|
DonkeyHotay posted:I recently bought a 4k monitor with free sync. My 970 is definitely getting a little long in the tooth and I was wondering what the best way to get to 4k on AMD's current lineup, seeing as they don't seem to have any high end updates coming down the line until next year. It looks like you might get a used r9 fury on eBay for 350ish, what's the state of crossfire these days if I wanted to run two? wait till vega do this and see what is being offered.
|
#
?
Jun 4, 2016 03:13
|
|
|
|
| # ? May 15, 2024 14:25 |
|
|
DonkeyHotay posted:OK but I was mostly asking about crossfire, which still comes in cheaper than the street prices for a founder's edition. Yes, but it is a bad idea
|
|
#
?
Jun 4, 2016 05:21
|
|
|
SwissArmyDruid posted:So maybe we'll see Intel doing Async compute on their igpus down the line? Pretty sure they'll have to do it sooner or later because it's a part of DX12/Vulcan? Also, console people have reported up to 30% higher perf with using async compute before, and I'm sure they'll squeeze even more out after some years. Meanwhile, nvidia apparently isn't getting as good scheduling for 2 more years, this despite all the compute customers yelling they want what AMD has. I talk to some HPC guys every now and then and they're super hyped about this ~brand new tech~, that's supposedly been imminent forever now, and also annoyed, because they're mostly locked into CUDA now and can't switch. Here's hoping AMD gets their CUDA to OpenCL compiler working, it might end up being a serious thing for them.
|
|
#
?
Jun 4, 2016 05:53
|
|
|
I need suggestions on my use case. I run an i5-4570 with a GTX660 (Not Ti). I desire better framerates. However, I currently run two 1080p monitors and have no inclination to upgrade them at this stage. This is partially because I like them and have never felt hindered by lack of real estate and they're good IPS screens so they have a nice image. The other reason is I have no desire to upgrade to 4k (Which is what I feel is the next logical step at this stage, skipping past 1440p) because a) they cost a fucktonne and b) there are no cards that can push a consistent and high framerate at 4k. I guess I was kind of disappointed with the GTX1080 in that regard. 4k is being pushed as the next 1080p, and with good cause, but to that end the GTX1080 can't give a good ~60FPS a majority of the time which kind of defeats the point of it. I guess what I'm getting at is I feel it's complete overkill to get a GTX1080 for 1440p and under compared to a lesser but still proficient card and it doesn't cut it at 4k in a way I'd want. So that's my use case. I'm seriously leaning towards a RX480 for my next card if it comes out <AU$500 because it purports to have GTX970ish performance for less than they're going for. Am I right in assuming anything better is going to be complete overkill for 1080p gaming? When 4k becomes an attainable thing I'll consider upgrading to a kicking rad card but for now I think the RX480 would be the way to go, provided it lives up to its targets and claims, correct? PS: Not in a hurry to upgrade, so I'ma see how this 480vs1060 thing pans out.
|
|
#
?
Jun 4, 2016 06:08
|
|
|
You answered your own question assuming your monitors are of the 60hz variety. Wait for the reviews and take whatever card you feel delivers the performance you want for the price you want. Edit: You could also wait for the next Titan or 1080 Ti which should deliver solid 60 fps at 4k.
|
|
#
?
Jun 4, 2016 06:14
|
|
|
Truga posted:
I was hoping Pascal would reverse the reputation of nvidia gpus not aging very gracefully and you're saying that's exactly whats not going to happen?
|
|
#
?
Jun 4, 2016 06:20
|
|
|
Seamonster posted:I was hoping Pascal would reverse the reputation of nvidia gpus not aging very gracefully and you're saying that's exactly whats not going to happen? I'm not sure. I'm certain nvidia will do everything in their power to make the driver magically fix everything game developers throw at it, but I don't have enough knowledge to tell you whether or not they'll get "full performance" out of it. Also who knows if games will even use this extensively on PC, seeing how the market shares are. That said, 1080 is faster than anything AMD has out, and will likely stay that way for quite a while. And yeah, nvidia cards don't age as well as AMD cards, but 780 Ti is still a *good* card today. It's not anything to write home about anymore, but it's definitely not bad. It's just not quite as good as 290x anymore.
|
|
#
?
