nVidia GT240 success

If you think it might be a driver problem, see viewforum.php?f=79

Moderators: slegrand, Site Moderators, PandeGroup

nVidia GT240 success

Postby dogbert2 » Wed Nov 24, 2010 11:17 pm

Just wanted to report success using a nVidia GT240 card (XFX) with the following clock values (1GB of GDDR3 onboard):

GPU Clock: 650 (default is 550)
Memory: 800Mhz (no change)
Shader: 1650Mhz (default is 1340)

nVidia drivers 260.99 in use

Is the GPU 3 client that much better than the stable 6.23 GPU, or have additional optimizations been put into the new client? :lol:

I like the behavior of the new client, what should I be on the lookout for in terms of reporting problems?
dogbert2
 
Posts: 5
Joined: Wed Nov 24, 2010 11:08 pm

Re: nVidia GT240 success

Postby divery4eyes » Thu Nov 25, 2010 12:26 am

You still have a little headroom on the shader clock, about 200 Mhz(your miles may vary).
Drop the Gpu clock down to 600 and the memory down to ~710.
Depending on the case temps should be around 60 C.

My bad you asked about problems :lol:
over 30,000 werk units
"Everyone is in the same boat. Grab an oar, and paddle some WU's." -RAH

Still GPU folding with 8.2 teraflops
divery4eyes
 
Posts: 84
Joined: Tue Jun 03, 2008 10:10 pm
Location: off of I40 in Arkansas

Re: nVidia GT240 success

Postby bruce » Thu Nov 25, 2010 1:39 am

Welcome to the foldingforum, dogbert2.

For the GT240, you can choose either the 6.23 client or the 6.32 client. Many projects are limited to the g80/200 series hardware and they typically use FahCore_11. Performance will be identical since you'll be using the same version of FahCore_11. The 6.32 client is an upgrade which adds support the 400/500 series GPUs (aka "Fermi"). There are new projects and a new core (FahCore_15) which can be run only with that client. If you have that client, you will have access to both sets of projects and both cores.

FahCore_15 is still being developed, Eventually all projects will be migrated to it, but that won't be soon. Until the phase out the projects for FahCore_11 and the 6.23 client, you're fine on either one.
bruce
 
Posts: 21751
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: nVidia GT240 success

Postby dogbert2 » Sat Nov 27, 2010 12:49 pm

I have changed the settings to 600 (GPU), 710 (memory), and 1800 (shader), the impact seen so far is about a 10C drop in GPU temperature to 50C (from 60C), so that's a good thing. However, I found that it grabbed the following work unit:

[12:39:13] + Processing work unit
[12:39:13] Core required: FahCore_11.exe
[12:39:13] Core found.
[12:39:13] Working on queue slot 02 [November 27 12:39:13 UTC]
[12:39:13] + Working ...
[12:39:14]
[12:39:14] *------------------------------*
[12:39:14] Folding@Home GPU Core
[12:39:14] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[12:39:14]
[12:39:14] Compiler : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86
[12:39:14] Build host: amoeba
[12:39:14] Board Type: Nvidia
[12:39:14] Core :
[12:39:14] Preparing to commence simulation
[12:39:14] - Looking at optimizations...
[12:39:14] DeleteFrameFiles: successfully deleted file=work/wudata_02.ckp
[12:39:14] - Created dyn
[12:39:14] - Files status OK
[12:39:14] - Expanded 46676 -> 252912 (decompressed 541.8 percent)
[12:39:14] Called DecompressByteArray: compressed_data_size=46676 data_size=252912, decompressed_data_size=252912 diff=0
[12:39:14] - Digital signature verified
[12:39:14]
[12:39:14] Project: 5768 (Run 1, Clone 215, Gen 69)
[12:39:14]
[12:39:14] Assembly optimizations on if available.
[12:39:14] Entering M.D.
[12:39:20] Tpr hash work/wudata_02.tpr: 171187576 418112979 2956194620 3567714971 2588686457
[12:39:20]
[12:39:20] Calling fah_main args: 14 usage=100
[12:39:20]
[12:39:20] Working on Protein
[12:39:21] Client config found, loading data.
[12:39:22] Starting GUI Server

which shows in the GUI display as 'Test Protein A', does this mean that there is a problem (since it shows 'Core Is Not Running Now)? :o
dogbert2
 
Posts: 5
Joined: Wed Nov 24, 2010 11:08 pm

Re: nVidia GT240 success

Postby derrickmcc » Sat Nov 27, 2010 1:06 pm

dogbert2 wrote:I have changed the settings to 600 (GPU), 710 (memory), and 1800 (shader), the impact seen so far is about a 10C drop in GPU temperature to 50C (from 60C), so that's a good thing. However, I found that it grabbed the following work unit:

