Thinking of upgrading a cruncher. Best @ 3930k priceline ?

[Sunny129]

Well, one of my 470's has an odd problem. It gets over 100c even at 100% fan, AND I already cleaned out the dust.

So I bought 2 560ti's for the new box.

i wish i were in a position to just throw money at my DC problems too

...joking aside, is cleaning out the dust the extent of your GTX 470 maintenance? if so, i strongly suggest removing the fan/heatsink and re-seating it w/ some quality thermal compound (Arctic Silver 5, Noctua NT-H1, or the like). if its an OE cooler that's never been removed before and the card has been running at those high temps for some time now, there's a good chance that the thermal compound applied by the manufacturer on the assembly line has caked over and lost its effectiveness by now...plus they probably used some cheap thermal grease anyways...that's what all the manufactures use.

at any rate, i'm not sure if its worth the hassle (perhaps that's why you said f*ck it in the first place, and just bought replacement GPUs). either way, its probably worth at least some of your time to try and fix the GTX 470 so you can sell it and recoup some money.

 

[blckgrffn]

You could send that 470 to me instead of that check, I'll try to get it to work

I did ship that XP-64 out earlier today, DC# coming to an PM inbox near you

In any case, I would recommend VMware Player over virtual box, it is likely to get you better performance than Virtual Box (it uses the commercially supported Workstation engine, which needs to have good performance if there is any reason for it to exist at all). It is even easier to use, IMHO.

Also, you might run into some upper bounds w/regards to the number of virtual CPUs you can give to a VM? Ah, I see Virtual Box totally trumps VMware player in that regard. Scratch my earlier suggestion and stick with Virtual Box, I guess

 

[Uppsala9496]

This hardware upgrade thread was peer pressuring me into doing something. So I managed to install an antec kuhler 620 in a thermaltake V3 case. Fortunately I have a matx board. Radiator/fan went in the far top fan placement (120mm or 140mm). Couldn't line it up with the 120mm fan holes since the radiator would hit either the ram or some caps. Managed to offset the placement to get it in. Wow is it significantly quieter than the hyper 212+ (dual fans) that I had. I'm sorely tempted to pick up the 620 about once a month to replace all of my heatsinks. Installed in 2 of 5 machines right now.

 

[Markfw]

OK< Now I feel really stupid.... The problem was that the fans on the twin frozer were frozen. I played with them, and then they started working. Now running 72c. Crap, just spent over $400 that I didn't really need to . At least come xmas for the race I will have 2 more boxes to turn on.

 

[blckgrffn]

This hardware upgrade thread was peer pressuring me into doing something. So I managed to install an antec kuhler 620 in a thermaltake V3 case. Fortunately I have a matx board. Radiator/fan went in the far top fan placement (120mm or 140mm). Couldn't line it up with the 120mm fan holes since the radiator would hit either the ram or some caps. Managed to offset the placement to get it in. Wow is it significantly quieter than the hyper 212+ (dual fans) that I had. I'm sorely tempted to pick up the 620 about once a month to replace all of my heatsinks. Installed in 2 of 5 machines right now.

Wow, really? I've been having great like with the 212+ being silent in my installations, even with a single fan!

Of course, I don't do extra volts or anything, so that might make the difference. My OC's are very mild.

:thumbsup: for you though, and that has me more determined to try H2O cooling than ever...

OK< Now I feel really stupid.... The problem was that the fans on the twin frozer were frozen. I played with them, and then they started working. Now running 72c. Crap, just spent over $400 that I didn't really need to . At least come xmas for the race I will have 2 more boxes to turn on.

Blame it on the surgery I would...

 

[Markfw]

Well, I am taking 650 mg of percoset evry 8 hours for pain.

 

[somethingsketchy]

Well, I am taking 650 mg of percoset evry 8 hours for pain.

That'll do it. That's some potent stuff you're taking (or so I have been told). Nonetheless at least the GPU is usable again. That's a bonus.

 

[Sunny129]

hey Mark, which specific model GTX 560 Ti's did you get?

