x1900 question

the services have been disabled for a few days now and ive reset inbetween, i uninstalled ATT completely when changing to .949, reset and installed new version, the speeds are set on a profile which currently, while i try to get this issue sorted out, im starting up manually.
so initially the computer boots into windows at default 2d timings, then i select my profile with the faster speeds, extra voltage needed etc, and yes the profile definately is kicking in :) then i start up F@H last of all, i dont get why its going slow though, its increased marginally from ~9mins to ~8mins in everything that iv'e done, OCing the CPU, memory and the card itself :(
 
yes according to task manager its using the whole cpu, its only on a single core opteron, the only thing i can think of (well assuming this card isnt just messed up) is ATT isnt changing the speeds even when it says it is, im intedning to get something else to monitor the gpu core/memory at the same time, can anyone recommend an app that i can use at the same time which wont affect ATT setting the speeds?
Im thinking riva tuner but i dont know if i can run both at the same time, doesnt need to be complex either, just something to monitor them
 
hmmm - I wonder if there's enough other processes running on the single core that the performance of the GPU client ends up being affected :confused:
I don't see how it could make that much difference though, I'm running both the GPU and a CPU client at "low" on my X2 so any other processes should affect both clients by a similar amount

totally flumoxed on this one I'm afraid :o
 
its a fresh windows install, + logitech mouse drivers, ATT and F@H running and thats it, so i dont think its that, thing is its not guna be on that long each day and i dont want it to keep missing deadlines, which i think it may do if each frame is going to take 8mins :( may just give up and put a cpu client on it but i really dont want to waste the gpu if i can just work out whats holding the stupid thing back
 
rich99million said:
hmmm - I wonder if there's enough other processes running on the single core that the performance of the GPU client ends up being affected :confused:
I don't see how it could make that much difference though, I'm running both the GPU and a CPU client at "low" on my X2 so any other processes should affect both clients by a similar amount
totally flumoxed on this one I'm afraid :o

Hi Uncle Rich,

Last night I started a test:

Setup Before:
1x Core 100%, Idle
1x Core 0%
1x GPU 100%, Idle
GPU Frame Times averaged 5.8 to 5.9

Current Setup:
1x Core 100%, Idle
1x Core 100% Low
1x GPU 100%, Idle
GPU Frame Times averaged 5.8 to 5.9 thus far.
I also brought down the CPU clocks to 3.2 and maintained a high FSB 400MHZ. For the moment things seem to be stable.

For the first GPU WU, it did not seem to have much effect on the GPU work unit times, but will have to see what happenes on the Core WU and subsequent GPU WU's.
.
 
wow - I was sure that people had tried out all sorts of combinations and come to the conclusion that it needs a whole core
I'll give that a go on my X2 aswell I think, if the GPU will happily work with half a core then I've just wasted half a core of 24/7 work for the last week :o :p
 
rich99million said:
wow - I was sure that people had tried out all sorts of combinations and come to the conclusion that it needs a whole core
I'll give that a go on my X2 aswell I think, if the GPU will happily work with half a core then I've just wasted half a core of 24/7 work for the last week :o :p


No, I tried that on sunday and GPU does need a full core still :confused:
 
What one!? :confused:

Setting:

CPU Client 1: Low 100%
CPU Client 2: Idle 100%
GPU Client 1: Idle 100%

= GPU getting no time at all pretty much...

Im sure they haven't changed the core to not need 100% polling...
 
from KE1HA's results page the GPU should be set to low not idle as posted above, will try that now

tried before with 1 CPU low and the other two clients idle and got the same as conc

edit: Test 7 of the blue results is what I'm looking at

7 Idle 100% 1:10:30 31.25 low 80% 0:16:52 512.25 low 100% 0:06:16 758.30 1301.80
 
rich99million said:
from KE1HA's results page the GPU should be set to low not idle as posted above, will try that now

tried before with 1 CPU low and the other two clients idle and got the same as conc

edit: Test 7 of the blue results is what I'm looking at


80% CPU makes sence, gives the GPU some time atleast.

GPU @ 100% Low
CPU1 @ 100% Low
CPU 2 @ Idle

Will give GPU and CPU1 100% of each core... giving CPU2 nothing..
 
rich99million said:
from KE1HA's results page the GPU should be set to low not idle as posted above, will try that now

tried before with 1 CPU low and the other two clients idle and got the same as conc

edit: Test 7 of the blue results is what I'm looking at
Opps... I need to set it properly as well ... LMAO ... well when I get home will have have to do the tweeks.


From that Graph, it looks like the following would be a good compromise:

1x Core 30% @ Low
1x Core 100% @ Low
1x CPU 100% @ Low

should Yeild 1200+ PPD provided you get 600pt CPU WU's.
.
 
Last edited:
KE1HA said:
Opps... I need to set it properly as well ... LMAO ... well when I get home will have have to do the tweeks.
.


tbh m8 I think one of your CPU cores is doing no work atm. Thats what I get here with those settings.

Persoanlly im going to wait until stanford say what they have done to the polling. Keep it at 1 core for GPU and 2 CPU clients on my faster core 1.
 
Concorde Rules said:
tbh m8 I think one of your CPU cores is doing no work atm. Thats what I get here with those settings.

Persoanlly im going to wait until stanford say what they have done to the polling. Keep it at 1 core for GPU and 2 CPU clients on my faster core 1.
That's what I have been doing for the most part, one core @ 0% used by the GPU for polling. So if I'm not slowing down the GPU or the other core by having both Cores running then any PPD from the second core would be a plus.
.
 
KE1HA said:
That's what I have been doing for the most part, one core @ 0% used by the GPU for polling. So if I'm not slowing down the GPU or the other core by having both Cores running then any PPD from the second core would be a plus.
.
oh well I'm going to stick with just 1 CPU and the GPU both at low - I only have 1GB RAM so 3 clients with large WUs is going to leave me short for anything demanding
 
rich99million said:
oh well I'm going to stick with just 1 CPU and the GPU both at low - I only have 1GB RAM so 3 clients with large WUs is going to leave me short for anything demanding
Yeppers, Concorde was right, the second core was doing little to nothing by way of it's own WU, so I've reconfigured them all.

1x Core @ 100% and Low
1x Core @ 50% and Low
1x GPU @ 100% and Low

We'l see how that progresses. Have 2GB somewhat zippy RAM, so shouldn't be to tasking on that front.

I did check Core-1 and it's processing a 2135 @ 21.1 Mins / frame, and the GPU had done 70 frames @ 5.7 mins / frame. It was the second Core that wasn't doing squat !! :D
.
 
Back
Top Bottom