Jun 4, 2016 06:37
|
|
|
Truga posted:And yeah, nvidia cards don't age as well as AMD cards, but 780 Ti is still a *good* card today. It's not anything to write home about anymore, but it's definitely not bad. It's just not quite as good as 290x anymore. Why on earth does this phenomenon happen? How do AMD GPUs get comparatively faster over time than their nVidia counterparts? I've seen it talked about here like it's common knowledge, but it's utterly confusing to me.
|
|
#
?
Jun 4, 2016 06:49
|
|
|
Twerk from Home posted:Why on earth does this phenomenon happen? How do AMD GPUs get comparatively faster over time than their nVidia counterparts? I've seen it talked about here like it's common knowledge, but it's utterly confusing to me. AMD (ATi?) keeps jamming new tech into their cards as fast as they can, because why not. Their main competitor is nvidia, they want every edge they can get. Sometimes it doesn't quite work out (Fury's HBM issues, 20nm failures, etc), sometimes it makes nvidia sweat real hard. Last time that happened was HD5000 series though... Nvidia, I imagine, is going for planned obsolescence, because nvidia's main competitor is previous generation nvidia cards. So, they only put 3 gigs of ram on their flagship card, and it's now obsolete in any game that has enough vram requirements to go past that (every game made recently if you play on ultra). I was fully expecting 980Ti to only have 4 gigs of ram, and wouldn't have bought it, but then it came with 6 so I imagine it'll hold out a bit longer than usual. 980Ti was an outlier for nvidia in many ways I think.
|
|
#
?
Jun 4, 2016 06:56
|
|
|
Twerk from Home posted:Why on earth does this phenomenon happen? How do AMD GPUs get comparatively faster over time than their nVidia counterparts? I've seen it talked about here like it's common knowledge, but it's utterly confusing to me. Also from what I understand AMD drivers handled DX11 pretty badly (at least compared to Nvidia ), so any advantage it had in the HW would be held back by the SW. Recently they started improving the driver performance, so they've regained some of that speed back.
|
|
#
?
Jun 4, 2016 07:47
|
|
|
There was that story doing the rounds last year where an AMD exec claimed that their drivers were only allowing the cards to use 3% of their total potential in the worst possible cases. Probably talking about any number of DX11 titles that use gameworks.
|
|
#
?
Jun 4, 2016 08:22
|
|
|
Is there also the bit where Nvidia will send engineers to work with you on your pipeline for Nvidia specific enhancements? I remember ATI having to do a lot of catchup with that sort of thing which always meant there was a bigger gap at the start of the line than towards the end.
|
|
#
?
Jun 4, 2016 10:22
|
|
|
Evidently this thing is running 1070M or 1080M parts in SLI in a reduced-size desktop, so I wouldn't buy a notebook with high-end discrete graphics just yet: http://www.tweaktown.com/news/52317/msis-new-vortex-g65-rocks-nvidias-mobile-pascal-gpus-sli/index.html Also, if this isn't snake oil, it might finally get me interested in CLC down the road: http://www.tweaktown.com/news/52386/raijintek-demos-pumpless-liquid-cooling-system-computex/index.html BIG HEADLINE fucked around with this message at 11:28 on Jun 4, 2016 |
|
#
?
Jun 4, 2016 11:24
|
|
|
This is probably a dumb question - I'm hoping to buy a 1070 (or possibly a 480) but my dodgy old motherboard has a faulty PCI-E 2.0 x16 slot that I can't use and the only other slot is x4. How much is this going to affect performance? It's running with an overclocked 2500k CPU.
|
|
#
?
Jun 4, 2016 15:55
|
|
|
franchise1 posted:This is probably a dumb question - I'm hoping to buy a 1070 (or possibly a 480) but my dodgy old motherboard has a faulty PCI-E 2.0 x16 slot that I can't use and the only other slot is x4. You will lose some performance but not a devastating amount, I think it's okay as a stop-gap until you get a new motherboard. https://www.techpowerup.com/reviews/NVIDIA/GTX_980_PCI-Express_Scaling/21.html https://www.techpowerup.com/reviews/AMD/R9_Fury_X_PCI-Express_Scaling/18.html
|
|
#
?
Jun 4, 2016 16:19
|
|
|
repiv posted:You will lose some performance but not a devastating amount, I think it's okay as a stop-gap until you get a new motherboard. Looks like I should expect to get somewhere around 85-95% of the performance depending on the resolution then. Thanks for that.
|
|
#
?
