my upgrade

Message boards : Number crunching : my upgrade
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4

AuthorMessage
Profile zoom3+1=4
Volunteer tester
Avatar

Send message
Joined: 30 Nov 03
Posts: 65709
Credit: 55,293,173
RAC: 49
United States
Message 1022395 - Posted: 4 Aug 2010, 4:56:29 UTC - in response to Message 1022392.  

What if you would just turn the psu into always on ? I remember reading about that when I planned to turn one of my old power supplies into a poor mans lab psu. It only takes a power resistor. Here is the link to the page. You might have to dig a little. http://www.wikihow.com/Convert-a-Computer-ATX-Power-Supply-to-a-Lab-Power-Supply


That would work or So would this below and I like the one below better.


http://www.performance-pcs.com/catalog/index.php?main_page=product_info&cPath=34_186&products_id=3163
The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's
ID: 1022395 · Report as offensive
Jon

Send message
Joined: 12 Aug 09
Posts: 157
Credit: 139,063,241
RAC: 0
United States
Message 1022396 - Posted: 4 Aug 2010, 4:58:01 UTC

Here is it. Looks like pin #3 was the problem.



Jon
ID: 1022396 · Report as offensive
Profile zoom3+1=4
Volunteer tester
Avatar

Send message
Joined: 30 Nov 03
Posts: 65709
Credit: 55,293,173
RAC: 49
United States
Message 1022398 - Posted: 4 Aug 2010, 5:06:39 UTC - in response to Message 1022396.  

It's like I thought, It seems that the Lian Li adapter has wires that can't transmit the power the primary power supply is trying to feed to the motherboard. Of course the Lian Li is made by a PC case maker and not a power supply maker which would use shorter wires and hopefully with more copper in those wires. The last cable with the switch would be ok I'd think, As there's no adapter cable to overload.
The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's
ID: 1022398 · Report as offensive
Profile Sutaru Tsureku
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 1022416 - Posted: 4 Aug 2010, 8:45:31 UTC - in response to Message 1022374.  
Last modified: 4 Aug 2010, 8:49:20 UTC

AFAIK, at least PCIe 2.0 x8 (or PCIe 1.0 x16) speed is needed for well CUDA crunching.
So if Jon's GIGABYTE mobo support this also he could let run 7 GTX480. WOW!


EDIT: But maybe faster PCIe speed is needed if 2+ WUs/GPU?

I crunch for S@H with a 16x and a 4x pci-e slot as I have a GTX295 in each, So I don't know where You got the 8x requirement, As 4x or possibly even 2x would do just fine.

I didn't said it wouldn't work with less speed.

As I build my GPU machine (for ~ 1 1/2 years) I looked around in differenct forum and I saw, that you have best/max CUDA performance if you have at least PCIe 2.0 x8 or PCIe 1.0 x16 slot for each GPU.

Also a member made a test here in the forum. Because of the max limit of one year in the forum search, I can't find it.
He tested his GPU in all slot speeds and saw that the GPU performance went down on slower slots.
IIRC, PCIe 1.0 x4 had ~ 10 % less, and PCIe 1.0 x8 ~ 3 % less performance. (But this values could be wrong, as I'm an old man ;-)
But for sure I remember, at least PCIe 1.0 x16 or PCIe 2.0 x8 needed for max/best CUDA performance.

If I look in TaskManager if my system crunch CUDA, I see that the GPU get support from the CPU. Also faster system RAM decrease the CUDA calculation time. So the CPU and system RAM is important for CUDA. So also the PCIe slot.

Here at SETI@home my GPU get ~ 45 secs. support/~ 10 mins. WU.
If I compare with MilkyWay@home, my GPU make all WUs in ~ 10 mins. there, the WU get ~ 8 secs. support from the CPU.
So there are projects which benefit more or less from faster PCIe speed.
ID: 1022416 · Report as offensive
Jon

Send message
Joined: 12 Aug 09
Posts: 157
Credit: 139,063,241
RAC: 0
United States
Message 1022580 - Posted: 4 Aug 2010, 21:57:42 UTC
Last modified: 4 Aug 2010, 22:01:28 UTC

Those flexible adapters are not really working at the moment. Put 4 cards right on the mb and it works.

It's drawing ~1050w doing 4 cards at 2wu's/gpu only. gpu usage is 90%.