Code: Select all
[12:39:13] + Processing work unit
[12:39:13] Core required: FahCore_11.exe
[12:39:13] Core found.
[12:39:13] Working on queue slot 02 [November 27 12:39:13 UTC]
[12:39:13] + Working ...
[12:39:14]
[12:39:14] *------------------------------*
[12:39:14] Folding@Home GPU Core
[12:39:14] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[12:39:14]
[12:39:14] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86
[12:39:14] Build host: amoeba
[12:39:14] Board Type: Nvidia
[12:39:14] Core      :
[12:39:14] Preparing to commence simulation
[12:39:14] - Looking at optimizations...
[12:39:14] DeleteFrameFiles: successfully deleted file=work/wudata_02.ckp
[12:39:14] - Created dyn
[12:39:14] - Files status OK
[12:39:14] - Expanded 46676 -> 252912 (decompressed 541.8 percent)
[12:39:14] Called DecompressByteArray: compressed_data_size=46676 data_size=252912, decompressed_data_size=252912 diff=0
[12:39:14] - Digital signature verified
[12:39:14]
[12:39:14] Project: 5768 (Run 1, Clone 215, Gen 69)
[12:39:14]
[12:39:14] Assembly optimizations on if available.
[12:39:14] Entering M.D.
[12:39:20] Tpr hash work/wudata_02.tpr:  171187576 418112979 2956194620 3567714971 2588686457
[12:39:20]
[12:39:20] Calling fah_main args: 14 usage=100
[12:39:20]
[12:39:20] Working on Protein
[12:39:21] Client config found, loading data.
[12:39:22] Starting GUI Server

which shows in the GUI display as 'Test Protein A', does this mean that there is a problem (since it shows 'Core Is Not Running Now)? :o

There is only a problem if the log is still stuck at "Starting GUI Server".
If it has continued and is displaying something like:
[13:29:36] Completed 1%
then you are ok.
Image
User avatar
derrickmcc
 
Posts: 329
Joined: Fri Jul 24, 2009 12:30 am
Location: Malvern, UK

Re: nVidia GT240 success

Postby dogbert2 » Sat Nov 27, 2010 1:37 pm

Thanks...I would hate to think the silly thing is broken :)
dogbert2
 
Posts: 5
Joined: Wed Nov 24, 2010 11:08 pm

Re: nVidia GT240 success

Postby new08 » Wed Dec 08, 2010 10:35 pm

I have the GT240 XTX 512MB version DDR5 memory- and it goes like a train.
Due to the different memory clocks on our versions, the figures look' way over 'for mine, but they work fine.
Core 650
Shader 1700
Memory 1800
The utility GPU Z, shows half clock rates against the default memory setting.[ie: 900 MHz ; default= 1700]
Occasionally, if I stress the card by testing in any way, the clock goes back to 2D settings, but still runs a unit at about 1/3rd pace, which isn't too bad.
Sometimes a reboot is needed to re-instate the full speed, if the use of Riva Tuner 'reloading Video drivers' doesn't do it. [That is the position now - it used to work!]
A good stable card, that runs at about 63C max, on these top settings.
The card fan barely makes any difference, but the PSU fan does suck some draught through every three minutes or so, for 30 secs.
To top this off - I'm running an 8400GS NV card in the adjacent PCI [NOT PCIe] slot, at about 10% of the GT240 output of 5Kp/day.
Image
User avatar
new08
 
Posts: 292
Joined: Fri Jan 04, 2008 11:02 pm
Location: England

Re: nVidia GT240 success

Postby codysluder » Thu Dec 09, 2010 5:54 am

The exact cause of switching back to 2D clocks is unknown, but there's a theory that it's because something was overclocked or overheated too much and the card is protecting itself. Remember 63C is just measured at one spot. If the RAM gets too hot or the GPU gets too hot or the VRs get too hot you still may find that the point that the temperature is being measured didn't get too hot.

If there is any way that the load can stress the card sufficiently to go back to 2D clocks, you're pushing it too hard. Back off a little and you'll gain by never having it go into GPU safe mode at 2D clocks.

