Message boards :
Number crunching :
GPU Wars 2016: GTX 1050 Ti & GTX 1050: October 25th
Message board moderation
Previous · 1 . . . 13 · 14 · 15 · 16 · 17 · 18 · 19 · Next
Author | Message |
---|---|
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13727 Credit: 208,696,464 RAC: 304 |
"Faster, better, cheaper"? Faster, yes. Better, yes. Cheaper, not quite yet. Available, any day now... Grant Darwin NT |
shizaru Send message Joined: 14 Jun 04 Posts: 1130 Credit: 1,967,904 RAC: 0 |
"Faster, better, cheaper"? Yes, that's the one :) Just using it as an example of "there's no such thing as a free lunch". Here's archae's thread over at Einstein for anybody that's missed it (lots of good info there so head on over) https://einstein.phys.uwm.edu/forum_thread.php?id=12101 Hmmm 80C does seem a bit high... but not quite a Fermi toaster :P Anyway, archae appears to understand chip design better than anyone so I'll shut up now. I'm sure he'll let us know if it's the driver, the TIM, the FE's fan design, or just plain Pascal :) |
archae86 Send message Joined: 31 Aug 99 Posts: 909 Credit: 1,582,816 RAC: 0 |
Hmmm 80C does seem a bit high... but not quite a Fermi toaster :P Currently I'd score it as a lazy fan speed curve in the default settings. I used MSI Afterburner to impose a somewhat more ambitious fan speed curve, and found the Einstein BRP6 work to run at 63.5C in a condition where it had been running about 76 (the room was a good bit cooler in the wee hours of the morning than it had been late afternoon of the previous day). That actually caused the core clock rate to move up a little (all the rates I've seen on Einstein are far above the published boost rate, let alone the published base rate--presumably because Einstein BRP6 work does not get the GPU power consumption up as a high as presumed for the published rates). I don't have the exact comparative fan speeds for the two cases at hand, but the default settings running BRP6 had the fan at about 53% (and pretty darn quiet), while on my first try revised curve it settled in at something like 73% (noticeably louder, but not really obnoxious). The Founders Edition fan is pretty good on the "agreeable sound" front, but does get noticeably noisier to my ear in my environment as the indicated % speed goes up above 55 or so. I've got a 140mm bottom fan currently running about 800 rpm on the case. I plan to try turning that up a bit--perhaps the 1070 will like getting more cool air input. But given the case design and my other components, I'm not sure there is much to gain there. Meanwhile, I did run my 1070 here at SETI on the stock application, with stock 1070 settings (other than moderately boosting fan speed.). For work of type SETI@home v8 v8.12 (opencl_nvidia_SoG), with guppi vlar work, the elapsed time running a single GPU task, with one SETI task, was pretty close to 10 minutes 35 seconds. I noticed that over the progress of a single WU, there was a strong and repeatable pattern of variation in the GPU load %, and strong shorter-term fluctuations in memory controller load, bus interface load, and GPU power consumption. I'll go back to Einstein for the moment, but will entertain suggestions as to test cases here which might be of use. I've not been working here for some time, so if an interesting test case involves non-default code or settings, I'll appreciate being pointed to suitable instructions. For Einstein work, I'm encouraged. The power efficiency is really very good, I think, and the single-card output is pretty high in the pack. A lower purchase price, and quieter, more capable fans would make it something easy to recommend more broadly. Maybe Nvidia will offer a 1060 or a 1050 at a dramatically lower price which retains enough Einstein performance and power efficiency to be really broadly attractive. |
arkayn Send message Joined: 14 May 99 Posts: 4438 Credit: 55,006,323 RAC: 0 |
Grant I think you are describing a straight-up die shrink of an exact same chip design... Whereas the GP104 is a new architecture so there's no baseline really. Running 2 tasks on the 1070, I am at 55c and the fan at 59%. Running 2 tasks on the 960, I am at 41c and the fan at 32%. Of course the 1070 is spanking the 960. 6/21/2016 20:39:06 | | CUDA: NVIDIA GPU 0: GeForce GTX 1070 (driver version 368.39, CUDA version 8.0, compute capability 6.1, 4096MB, 3048MB available, 6463 GFLOPS peak) 6/21/2016 20:39:06 | | CUDA: NVIDIA GPU 1: GeForce GTX 960 (driver version 368.39, CUDA version 8.0, compute capability 5.2, 4096MB, 3067MB available, 2618 GFLOPS peak) 6/21/2016 20:39:06 | | OpenCL: NVIDIA GPU 0: GeForce GTX 1070 (driver version 368.39, device version OpenCL 1.2 CUDA, 8192MB, 3048MB available, 6463 GFLOPS peak) 6/21/2016 20:39:06 | | OpenCL: NVIDIA GPU 1: GeForce GTX 960 (driver version 368.39, device version OpenCL 1.2 CUDA, 4096MB, 3067MB available, 2618 GFLOPS peak) 6/21/2016 20:39:06 | | OpenCL CPU: AMD FX(tm)-8320 Eight-Core Processor (OpenCL driver vendor: Advanced Micro Devices, Inc., driver version 1084.4 (sse2,avx,fma4), device version OpenCL 1.2 AMD-APP (1084.4)) |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13727 Credit: 208,696,464 RAC: 304 |