Jun 4, 2016 16:28
|
|
|
mmkay posted:Also from what I understand AMD drivers handled DX11 pretty badly (at least compared to Nvidia ), so any advantage it had in the HW would be held back by the SW. Recently they started improving the driver performance, so they've regained some of that speed back. AMD's also got the consoles, so console games have been shifting to targeting GCN, which helps AMD cards gain back performance. I definitely think there's some architecture stuff going on because NV isn't consistent between architectures. For example the 780 ti is often behind the 970 but in the Division the 290 wins big but also the 780 Ti beats the 970 at 1080. If it were lol NV planned obsolescence that sort of thing wouldn't happen, older cards would fall steadily behind but instead there's some games that reward Kepler's strengths. NV makes somewhat shortsighted cards in a way but they're the right cards for when they launch, which is what matters to drive sales. AMD makes cards that will look like cards that will be targeted in the future but that's because they built that target. Also NV doesn't leave performance on the table to be gotten by those big drivers AMD has done. So AMD is for a lot of reasons both better positioned for the long term and worse positioned for the short term. Hopefully I'm right that their RTG drivers have been more on point (didn't they have launch day for total warhams and NV not?) and they'll do better in comparison.
|
|
#
?
Jun 4, 2016 16:39
|
|
|
Truga posted:Pretty sure they'll have to do it sooner or later because it's a part of DX12/Vulcan? Intel GPUs have supported compute since Ivy Bridge. Any chip that claims Vulkan support *must* support compute. It's not an optional feature. mmkay posted:Also from what I understand AMD drivers handled DX11 pretty badly (at least compared to Nvidia ), so any advantage it had in the HW would be held back by the SW. Recently they started improving the driver performance, so they've regained some of that speed back. They've also been exposing more and more GCN-specific functionality via extensions. Things like shader intrinsics which were previously only directly accessible on consoles and a Vulkan extension to enable relaxed primitive ordering, both of which can provide a performance boost when used properly. GCN does seem to be falling a little behind in certain areas though. Polaris looks like it still won't have support for conservative rasterization or fragment shader interlock to do advanced blending, although the latter could also be achieved by using Vulkan's concept of render passes.
|
|
#
?
Jun 4, 2016 16:53
|
|
|
The_Franz posted:Intel GPUs have supported compute since Ivy Bridge. Any chip that claims Vulkan support *must* support compute. It's not an optional feature. I think the difference is between spec compliant implementations and implementations that actually give a performance boost. The mainstream understanding is that the latter only is what counts.
|
|
#
?
Jun 4, 2016 17:23
|
|
|
ahmeni posted:Is there also the bit where Nvidia will send engineers to work with you on your pipeline for Nvidia specific enhancements? I remember ATI having to do a lot of catchup with that sort of thing which always meant there was a bigger gap at the start of the line than towards the end. It's not even "NVIDIA-Specific Enhancements" -- it's "hey this thing you are doing performs poorly, can you do it in this almost equivalent way that fits our hardware better?" The changes are either a) Net performance wins across the board b) Improvements for both NVIDIA and AMD that might be a bigger deal for NVIDIA c) Improvements mostly for NVIDIA, breaking even or with a slight detriment to AMD but increasing image quality (always with the option of having it be NV-only if it hurts AMD performance). d) An NV-Specific improvement that works better as a driver optimization. AMD doesn't have nearly the same manpower devoted to this sort of work. That's the real story behind GPUOpen and Mantle -- AMD isn't able to do the same level of developer support as NVIDIA, so they are constantly trying to encourage developers to do it all themselves. There are inherently appealing things about lower level APIs and open effects libraries, but don't think for a second it's not also a calculated strategic play to minimize AMD's relative weakness and NVIDIA's strength. The_Franz posted:Intel GPUs have supported compute since Ivy Bridge. Any chip that claims Vulkan support *must* support compute. It's not an optional feature. And the difference is that when NVIDIA tries to get people to use that technology, AMD cries about it being bloat and shovelware.
|
|
#
?
Jun 4, 2016 17:24
|
|
|
how about AMD releases cards
|
|
#
?
Jun 4, 2016 18:19
|
|
|
Do we have a good idea of what the 3rd party 1070s are looking like? Wondering if there's a short form factors coming in.
|
|
#
?
