Should I need just a single free CPU for each GPU? So if other tasks use 22 of the 24 I should be OK, correct?

I've set the usage down to 90% for the night, let's see what happens.

Mostly for higher end gpus you will need a cpu per gpu task. For some gpu tasks, you can only run one per card. For high end gpus under Windows you can sometimes run 2-3 gpu tasks. This is common with gtx 1080 and up. A gtx 1060 3GB will not run multiple gpu tasks without significant reduction in total processing speed.

Mostly you want to limit your BOINC cpu cores usage to no more than 90% of the available cores. This gives the system the extra processing without suspending BOINC when non-Seti processing is needed.

For an 8 thread machine, 75% seems to work.

Tom

---

A proud member of the OFA (Old Farts Association).

Some people say they need to reserve more to have the GPUs run normally. Right now it appears you are only running 1 GPU, I don't see any 570 results.

It looks like Einstein grabbed the 570.

I'm getting a lot of "suspicious pulse result" errors postponing SETI tasks. I restarted the Mac Pro to see if that helps.

---

I'm getting a lot of "suspicious pulse result" errors postponing SETI tasks.

Those seem to come and go with different Apple updates. A while back it was pretty common, so, I built a new AMD App that appears to solve it. While the App was still at Beta another Apple update came along and the problem basically disappeared. The New App was left to wither at Beta while Apple updates come and go. Restarting probably won't help, the next Apple update might, as well as the newer AMD App at Beta and Crunchers Anonymous. The one I'm running is the v3710 one at Crunchers Anonymous and I've never gotten the 'suspicious pulse' message. You might go over to Beta and see if the v3610 one there is any better, or install the one at Crunchers Anonymous, or, wait for the next Apple update. Hmmm, the v3610 (8.22 - opencl_ati5_mac) App has been at Beta since 17 Oct 2017, 23:48:14 UTC while people still report of getting "suspicious pulse result".

Kinda hard to test something if the subject swaps projects, like the 570.
Oh well, I guess I'll go back to trying to get the v0.99b1 Special App to work on a Mac.

TBar->

The links at Crunchers Anonymous were broken but are fixed as of Sunday. If that was your doing, thanks.

My Mac Pro with the two Radeons was running too hot. Hardware Monitor doesn't read GPU card temperatures, I found another app that did. It claimed the 570 was at 185 F! That may be dubious, for a short time it claimed an empty memory slot was at 248 F.

I kept the 570 in the dual-CPU and moved its 560 to the single-CPU Mac Pro that already had a 560. This is a better power balance, instead of one having 240W of GPUs (150 + 90), one has 150 and the other 180 (2 * 90). This also let me put all my cards back in the dual-CPU since it has to run Mojave and is supposedly a workhorse.

I'm having some stability issues with Mojave on the Mac Pros. I may revert the single-CPU to High Sierra since it doesn't have to run Mojave. I'll try the custom apps once I get these two Macs settled down.

---

Yes, someone PMed me and said the links were broken. Seems the Attachment process can't find the attachment directory. You have to post the links using a different method. It also seems I can't find most of the old packages, I'll have to look around a little more.

Most of the Mac tools don't give good GPU temps, the best way to see the temps is to run Windows or Linux and see what they show. I found that if I ran the I/O fan in Macs Fan Control up to around 2000rpm the GPUs stayed reasonable. Macs Fan Control works in Windows. Modern electronics are a marvel, for the last 10 years or so they automatically throttle if they run too hot. So, the way I look at it, if the completion times are normal then the card isn't throttling, and thus must not be too hot. If the times are slow, then boot to Windows and see if there is a problem. Back before there were Mac SETI Apps, I ran the Mac in Windows to run SETI. I haven't booted my Mac to Windows or Linux in a very long time, and my completion times are what are to be expected. Actually, if you compare my 570 times to others, it's doing very nicely. Unfortunately, the AMD High Sierra driver needs SSE4.2, which the 3,1 doesn't have, so, I run Sierra where the 570 works normally, and so far I've been able to compile the CUDA App in CUDA 9.1.

I found two reasonably priced GTX 1060s locally. One is a MSI GTX 1060 6GB, it's a single-fan shorter unit. The other is a Gigabyte GTX 1060 6GB, longer with dual fans. Is one better than the other for compute power and power consumption?

TIA,
Sam

---

Sam not quite on topic but I have a pair of 3GB 1060s.
One is a single fan EVGA the other is a dual fan Gigabyte. For crunching the dual fan runs much cooler.

---

Sam not quite on topic but I have a pair of 3GB 1060s.
One is a single fan EVGA the other is a dual fan Gigabyte. For crunching the dual fan runs much cooler.

Thanks, That's perfectly on topic and just the feedback I'm looking for.

---

Sam not quite on topic but I have a pair of 3GB 1060s.
One is a single fan EVGA the other is a dual fan Gigabyte. For crunching the dual fan runs much cooler.

Thanks, That's perfectly on topic and just the feedback I'm looking for.

I've always wondered what shortcuts were taken to produce those short cards. ;-)

Cheers.