Jon
ID: 1022580 · Report as offensive
Profile soft^spirit
Avatar

Send message
Joined: 18 May 99
Posts: 6497
Credit: 34,134,168
RAC: 0
United States
Message 1022597 - Posted: 4 Aug 2010, 23:29:38 UTC - in response to Message 1022580.  

Those flexible adapters are not really working at the moment. Put 4 cards right on the mb and it works.

It's drawing ~1050w doing 4 cards at 2wu's/gpu only. gpu usage is 90%.

Jon


4 of those things is still a beast. How are your temperatures doing?

Janice
ID: 1022597 · Report as offensive
Jon

Send message
Joined: 12 Aug 09
Posts: 157
Credit: 139,063,241
RAC: 0
United States
Message 1022605 - Posted: 5 Aug 2010, 0:32:00 UTC - in response to Message 1022597.  

Temps of the gpu are in the upper 70's. cpu is 60. Tomorrow i'll put it in a different room, under the a/c vent to cool it off!

Jon
ID: 1022605 · Report as offensive
Profile Fred J. Verster
Volunteer tester
Avatar

Send message
Joined: 21 Apr 04
Posts: 3252
Credit: 31,903,643
RAC: 0
Netherlands
Message 1022679 - Posted: 5 Aug 2010, 12:15:24 UTC - in response to Message 1022605.  
Last modified: 5 Aug 2010, 12:25:15 UTC


Temps of the gpu are in the upper 70's. cpu is 60. Tomorrow i'll put it in a different room, under the a/c vent to cool it off!

Jon



But you've got a Hell of a Crucher and can heat a (small) room, too, ;-)

I've often wondered, why PSU cables are so small/thin, compaired to the mains cable, which carries 1/10 of the Input-current, although most PSU's > 650Watt have 4x 12V @ 17-20A. (When in the U.S.A and have 120Volts)
Still 2.5x main-current , while cables are mostly thinner then their mains supply. Not to mention connectors/splitters/switches.
(You can welt with a 100 Amps!)
ID: 1022679 · Report as offensive
ToxicTBag

Send message
Joined: 5 Feb 10
Posts: 101
Credit: 57,197,902
RAC: 0
United Kingdom
Message 1022730 - Posted: 5 Aug 2010, 15:30:38 UTC

Good to see you making progress with your rig Jon, must admit it certainly grabbed my attention! Can i ask why do you think the riser cables did not work or are they ok now?
Best of luck
ID: 1022730 · Report as offensive
Profile hiamps
Volunteer tester
Avatar

Send message
Joined: 23 May 99
Posts: 4292
Credit: 72,971,319
RAC: 0
United States
Message 1022839 - Posted: 5 Aug 2010, 22:50:14 UTC - in response to Message 1022396.  

Here is it. Looks like pin #3 was the problem.



Jon

Well I won't buy one of them, seems they didn't even need the extra wires could have made a strait adapter with just the Black and green jumpered off and it would be safer. Think I will stick to my home made jumper.
Official Abuser of Boinc Buttons...
And no good credit hound!
ID: 1022839 · Report as offensive
Profile zoom3+1=4
Volunteer tester
Avatar

Send message
Joined: 30 Nov 03
Posts: 65709
Credit: 55,293,173
RAC: 49
United States
Message 1022845 - Posted: 5 Aug 2010, 23:05:38 UTC - in response to Message 1022839.  
Last modified: 5 Aug 2010, 23:07:10 UTC

Here is it. Looks like pin #3 was the problem.



Jon

Well I won't buy one of them, seems they didn't even need the extra wires could have made a strait adapter with just the Black and green jumpered off and it would be safer. Think I will stick to my home made jumper.

Psu connector is below and the adapter with the green and black wire is above, th adapter has all those wires for the 1st psu to transmit the power to the motherboard, The green and black wires from the adapter merely tell the 2nd psu to turn on or to turn off in unison with the 1st psu, nothing more, As to some saying not enough ground wires, Their nuts, There are plenty of those, The adapters wires are just not thick enough to do the job and so smoke and a fire break out, Clearly I'd avoid a cable like this as It's made by a case manufacturer and not by a power supply manufacturer, Ultra made their wires short and I'd hope of a thicker gauge of wire,
(a smaller gauge number = a larger gauge wire = more carrying capacity).