Jun 4, 2016 19:35
|
|
|
lurksion posted:Do we have a good idea of what the 3rd party 1070s are looking like? Wondering if there's a short form factors coming in. There's this motherboard with a 1070 built into it http://www.overclock3d.net/articles/gpu_displays/colorful_show_off_an_intel_skylake_motherboard_with_a_built_in_gtx_1070_gpu/1
|
|
#
?
Jun 4, 2016 19:39
|
|
|
Hubis posted:AMD doesn't have nearly the same manpower devoted to this sort of work. That's the real story behind GPUOpen and Mantle -- AMD isn't able to do the same level of developer support as NVIDIA, so they are constantly trying to encourage developers to do it all themselves. Even GPUOpen doesn't inspire a lot of confidence in their available manpower - the first public source dump of TressFX3 worked properly on NV and Intel but had gross artifacts on GCN. Not only did the library not working on their own hardware slip through testing, it took them over a month to fix it 
|
|
#
?
Jun 4, 2016 19:42
|
|
|
BIG HEADLINE posted:Evidently this thing is running 1070M or 1080M parts in SLI in a reduced-size desktop, so I wouldn't buy a notebook with high-end discrete graphics just yet: http://www.tweaktown.com/news/52317/msis-new-vortex-g65-rocks-nvidias-mobile-pascal-gpus-sli/index.html I read that the Raijintek is already ready for production--did they ever list a cooling capacity wattage? I'm leery of how quickly the pumpless AIO will be able to recirculate its coolant from the rad back down, especially if ambient temps aren't ideal. Cooler Master put up their new AIO solution (already in the market) that should require significantly less pump noise. https://www.techpowerup.com/223069/cooler-master-aio-masterliquid-pro-series-detailed Also, let's do away with jury-rigging our own AIOs onto our graphics dies altogether: https://www.techpowerup.com/223045/colorfuls-unique-air-liquid-cooling-solution-pictured That's an AIO crammed onto the GPU, no extra parts required. Sidesaddle Cavalry fucked around with this message at 19:53 on Jun 4, 2016 |
|
#
?
Jun 4, 2016 19:43
|
|
|
THE DOG HOUSE posted:how about AMD releases cards Seriously, they've released a grand total of four new chips since GCN 1.0 - Bonaire, Hawaii, Tonga, and Fiji. Four chips in four years, across all market segments. They're like the Valve of computer hardware. Paul MaudDib fucked around with this message at 20:04 on Jun 4, 2016 |
|
#
?
Jun 4, 2016 19:48
|
|
|
Paul MaudDib posted:Seriously, they've released a grand total of four new chips since GCN 1.0 - Bonaire, Hawaii, Tonga, and Fiji. Four chips in four years, across all market segments. Including getting sidetracked by getting a bit of a huge number of sales.
|
|
#
?
Jun 4, 2016 20:07
|
|
|
I had been planning on buying a 1070 as soon as it comes out, but now I find myself hesitating and thinking about throwing down the extra couple hundy for the 1080. Talk me into/out of it, if you feel like. My use case is pretty much exclusively gaming. I like 1440p, but don't mind dropping to 1080 if it's the difference between ~30ish FPS and 45+ at high settings. Don't have any desire to leap into VR.
|
|
#
?
Jun 5, 2016 00:49
|
|
Spiritus Nox posted:I had been planning on buying a 1070 as soon as it comes out, but now I find myself hesitating and thinking about throwing down the extra couple hundy for the 1080. Talk me into/out of it, if you feel like. The 1070 does like 80+ FPS at max settings in basically everything at 1440p, there is no reason to get a 1080, especially since the difference between the two cards is looking to be smaller than everyone thought once you take overclocking into account.
|
|
|
#
?
Jun 5, 2016 00:56
|
|
|
AVeryLargeRadish posted:The 1070 does like 80+ FPS at max settings in basically everything at 1440p, there is no reason to get a 1080, especially since the difference between the two cards is looking to be smaller than everyone thought once you take overclocking into account. Mm. Good to know. I suspected as much, but the 1080's earlier release date was making my resolve waver just a bit. Thanks.
|
|
#
?
