CPUs.
|
|
#
?
Jan 22, 2016 22:26
|
|
|
Rabid Snake posted:I can't seem to find which HDMI version my R9 290 has. Can my R9 290 support 1440p @ 60hz over HDMI? Or do I have to use a displayport cable? My monitor only has one displayport input but two HDMI inputs. I have my keyboard/mouse hooked up to the USB hub on my monitor and it'd be nice to only switch out the USB wire when switching from my MBP to my Desktop PC. Some cursory Googling says the R9 290 and other 200 series cards only have HDMI 1.4. Looks like you should be good, from what I found the 1.4 revision supports 1440p up to 80Hz but higher resolutions like 2K/4K cap out around 25-35Hz
|
|
#
?
Jan 22, 2016 22:51
|
|
|
Rabid Snake posted:I can't seem to find which HDMI version my R9 290 has. Can my R9 290 support 1440p @ 60hz over HDMI? Or do I have to use a displayport cable? My monitor only has one displayport input but two HDMI inputs. I have my keyboard/mouse hooked up to the USB hub on my monitor and it'd be nice to only switch out the USB wire when switching from my MBP to my Desktop PC. The NVIDIA GTX 900 series are the only cards on the market with HDMI 2.0 at the moment. I think HDMI can do 60hz at 1440p but I'm not positive.
|
|
#
?
Jan 22, 2016 23:38
|
|
|
Where's the foreseen need of 32GB of VRAM?
|
|
#
?
Jan 23, 2016 05:56
|
|
|
Tab8715 posted:Where's the foreseen need of 32GB of VRAM? In response to the quad-die thing? Having enough bandwidth and capacity to feed 4x600mm^2 dies. To the Titan thing? The compute market wants literally as much as they can possibly have and will pay out the rear end for it. The more memory you have, the better GPU compute works. Having to start and stop is the slowest part of GPGPU and more memory means a larger scale and thus less stopping. Paul MaudDib fucked around with this message at 06:00 on Jan 23, 2016 |
|
#
?
Jan 23, 2016 05:57
|
|
|
http://techreport.com/news/29633/jedec-makes-gddr5x-official 10-14 Gb/sec bandwidth per-die, about double GDDR5. http://techreport.com/news/29632/rumor-amd-may-cut-radeon-r9-fury-prices-soon Called it~ e: I'm taking that back. A rumor isn't definitive proof. SwissArmyDruid fucked around with this message at 12:12 on Jan 23, 2016 |
|
#
?
Jan 23, 2016 12:04
|
|
|
Tab8715 posted:Where's the foreseen need of 32GB of VRAM? There are some edge cases that will require 32GB VRAM, but 64GB VRAM is than anyone will ever need.
|
|
#
?
Jan 23, 2016 12:20
|
|
|
FSMC posted:There are some edge cases that will require 32GB VRAM, but 64GB VRAM is than anyone will ever need. Famous last words? I know Bill Gates didn't actually say the thing that's attributed to him, but still.
|
|
#
?
Jan 23, 2016 12:23
|
|
|
FSMC posted:There are some edge cases that will require 32GB VRAM, but 64GB VRAM is than anyone will ever need. HPC loooooooooooooves memory.
|
|
#
?
Jan 23, 2016 16:29
|
|
|
xthetenth posted:HPC loooooooooooooves memory. Truth. There are some applications in my line of work where we could easily use 1TB of memory on a GPU.
|
|
#
?
Jan 23, 2016 17:01
|
|
|
I work on high resolution video processing using neural networks as a hobby. My entire architecture would change (for the better) if I could treat the GPU memory like it's uh... actually memory rather than a manually managed L2 cache in practice.
|
|
#
?
Jan 23, 2016 20:44
|
|
|
|
| # ? Apr 29, 2024 12:51 |
|
|
Is there any kind of technical documents on how Nvidia gets the reduced input latency on the GTX 950?
|
|
#
?
Jan 23, 2016 22:22
|
|