Several questions about hardware components
21 Jan 2013 18:21:25 UTC
Topic 196758
(moderation:
I currently have one Gigabyte Windforce GTX 670 which I have overclocked slightly past what it came with.
NOTE: I CANNOT pick up another Windforce 670 for this particular case (CM Storm Scout I), it will not fit. I don't even know how I wiggled the first one in.
1.) How significant of a performance increase would I see with these computations if I added another 670 in SLI?
2.) Should I enable my GPU overclocking profile 24/7 while it's working?
3.) The Windforce 670 stays extremely cool even when running Heaven benchmark at max. Undoubtedly there will be heat issues with a 2nd 670. Does anyone have any experience with this particular 670 and finding a cool-running partner card for SLI? The Windforce can also not be moved out of the first PCIe slot, not enough room for the shroud in the second slot.
Last resort: Do I just need to get a bigger case for better airflow + more room for 2nd Windforce 670?
I currently have an i5 3570K, overclocked to 4.2Ghz.
1.) I've seen reports of some CPU's having too many errors, and their computations thrown out. How can I avoid this with this processor?
2.) Would it be more effective to underclock the CPU to reduce temps/stress, but thereby slowing down processing?
3.) Is there a particular server-class CPU that is cheap (under 300) that would be better suited for projects like SETI@Home?
Thank you guys, you're awesome.
Several questions about hardware components
)
Just a tip, the I5 (i have few with 2x670) are a low end CPU, so leave all the cores you could to feed the GPU´s, who many, depends on your particular system, but there is a easy way to find.
Use a program like GPUz and look the GPU usage, increase/decrease the number of avaiable core until you find the best value (the one that produces the best GPU usage %)
In this I5 with a 690/670 i use 25% only, no CPU work is done when running 6 E@H GPU WU, but each system works diferent.
About the heat, the 670 is extreamely cold, you don´t be in any problem by ussing 2 on a single case, even with a normal 670 (not windforce) Specialy in E@H, in SETI if you use optimized apps and runs 2 or more WU at a time you just need to increase the fan speed a little (i use evga precision to do that). BTW SETI does not have the CPU usage limitation of E@H.
RE: Just a tip, the I5 (i
)
I'm sorry I'm not sure I understand "Leave all the cores you could to feed the GPU". How can I do this specifically?
Thank you
RE: 1.) I've seen reports
)
The design and test of your CPU should, in principle, assure correct operation so long as you operate it within expected limits (mostly of temperature and power supply voltage).
Individual samples often have some margin to these requirements--which people take advantage of by overclocking. However the possibility to overclock is not uniform from sample to sample of even the exact same part--so it is a trial and error process. A problem for both the manufacturer and the overclocker is that there is no such thing as a complete speed test--so samples slip out which won't really run at specified conditions correctly if the CPU is requested to run an untested operation which unluckily relies on a path that has a sub-critical defect on it lowering the speed of that path compared to the rest of the chip compared to the usual production. Overclockers can fall in the same trap--relying on some test that internet folklore claims to be reliable (e.g. Prime95). It can still happen that your sample is slower running Einstein (possibly just for some WU) than on whatever test you use.
So the safest thing to do is to assure your case and HSF provide good cooling and well-regulated power, and avoid overclocking. Plenty of us overclock, some a little, and some a lot. As a practical matter, with good cooling and moderately higher power supply voltage than standard, many, many parts can be run on Einstein at clock rates appreciably above stock without detectable increase in error rate.
The biggest thing, for you and for the project, is to make sure if you do overclock that you watch for error, and be willing promptly to undo the overclock (or at least relax it), on any indication that your host is making appreciable errors.
If this sound more trouble than it is worth--don't overclock. If it sounds like fun--step right up.
Aside from overclocking--do take some care to assure case and CPU heatsink/fan air flow does not come to suffer substantial blockage from dust buildup. Lots of us have been stunned at what we find in cases we have not opened for a while.
RE: RE: Just a tip, the
)
My Boincs show in portuguese, so please forgive me if my translations is bad.
Go to computing preferences
Must show something like: In multiprocessed system use .... % of the processors.
that .... is the number (defoult = 100%)
I need to limit to 25% in this 2x670 I5 for example (4GPU WU+1 CPU task), but each sistem have it´s optimal value, you need to try each one (25,50,75 or 100% in the I5 case). But be aware that´s is completely diferent on an I7 hosts.