Wow is it significantly quieter than the hyper 212+ (dual fans) that I had.

were you using "inexpensive" fans or a not-so-well ventilated case by any chance? i've got a 212+ in one machine and a 212 EVO in another, both w/ dual Scythe SY1225SL12LM-P fans in push/pull. in the machine whose case has slightly less efficient airflow and ventilation, i run the fans @ ~1000rpm, and the CPU (a 1090T @ 3.7GHz, 1.4v) stays under 50°C under 100% load 24/7. in my other machine (a 1090T @ 3.2GHz, 1.375v), the fans only have to run at ~800rpm to keep the CPU below 50°C under 100% load 24/7. they're both virtually inaudible to me, and i'm pretty stubborn about having a silent machine.

 

[Markfw]

2 of these :
http://www.newegg.com/Product/Produc...82E16814127565

I don't suppose you know how much more juice than a 460 these take for power ? (wattage)

 

[Diogenes2]

Anand shows it drawing 26 more watts at full load..

http://www.anandtech.com/show/5048/evgas-geforce-gtx-560-ti-2win-the-raw-power-of-two-gpus/3

 

[Sunny129]

Anand shows it drawing 26 more watts at full load..

http://www.anandtech.com/show/5048/evgas-geforce-gtx-560-ti-2win-the-raw-power-of-two-gpus/3

this comparison will have to do for now Mark...just take into account your GTX 560 Ti's hefty factory OC, and adjust your power draw estimate accordingly. while i know that my total system power draw w/ the GTX 560 Ti under ~90% load and the 1090T CPU under 100% load is ~320W, my EVGA GTX 560 Ti is also factory OC'ed (to 900mhz), and its default voltage is slightly higher than the reference voltage, so like yours, it probably draws a bit more power than a reference GTX 560 Ti. also, i have no had a chance to measure my system's power draw with only the GTX 460 installed, so i don't yet know the difference in power draw between the two under load.

 

[Markfw]

Well, I think I figured this out. I have a 950 @ 4 ghz and 2 460's that is using 540-550 watts from the wall at full load. THAT system has a AX1200, that I could remove and put the 850 in with room to spare. And that AX1200 can handle anything.

 

[blckgrffn]

Well, I think I figured this out. I have a 950 @ 4 ghz and 2 460's that is using 540-550 watts from the wall at full load. THAT system has a AX1200, that I could remove and put the 850 in with room to spare. And that AX1200 can handle anything.

Haha, I am just hoping that you remember that you've ordered all of these parts when they start showing up!

At least you'll have this thread to look back at

 

[VirtualLarry]

Well, I think I figured this out. I have a 950 @ 4 ghz and 2 460's that is using 540-550 watts from the wall at full load. THAT system has a AX1200, that I could remove and put the 850 in with room to spare. And that AX1200 can handle anything.

I'll have to borrow the K-A-W from my friend (I stupidly gave him both of them, instead of keeping one for myself), and check out my new main rig. It's an X6 1045T @ 3.51, and two GTX460 1GB OC (715Mhz). I have an ABS 1050W PSU in there, 80Plus Silver. At least I'm not really wasting watts due to inefficiency.

 

[Markfw]

I'll have to borrow the K-A-W from my friend (I stupidly gave him both of them, instead of keeping one for myself), and check out my new main rig. It's an X6 1045T @ 3.51, and two GTX460 1GB OC (715Mhz). I have an ABS 1050W PSU in there, 80Plus Silver. At least I'm not really wasting watts due to inefficiency.

I had a 1090T, and I can't remember what it took, but it was a fair amount (thinking the system was over 500 watt from the wall)

As for efficiency, I wil never get anything but the gold, >90% efficient ones. like the AX series from Corsair.

 

[Sunny129]

Mark, i plugged my machine w/ the 1090T, the GTX 560 Ti, and the GTX 460 into the Kill-a-watt meter and got ~450W at the wall with everything under load. that's compared to ~320W at the wall before i put the GTX 460 in...so i suppose the GTX 460 itself draws approx. 130W under a ~90% load. keep in mind though that mine is factory OC'ed to a whopping 850mhz.

 

[Markfw]

OK, new rig up and running @ 4.5 ghz. The ppd is depressing.17,750 for the 560ti cards (one is having a problem right now @ 1k ppd) and 27k ppd for the smp unit. Time for some help on the unix vm thing !