The fan should do something. If at a constant load, you manually adjust the fan to a higher or lower setting you should see a measurable response at the temperature sensor. If not, then you need to clean and re-grease the junction between the GPU and the heatsink and check the air path for dust.
codysluder
 
Posts: 2128
Joined: Sun Dec 02, 2007 12:43 pm

Re: nVidia GT240 success

Postby new08 » Fri Dec 10, 2010 7:12 pm

I thought the fan paste on the GPU was likely suspect at first- but where the changing of speed has little apparent effect, the PSU fan sucks enough through to alter temp 1- 2 deg in 30 secs. This made me think the fan was struggling alone- due to airflow in the case, as a direct fan blowing on the area actually made things worse.
So , it may be cooled by air over the GT240 PCB itself [not a bad thing].
When I get the chance to have a poke around- I'll redo the cooling paste, anyway.
Also I backed off the top revs a touch to show respect! ;) Running through the night with 3 clients, makes this good sense.
User avatar
new08
 
Posts: 292
Joined: Fri Jan 04, 2008 11:02 pm
Location: England

Re: nVidia GT240 success

Postby Nathan_P » Fri Dec 10, 2010 7:27 pm

When overclcoking i always override the fan control, gpu's throw out a lot of heat so al my fans are forced to 100% even though they run 24/7. If you are still struggling to keep things cool get some decent case fans with a high CFM rating
Image
Nathan_P
 
Posts: 1409
Joined: Wed Apr 01, 2009 9:22 pm
Location: Jersey, Channel islands

Re: nVidia GT240 success

Postby new08 » Sun Dec 19, 2010 2:55 pm

I've redone the paste on the GPU [twice as the first was not enough]. This gives the same temps as previous, but the paste that I took off was the harder grey type, which was probably original.
I used the white Arctic silver 2 [ a gift from my local pc shop] to renew- and no problems making sure to remove all but a minor trace of the grey. I know they don't mix well
Still keeping to an absolute max of 64C - as beyond this seems a cliff edge, whatever the numbers are doing.
I think extra load from system use can tip the GPU over the edge! Mostly- utilities that address the Gfx bus, in some way.
[The specifics an unknown area to most of us, I wager.]
Not a great idea, really. Most units inadvertently 'so whacked' have recovered to run on, btw.
The new [Beta] client is better for this, seeming less likely to lose its place with data etc.although it happened once or twice during setting up trials.
I don't think I need a case fan, but running a CPU client to load the PSU ensures regular bursts of case air shift, much made to enter via the slot at the two GPUs.
User avatar
new08
 
Posts: 292
Joined: Fri Jan 04, 2008 11:02 pm
Location: England

Re: nVidia GT240 success

Postby new08 » Thu Sep 08, 2011 11:50 am

After a few months running on multiple units, now stable if NOT taken over the o/c limit!
The most temp the GT240 will run at, flat out, is around 70C in an ambient of 23C max.
I reported elsewhere that my old 8400GS [accidentally] ran OK- at over 100C , but this is not possible on these newer designed parallel path chips.
They really are a miracle of design. I wonder if any figures on how many of these [as well as other] GPUs are running on F@H?
The GT240 certainly seems to run well at only ~70 watts overhead.
User avatar
new08
 
Posts: 292
Joined: Fri Jan 04, 2008 11:02 pm
Location: England

Re: nVidia GT240 success

Postby divery4eyes » Thu Sep 08, 2011 9:55 pm

70C.....?
thats all???
I have repeatedly run my cards at over 75C, although these were the Palit, EVGA, and Gigabyte cards. Maybe those brands were better sourced, manufactured, whatever
divery4eyes
 
Posts: 84
Joined: Tue Jun 03, 2008 10:10 pm
Location: off of I40 in Arkansas

Re: nVidia GT240 success

Postby new08 » Wed Dec 14, 2011 7:55 pm

Well, it certainly performs up to the rated max folding @ 5000 ppd at under 70 deg C, but not much more-before EUEs occur.
This issue does depend on what units come down etc [ for stabilty comparisons] -but seems OK, to me.
Why would you push a particular card run past a temperature that causes it to crash?
As the substrates get thinner the temp withstand must reduce.
User avatar
new08
 
Posts: 292
Joined: Fri Jan 04, 2008 11:02 pm
Location: England

Re: nVidia GT240 success

Postby divery4eyes » Wed Dec 14, 2011 9:33 pm

Just for kicks.
I wanted to see how far the card would go without crashing.
So, i stuck my finger in the fan and let it stay there. The card, a Palit, went to 104 C
before i removed my finger from the fan and let it cool down.

I would still be folding with the card if it wasnt for the GTS 450. twice the points in half the space.
divery4eyes
 
Posts: 84
Joined: Tue Jun 03, 2008 10:10 pm
Location: off of I40 in Arkansas

Next

Return to V6 GPU3 beta (including Fermi) OpenMM

Who is online

Users browsing this forum: No registered users and 1 guest

cron