Message boards :
Number crunching :
Average RAC for GTX 670/680.?
Message board moderation
Previous · 1 · 2
Author | Message |
---|---|
arkayn Send message Joined: 14 May 99 Posts: 4438 Credit: 55,006,323 RAC: 0 |
|
W5DMG - Dave Send message Joined: 19 May 99 Posts: 155 Credit: 33,162,251 RAC: 0 |
Back when the GTX-200 series was 1st released, I had always thought that the main difference between the 260/70/80 was the clock speed of the GPU, the memory interface and the amount of cuda cores. So as long as the amount of cores increased the output increased. But this is not true with the 600 series, as we now have 3 times the cores and not 3 times the output. |
shizaru Send message Joined: 14 Jun 04 Posts: 1130 Credit: 1,967,904 RAC: 0 |
Hi there. Thanx Vyper. This kind of number crunching is hard to come by in the Number Crunching forum:D So Kepler finally shows it's true colors! Here's what your numbers are telling me: GTX 670 (GK104) has ~50% more RAC than a GTX 560Ti (GF114), ie the chip it "should" have replaced. "At 294mm2 GK104 is not Big Kepler... This is roughly 89% the size of GF114 [~330mm2], or compared to GF110 a mere 56% of the size. Inside that 294mm2 NVIDIA packs 3.5B transistors thanks to TSMC’s 28nm process, only 500M more than GF110 and largely explaining why GK104 is so small compared to GF110. Or to once again make a comparison to GF114, this is 1050M (53%) more than GF114..." AnandTech And both have a TDP of 170W. So you get (roughly) 50% more Seti performance in the same power envelope for a launch price that is/was $150 higher. The numbers are finally starting to fall into place. (670 launced at $400, 560Ti at $250) Edit: This isn't gonna make sense to most, but I think "crooked" is norwegian for DIYers/modders/"hackers":) |
W5DMG - Dave Send message Joined: 19 May 99 Posts: 155 Credit: 33,162,251 RAC: 0 |
Back when the GTX-200 series was 1st released, I had always thought that So is my info correct or incorrect..? The 270 out performed the 260 and the 280 out performed the 270 in SETI ouput because of what..? Not cores..? Duh ok there is more math to it.. :( |
shizaru Send message Joined: 14 Jun 04 Posts: 1130 Credit: 1,967,904 RAC: 0 |
Let's start over. First I'm gonna open Pandora's box and say (based on specs) I can't see how a 670 will do any better at Seti than a 660Ti. Their compute performance is an exact match. It's highly likely Seti crunchers need to forget about the 670 now that the 660 Ti is out (as far as Seti is concerned, NOT gaming) Second, only compare same generation cards when doing core number and shader speed comparisons. For the 6xx series nVidia doubled the cores for the same amount of compute performance. Third. There is no 270. There's a 275 that was actually better (at Seti) than the 280. So you can't go by model numbers either:) Questions: Are you dead-set on an RAC of 100k? Do you pay your own power bills? What's your budget for cards, power supply and case? |
W5DMG - Dave Send message Joined: 19 May 99 Posts: 155 Credit: 33,162,251 RAC: 0 |
I am not dead set on 100k, but certainly close to it. Yes I pay my power bills and usually early.. :) I have a CoolerMaster HAF case with 3 200mm fans. I also have a CoolerMaster 1000 watt power supply. Actually your wrong, there was a GTX-270. It came out 1st and the 275 replaced it. I have been keeping up with nVidia forever, because I am a gamer too. Finally I agree the 660Ti is almost equal to the 670. The only minor difference between the 2 is the memory interface. The memory interface deals with bandwidth, the higher the better. But that higher bandwidth is not needed for SETI performance. The topic 670/680 was because I had read somewhere here that the 670 could produce 100k by itself. But I just bought 2 EVGA superclocked 560 Ti's and am close to buying another set of them. At present I have yet to run any GPU's with the modified app. But that is about to change and I will be posting a new thread next week on how to use Lunatics modified app and what I need. Main reason I have yet to switch to Lunatics is, I have not seen any instructions in laymens terms. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13727 Credit: 208,696,464 RAC: 304 |
Main reason I have yet to switch to Lunatics is, I have not seen any instructions in laymens terms. You use something like CPU-z & GPU-z to determine exactly what type of processors you have & the instruction sets they support, you run the installer, select the most appropriate option for the CPU & GPU you have, select whether to run Multibeam, Astro Pulse or both, let it install. Grant Darwin NT |
tbret Send message Joined: 28 May 99 Posts: 3380 Credit: 296,162,071 RAC: 40 |
You didn't say if you were going to put all four of your contemplated cards in the same case on the same power supply. If you are, you might want to pay particular attention to the power draw of four superclocked GTX 560Ti's under "gaming conditions" or whatever heavy-use application you choose. I don't know one way or the other; I'm just concerned that it might be an issue. |
W5DMG - Dave Send message Joined: 19 May 99 Posts: 155 Credit: 33,162,251 RAC: 0 |
Well actually as I stated in another reply that I have a MSI mb that has 5 PCI-X slots, supports 4 double spaced cards as long as its a tower case. That case and mb has a Corsair 650 watt power supply I also have a newer Gigabyte mb that supports 2 PCI-E slots. It is the one with the 1000 watt psu. I plan on putting 2 560's in each system. Originally when I bought the MSI and Corsair I had 2 GTX 260's. Then I purchased a CoolerMaster 1000 watt and added a GTX-465. So I feel I will have enough power for both. Both mb's have AMD cpu's and proper cooling has always been important to me. I always plan on overclocking, but never really see any noticable improvemants. My main sytem now has a water cooled cpu, but water cools nothing else. Noisy computers are not an issue with me.. |
©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.