What looks like thicker wire may just be thicker insulation over thinner and unsafe wires.
The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's
ID: 1022845 · Report as offensive
Profile hiamps
Volunteer tester
Avatar

Send message
Joined: 23 May 99
Posts: 4292
Credit: 72,971,319
RAC: 0
United States
Message 1022858 - Posted: 6 Aug 2010, 0:02:13 UTC

Probably a cheap connector that was loose. Doubt it was the wire....
Official Abuser of Boinc Buttons...
And no good credit hound!
ID: 1022858 · Report as offensive
Profile Bill Walker
Avatar

Send message
Joined: 4 Sep 99
Posts: 3868
Credit: 2,697,267
RAC: 0
Canada
Message 1022860 - Posted: 6 Aug 2010, 0:17:33 UTC - in response to Message 1022858.  

Probably a cheap connector that was loose. Doubt it was the wire....


Agreed. The heat was apparently strongest right where the pin goes into the socket. Could be a poor fit, or foreign material trapped in the interface, or some manufacturing defect that resulted in only a very small contact area between the pin and socket.

ID: 1022860 · Report as offensive
Jon

Send message
Joined: 12 Aug 09
Posts: 157
Credit: 139,063,241
RAC: 0
United States
Message 1022882 - Posted: 6 Aug 2010, 1:54:47 UTC - in response to Message 1022730.  

Good to see you making progress with your rig Jon, must admit it certainly grabbed my attention! Can i ask why do you think the riser cables did not work or are they ok now?
Best of luck


I tested one when i got them, the only good one. I found out that 2 of the 3 adapters were bad! I have more coming so I'll continue to go ahead with this project.


Jon
ID: 1022882 · Report as offensive
Profile zoom3+1=4
Volunteer tester
Avatar

Send message
Joined: 30 Nov 03
Posts: 65709
Credit: 55,293,173
RAC: 49
United States
Message 1022929 - Posted: 6 Aug 2010, 5:49:11 UTC

Jon If You need a second psu there is the VisionTek 450w psu on amazon, It mounts in a 5.25" drive bay and has 40A on the 12v rail, Plus It's designed to work with a primary psu.


The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's
ID: 1022929 · Report as offensive
Profile evilspoons
Avatar

Send message
Joined: 30 Jul 99
Posts: 50
Credit: 8,469,307
RAC: 4
Canada
Message 1024055 - Posted: 9 Aug 2010, 22:36:47 UTC

FWIW, the battery connectors on a lot of hobby-level radio control cars are similar in design to the ATX mobo connector("Molex" in PC lingo, "Tamiya" in RC lingo).

The RC guys hate the "Tamiya" connector because it melts under high-current applications. This may be what's happening.

As to when someone mentioned above that 0V/ground carries no current, that should be rephrased to "SHOULD carry no current." You can get awful ground loops that destroy things if you muck around too much with multiple ground planes (i.e. more than one power supply). Personally, I'd be looking for an adapter that does something to tie together the grounds of these PSUs (obviously without making them fight each other).

Just my two cents. By the way, this thing is terrifying (and awesome).
ID: 1024055 · Report as offensive
Profile soft^spirit
Avatar

Send message
Joined: 18 May 99
Posts: 6497
Credit: 34,134,168
RAC: 0
United States
Message 1024059 - Posted: 9 Aug 2010, 23:08:46 UTC

I theenk you need a beeger box

re: grounds..

It should be okay to strap the two power supply grounds, as long as they are not providing any shielding. Also be sure both units run from the same potential(same side of the circuit breaker panel) otherwise strange and sparky things could happen.

That grounding may be common to the case anyway. You should be in good shape.
Janice
ID: 1024059 · Report as offensive
Jon

Send message
Joined: 12 Aug 09
Posts: 157
Credit: 139,063,241
RAC: 0
United States
Message 1024546 - Posted: 13 Aug 2010, 13:43:44 UTC

I decided to stay with only 4 cards. I fixed the problem with adapters(sloppy soldering). The 5th card will work but only lasts for a few minutes and you might get a few wu errors. Maybe it's Windows or maybe motherboards are not ready for 5-480 cards!

Jon
Jon
ID: 1024546 · Report as offensive
Previous · 1 · 2 · 3 · 4

Message boards : Number crunching : my upgrade


 
©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.