Jun 5, 2016 01:40
|
|
|
This question might be better suited for another thread but it's video card related, so why not. I plan on getting a 1070 (oh man going from a 560Ti >_<) but my monitors are two dell u2412M, which have no HDMI, newer video cards don't have 2x DVI ports generally it seems, so I think I need to get a special HDMI to DVI cable correct? Is this what I'm looking for? So I can hook my two monitors to a modern video card. https://www.amazon.ca/AmazonBasics-HDMI-DVI-Adapter-Cable/dp/B00NH11X64/?s=electronics&ie=UTF8&qid=1465087569&sr=1-21
|
|
#
?
Jun 5, 2016 01:57
|
|
|
Im_Special posted:Is this what I'm looking for? So I can hook my two monitors to a modern video card. https://www.amazon.ca/AmazonBasics-HDMI-DVI-Adapter-Cable/dp/B00NH11X64/?s=electronics&ie=UTF8&qid=1465087569&sr=1-21 You could also do DP -> DVI
|
|
#
?
Jun 5, 2016 02:08
|
|
|
For your purposes HDMI to DVI is virtually a direct conversion and yes that will work
|
|
#
?
Jun 5, 2016 02:21
|
|
|
Psy890 posted:You could also do DP -> DVI I'm not really sure what this is, a quick google lookup, make it look to me like it's just a versatile display converter for hdmi/dvi/vga? Is there any advantages over the hdmi to dvi cable I linked to above? I assume I'll still be limited to DVI standards or whatever like no audio only digital, etc.
|
|
#
?
Jun 5, 2016 02:40
|
|
|
AVeryLargeRadish posted:The 1070 does like 80+ FPS at max settings in basically everything at 1440p, there is no reason to get a 1080, especially since the difference between the two cards is looking to be smaller than everyone thought once you take overclocking into account. For 1440p, does that include ultrawides at 1440p typically?
|
|
#
?
Jun 5, 2016 03:23
|
|
|
Luftwaffel posted:For 1440p, does that include ultrawides at 1440p typically? I couldn't find any ultrawide benchmarks for the 1070 specifically but here are some for the 1080 and Titan X. The Titan X drops into the 50s in some of the most demanding games but I would expect the 1070 is able to do 60+ FPS in pretty much any game at 3440x1440 especially if overclocked. http://techgage.com/article/nvidia-geforce-gtx-1080-review-a-look-at-4k-ultra-wide-gaming/
|
|
#
?
Jun 5, 2016 03:46
|
|
Luftwaffel posted:For 1440p, does that include ultrawides at 1440p typically? No, ultrawide adds ~27% more pixels to process and your frame rate will be somewhat lower. However a 1070 should still do very well at 3440x1440, especially once overclocked.
|
|
|
#
?
Jun 5, 2016 03:46
|
|
|
Im_Special posted:I'm not really sure what this is, a quick google lookup, make it look to me like it's just a versatile display converter for hdmi/dvi/vga? Is there any advantages over the hdmi to dvi cable I linked to above? I assume I'll still be limited to DVI standards or whatever like no audio only digital, etc. Most DisplayPort devices send a DVI/HDMI signal as well that can be filtered out and turned into a DVI connection by a passive adapter. This should work just fine, but HDMI and DVI use identical signaling so that definitely will.
|
|
#
?
Jun 5, 2016 05:30
|
|
|
Im_Special posted:I'm not really sure what this is, a quick google lookup, make it look to me like it's just a versatile display converter for hdmi/dvi/vga? Is there any advantages over the hdmi to dvi cable I linked to above? I assume I'll still be limited to DVI standards or whatever like no audio only digital, etc. HDMI and DVI are more or less the same thing, just with a different non-latching connector so that idiots don't rip their devices apart or whatever. DVI can actually carry an audio signal too. DisplayPort is actually different both in physical connector, signalling protocol, and voltage levels, but most GPUs can talk the HDMI standard across the DP connector if you use a passive adapter to connect it.
|
|
#
?
Jun 5, 2016 05:31
|
|
|
|
| # ? May 15, 2024 14:25 |
|
|
Spiritus Nox posted:I had been planning on buying a 1070 as soon as it comes out, but now I find myself hesitating and thinking about throwing down the extra couple hundy for the 1080. Talk me into/out of it, if you feel like. Even a 1080 can't do 120fps @ 1080p or 60fps @ 4k. I'ma wait for the 1080 Ti.
|
|
#
?
Jun 5, 2016 06:36
|
|