Running 2 tasks on the 1070, I am at 55c and the fan at 59%. Founders edition, single centrifugal (blower) fan? eg Running 2 tasks on the 960, I am at 41c and the fan at 32%. Non reference card, larger than reference heatsink with multiple fans? eg Both in the same system? GTX 960 closest to edge of motherboard (bottom of case), GTX 1070 closest to CPU? GTX 1070 150W GTX 960 120W GTX 960 draws 20% less power, most likely has a better cooler, and possibly better position in case (not drawing in heat from the back of other card)? I suspect they're the most likely reasons for the temperatures seen here; when I had 2*GTX 750Tis in my C2D, the one drawing air in from the back of the other video card was only a couple of degrees warmer, but the fans were about 10-12% faster to keep it there. Grant Darwin NT |
AMDave Send message Joined: 9 Mar 01 Posts: 234 Credit: 11,671,730 RAC: 0 |
|
arkayn Send message Joined: 14 May 99 Posts: 4438 Credit: 55,006,323 RAC: 0 |
Running 2 tasks on the 1070, I am at 55c and the fan at 59%. Yep. http://www.newegg.com/Product/Product.aspx?Item=N82E16814125870 https://www.amazon.com/gp/product/B00V4HY522/ I did have a 760 in the case with the 960, but it came out, the 960 moved down to the 4x slot and the 1070 went into the 16x slot. |
W-K 666 Send message Joined: 18 May 99 Posts: 19048 Credit: 40,757,560 RAC: 67 |
For historical perspective: Thanks for that, passed it on to some friends. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13727 Credit: 208,696,464 RAC: 304 |
For historical perspective: Good link. Another interesting one is to compare cards over the last 12 or so years, and clicking on price/performance at the top of the graph gives another perspective on things. PassMark Direct Compute benchmark comparisons. Grant Darwin NT |
AMDave Send message Joined: 9 Mar 01 Posts: 234 Credit: 11,671,730 RAC: 0 |
For historical perspective: Interesting indeed.  The Videocard Value (G3D Mark / $Price ) of the GTX 750 Ti (and my GTX 950) surpasses that of the GTX 970. |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Well, I'm happy to announce that my Founders Edition 1080 is now enjoying life as Hybrid. Running 1.89 GHz at 43C. Much nicer than that 76C I was seeing since I had installed it. Not the prettiest belle at the ball, but it's about function. I'm also happy to report that removing of the backplate was not required, though a special tool was needed to get the power plug of the fan out of it's hole. Here's a pic of the final product, had to leave the front half of the shroud off but otherwise it runs fine. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13727 Credit: 208,696,464 RAC: 304 |
Running 1.89 GHz at 43C. Almost 2Ghz and less than 50°c, damn that's nice. had to leave the front half of the shroud off but otherwise it runs fine. Yeah, no issue there. You need the full shroud to get the air to do through the heatsink. With the water block there, it's not needed; although you do need the shroud around the fan, otherwise it won't blow any air for the remainder of the card that isn't cooled by the waterblock. Grant Darwin NT |
Al Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 |
Zalster, where did you pick that waterblock up at? I received an email from Koolance yesterday for a waterblock for the 1070/80, in fact I pulled it up within 30 seconds of getting the email, and the price showed as $0.00, so I added one to my cart, chose express shipping for $13 and placed the order. Not surprisingly, the order was cancelled later in the afternoon, and I called them for fun about it, they said that the email went out early, and no, I hadn't won the prize of a free block for being the first person to order one (dang), but they should be in soon. I checked back today, they had a price on them this time, as well as a notify button. Not exactly cheap ($130), how much did yours cost? Not sure how it's temps would compare to a full coverage block, but I like the idea of self contained, none of the extra hardware (pumps, rads, hoses, etc) needed, just plug and play. *edit* Oh, and if you would, please tell me more about said special tool for removing the plug you mentioned. Was it included with the waterblock? |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Hey Al, Actually the kit is the hybrid kit for the 980Ti. You can use any of the hybrid kits for all of the 980, 980Ti and Titan Xs. They are all the same. The only difference is the shroud that comes with the kit, but since I'm the only one that sees it, I don't care what the shroud says. I ordered mine before the 1080 came out so that I would have it when my card arrived. Just been sitting on it until I decided to try modifying my card. It's a bit harder than previous cards but can be done. You can remove 4/5 of the shroud on the 1080, the very last part of the shroud