As far as I can tell, they increase the surface area of the heatsink by making it wider and increase the fan blade diameter along with the base rpm the fan runs. At least based on my two examples with a EVGA 1060 6GB single fan card and a Zotac Mini 1070Ti card.

If you place the card out in the open, for example on the outside of the gpu stack, it has no problems with cooling or running at the top frequencies that GPU Boost allows.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

As far as I can tell, they increase the surface area of the heatsink by making it wider and increase the fan blade diameter along with the base rpm the fan runs. At least based on my two examples with a EVGA 1060 6GB single fan card and a Zotac Mini 1070Ti card.

If you place the card out in the open, for example on the outside of the gpu stack, it has no problems with cooling or running at the top frequencies that GPU Boost allows.

Thanks Keith. I intend to use this card on my dual-Xeon server mobo which runs without a case. The mobo only has a single PCI slot so the card won't be crowded and will be well ventilated. I'll try for the dual-fan but the seller hasn't responded, so there's a good chance I'll end up with the single fan.

---

Actually I have a Single fan GTX 1060 SC. It's the Hottest, Noisiest GPU I have, to keep it cool you need to run the fan up to around 85% in open air. Fortunately, I only have one and it now runs on the Mining machine. When it was in the Mac it would Scream on some of the WUs, on others it would race up and down, very annoying. That's how you can tell a card is running Hot in a Mac, it Screams. Amazingly, I have the 'good' model. The SC model has a much better Heatsink than the plain single fan GTX 1060 which has a heatsink similar to the single fan 950 SC. The single fan 950 is another 'Hot' card with many of the same features. I'd stay away from both, the dual fan models are much more pleasant. As I said before, the Gigabyte 3 GB 1060 has a backplate, and any card with a backplate is too tight to place in a tight dual card arrangement, however, they work fine in the Mining machine nicely spaced apart. You can take the backplate off, but it removes half the cover, even though, it still seemed to work fine in the Mac.... and didn't scream.

I too have a single fan 1060. Used to have several. It’s the 6GB model so has an extra CU vs the 3GB models. More heat to handle.

They aren’t any wider than the dual fan cards. Just shorter. I don’t believe they took any “shortcuts”, I’m pretty sure that’s the reference nvidia PCB. It’s the same story on the RTX 2070, the reference nvidia PCB is a short board as well. EVGA didn’t start doing the triple slot single fan design until the Turing cards.

As for performance, my card is the EVGA “SC” model. Which means it has a better heat pipe and fin heat sink layout. It runs a tad warmer than the others, but nothing outrageous. I leave it on 50 % fan speed, inside a closed case with 5 other cards and the temps stay in the 50s and 60s under seti. Proper case ventilation does wonders.

Some models of the EVGA 1060, especially the 3GB models that are short, have a black coated cast aluminum heatsink that likely gives worse performance.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

I have a GTX 1060 3 GB and a GTX1050 Ti 4GB on two Windows PC. Both are very silent. The Task Manager on the 8.1 PC with the 1050 Ti does not show any GPU, while the Omen Command Center in a HP desktop with the 1060 GTX shows its temperature and that of the CPU, a Ryzen 5 1400. It happens often that the temperature of the CPU is greater than that of the GPU.
Tullio

---

nvidia reference GTX 1060 vs EVGA short GTX 1060


Good vs Bad EVGA heatsink on short cards

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

That's what I mean Ian, both those cards have very small PCB's compared to the full length PCB's in my ASUS Dual OC GTX 1060 3GB's and Gainward GeForce GTX 1060 3GB's, so what has been cut back on to produce such small PCB's?

Cheers.

again, nothing was cut back. that is the "stock" "default" configuration as designed by nvidia themselves.

the more appropriate question, is "what was added" by the other AIB manufacturers. a full length PCB is not the reference design. as shown in the pictures.

I have no doubt that the EVGA with it's shorter cooler, still outperforms the stock NVIDIA blower design since they are using a better/larger fan and cooler.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

More coolness? :-D

Cheers.

which kind? ;P

I had some of those ASUS dual cards in the past, and they performed pretty well. a little better than the short 1060s, IF you have the short 1060 with the good cooler and not the bad cast aluminum one. the 1060 doesnt really produce much heat, and uses 100w or less in seti, so it's not really all that difficult to keep cool in the first place.

the short 1060 definitely has their place. many guys doing SFF builds like them because thermals and noise are totally acceptable and they can fit in some really small cases.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

The OP has a couple of Macs. For now he says he won't use it in his Mac, however, if he ever does go back to High Sierra he will most likely change his mind.
Considering I have a 'Good' version, bought new from NewEgg, I can report if he does try it in his Mac, it will most likely Scream and be a bad experience.
If you turn up the IO fan any higher, then the IO fan will Scream, so, the best choice for someone with a Mac, is to stay away from this particular model. There are Tons of other dual fan 1060s out there,
why buy one you will be unhappy with? My advice for anyone with a Mac is to stay away from the Single fan 1060, I have already tried it with a New card, it runs Hot.