Help my ppd please ! 24 gig ram and a 3930@4.5 ghz. Winxp 64.

 

[Sunny129]

OK, new rig up and running @ 4.5 ghz. The ppd is depressing.17,750 for the 560ti cards (one is having a problem right now @ 1k ppd) and 27k ppd for the smp unit. Time for some help on the unix vm thing !

Help my ppd please ! 24 gig ram and a 3930@4.5 ghz. Winxp 64.

you can't really use Folding@Home as a true gauge of any GPU's compute power b/c its the one DC project out there where CPUs are capable of outperforming GPUs. your GTX 560 Ti would tear through SETI@Home Multibeam tasks, to name one of many such projects/applications. one of your GTX 560 Ti's is capable of netting significantly more PPD than the ~18,000 PPD the two of them are doing now provided you choose the right project to participate in. granted, PPD is just a superficial performance gauge - i can't say that one GPU that earns an average of 200,000 PPD in Milkyway@Home contributes to science any more than a GPU that contributes 30,000 PPD in Folding@Home, b/c the units of scientific contribution, while both called PPD, are not an apple-to-apples comparison. the bottom line is that Folding@Home will always make GPU contributions look dismal when you have a powerful CPU w/ many physical or virtual cores b/c those CPUs (not GPUs) are capable of earning the highest F@H PPD values.

 

[Markfw]

I am dedicated to F@H, and curing cancer. Aliens can wait.

What I would like is help on creating the vm unix job for F@H. I for the 24 gig of ram and XP 64 specifically so I could tun this job, as I was told it will work much better than the smp client.

Oh, and the 17,750 is for EACH 560ti card. I don't really cards about numbers, they just indicate how much work is being done for F@H.

And lastly, my dad died of cancer, one of the reasons I am so dedicated to this project.

 

[DooKey]

Mark, just hit this link and it will walk you thru setting up virtual box and the vm.

http://hardforum.com/showthread.php?p=1037529274#post1037529274

 

[Sunny129]

fair enough. i agree btw - aliens can wait. that's why i primarily participate in S@H Astropulse (a newer application that searches for far more important things in addition to aliens, such as black holes, pulsars, and other exotic deep space objects and phenomena), and not so much S@H Multibeam (the original application that pretty much only looks for little green men). unfortunately for the nVidia crowd, their hardware is much more productive on Multibeam than it is on Astropulse.

anyways, i'm glad you're so dedicated to F@H and curing cancer. just don't expect your GPUs to outperform your CPU, like they will in just about every other DC project that allows GPU computing.

 

[blckgrffn]

fair enough. i agree btw - aliens can wait. that's why i primarily participate in S@H Astropulse (a newer application that searches for far more important things in addition to aliens, such as black holes, pulsars, and other exotic deep space objects and phenomena), and not so much S@H Multibeam (the original application that pretty much only looks for little green men). unfortunately for the nVidia crowd, their hardware is much more productive on Multibeam than it is on Astropulse.

anyways, i'm glad you're so dedicated to F@H and curing cancer. just don't expect your GPUs to outperform your CPU, like they will in just about every other DC project that allows GPU computing.

To be clear, they are doing more science, you are just getting points skewed by the accouting system

 

[Sunny129]

To be clear, they are doing more science, you are just getting points skewed by the accouting system

yes, i should have been more specific w/ my last comment - i meant to say that one shouldn't expect their GPUs to outperform their highly-threaded CPUs strictly in terms of PPD. in terms of scientific contribution, i completely agree w/ you.

 

[blckgrffn]

yes, i should have been more specific w/ my last comment - i meant to say that one shouldn't expect their GPUs to outperform their highly-threaded CPUs strictly in terms of PPD. in terms of scientific contribution, i completely agree w/ you.

Sorry, I wasn't trying to be pedantic and correct you - I just wanted to be clear for Mark that despite their "low" scores GPUs are doing a lot of work - work that is fairly narrow in scope, granted - but a lot of it.

 

[geecee]

If you folks live near a Microcenter, see this thread and get another $15 off your purchase:
http://forums.anandtech.com/showthread.php?t=2233062