Tyler, Just checked your
)
Tyler,
Just checked your BRP cuda times and those are quite nice indeed.
http://einsteinathome.org/host/6486938/tasks&offset=0&show_names=1&state=3&appid=19
To see just how good your quad core will work maybe you should set your Einstein preferences to only run BRP cuda's and check the times after you finish all the GRP and GW S6's
Do you have your preferences set (GPU utilization factor of BRP apps) at 0.5 ?
Your processor OC and video card set OC'd looks to be running nice.
I have a few quad cores of older versions so they can't be OC'd but I run them set at 0.5 and run cuda X2 along with one of the other Einstein tasks and at the same time 2 Test for Theory's with no problems and I never leave a core empty just to add processor to this like I know many members do.
You can also check other members hosts on the stats page if you want to compare times and the thread here with times on all the different cards and add yours once you run it for a week or 2 since the RAC takes that long to hit the number.
You can also download GPU-Z to look at all the numbers for your video card and the temp while you have it running the cuda's and other tasks (the best card I have is the GeForce 660Ti superclocked so I never have tried a 670)
Welcome to the project.
RE: The design and test of
)
Great information, thanks. I could run the 3570K at 4.5Ghz but after reading your post I've decided to drop it to 4.2 just to be safe. My cooler seems to do great though (CM Hyper 212). I would be a bit sad if I started returning error results.
RE: Tyler, Just checked
)
Interesting. So I need to understand this, when you run the CPU at .5 that means you're allowing 50% utilization, correct? And when you set the GPU utilization factor of BRP apps at .5, it's doing the same thing right? I'm pretty sure I have the settings set to 100% as I'm at work most of the day so I just let it run as fast as it wants.
RE: Interesting. So I need
)
No, the setting for the GPU is to tell Boinc how much of the GPU will be utilized by one task, when you set it to 0.5 you tell Boinc that the task will us half of the GPU and thus Boinc will start 2 tasks to fully utilize the GPU. And with that follows the possibility to set 0.33 for 3 task and 0.25 for 4 task and so on...
The setting exist because often one GPU-task can't make full use of the GPUs power and running 2 parallel is often more efficient.
Remember to watch the memory usage when increasing the number of parallel task so you don't run out of video-RAM. GPU-z is a good program for checking on the status of a GPU. When GPU-z shows >90% GPU-load then the card is fully utilized and adding more parallel tasks won't gain anything further.
And to answer your first question from the initial post:
SLI won't help at all. But adding another GPU will definitely help, if it's another 670 you should expect about double the performance form only running one. That is if you CPU and other system components is up to the task of feeding 2 670s.
Could someone have the answer
)
Could someone have the answer and explain or give a clue how to configurate?, why the GTX690 is so ineficient in E@H? I have few 690/590 and runs 2 WU (4 on each GPU - CPU Usage >85%) and comparing for an 2x560 host the diference is incredible against the 690. The same not happens on SETI, GPUGrid or Collantz where a 690 is equivalent to almost 2x580 (i have few too) in processing power.
To obtain a little gain i was forced to divide the 2x690host in 2 690+670 diferent hosts.
RE: Could someone have the
)
The Einstein BRP4 CUDA application uses significant PCI-E bandwidth and is a hybrid type application that does processing on both the CPU and GPU. The GTX 690 is equipped with a PLX PEX 8747 which is a 48-lane PCI-E 3.0 switch. The PEX provides 16 PCI-E 3.0 lanes to each of the two GPUs. This will give the GPUs ample bandwidth to the PEX but the pipe between the PEX and CPU or NB PCI-E controller is also important. Since the two GPUs have to share the same pipe to the CPU or NB PCI-E controller, ideally you would install the cards in dedicated slots of at least PCI-E 2.0 x16. PCI-E 3.0 x16 would likely be the most optimal solution and will give roughly double the bandwidth of a PCI-E 2.0 x16 slot.
If I were to setup two GTX 690s for Einstein@Home, I would go with an x79 board and Intel 3820 processor. This particular CPU has 40 PCI-E 3.0 lanes available so that each 690 can have its own PCI-E 3.0 x16 slot.