that covers the vram fins can't be removed without removing the back plate and some of other things. This is what it looked like You can see the fan pin connection in the upper right corner but it's kind of blocked by the last part of the shroud. But I was able to get in there with this Used it to pull the pin out of it's plug. Some will figure out what this is. Just a spare I had lying around, lol I needed something long and angled as needlenose plyers weren't going to do the job. |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Al, Newegg has some open box for $79.99 http://www.newegg.com/Product/Product.aspx?Item=N82E16814998105R&nm_mc=KNC-GoogleAdwords-PC&cm_mmc=KNC-GoogleAdwords-PC-_-pla-_-Accessories+-+Video+Card-_-N82E16814998105R&gclid=CjwKEAjwka67BRCk6a7_h_7Pui8SJABcMkWR2_th_fswu_hyt7I40xtFDxB4ZnZv_1W77AcI8u8-YBoCM1nw_wcB&gclsrc=aw.ds |
Al Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 |
Surgical tools for performing card surgery, how appropriate! I'm suprised that the blocks from the previous series cards fit this one, I would have thought they would have changed things up as the chip size is quite a bit different if I recall correctly? Well, good to know then, but I just went to their site, and they don't list one as 'for' the 1080, I wonder if they are going to release one specifically for it, or just re-label it as 980/Titan/1080 compatible? Looking at their site, it appears they list 3 available, one for the Titan X (P/N: 400-HY-0990-B1) and one for the 980Ti (P/N: 400-HY-0996-B1) and one for the standard 980 (P/N: 400-HY-H980-B1). I've found often times that EVGA's numbering scheme is that higher the number, the higher performance, so I wonder if the one for the 980Ti is better performing than the Titan X? Any thoughts on that one? I just went on Amazons site to check on pricing on the 980Ti version, and found that they are selling it for $59! I did look at listings from all sellers, and found the next cheapest one was a new one from a guy who bought it, but then has to sell it because it doesn't work for his card? He said "Never used, only selling cause the product does not work with ACX cooler style GPUs so only purchase if you own the non-cooler version." Not sure exactly what that means, but man, for $60 it seems pretty hard to go wrong if it fits. Only bummer about it is that shipping (in stock) time on that one is 2-4 weeks, EVGA must be out of that model, because I tossed it into my cart, and it said Estimated delivery: July 13, 2016 - July 20, 2016. Wonder if it is the 'best' one, I'll have to do some reading up on that one. As mine is the non-reference, FTW version, might this be an issue? *edit* Just tossed the order in for one of them, but I will have to do some more research, after thinking about what the guy mentioned above, I think I would need a blower style card, and mine is an ACX type cooler, so that may be an issue? What are your thoughts? As I have 2-4 weeks before it ships, I'm not terribly worried, I can always cancel my order. Oh, and just noticed you answered my question, the difference in the P/N's is what is listed on the shroud. Yeah, like you, I really don't care what it says, just that it works! :-) |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
The hybrid kits are all the same. They just label them differently for each new set of cards. ACX means he got a card that has 2 fans on it, ie not a reference model that only has 1 fan. The hybrid kits only work on reference models as you need the original fan to cool the vram. What can I tell you, the mounting points are all the same. So you can use them on any of the 900 series, Titan X or the 1080 but again, it needs to be a reference design. |
Al Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 |
Crap, that is kind of what I was thinking after doing a little more digging after posting that last comment. Hmm... Maybe I can pick up a cheap reference blower cooler from someone who has put a full water block on their 1080. The only concern would be if the mounting locations (or anything else for that matter) is physically different on the non-reference cards compared to the reference versions, which would prevent me from doing this. The cost would still be lower than doing a full blown water cooled setup, with all the addl parts needed, I would think that it would probably be about half the cost. Fun living on the bleeding edge, eh? |
Lionel Send message Joined: 25 Mar 00 Posts: 680 Credit: 563,640,304 RAC: 597 |
The attached link goes to an article at Tom's hardware and talks about power consumption of the AMD Radeon RX480. According to the article, it looks as though the card draws more power than AMD indicates. http://www.dvhardware.net/article64693.html |
zoom3+1=4 Send message Joined: 30 Nov 03 Posts: 65738 Credit: 55,293,173 RAC: 49 |
And now a GTX1060 has been spotted in Hong Kong... $300-380 is the speculated price range... This is from a HardOCP forum thread... The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.