Frustratingly long LL :(

hj47 · 2009-04-01, 07:44

Hey again,

So I set up a new rig for 24/7 LL testing 2 months ago, but it seems as though the LL test I am working on is taking forever.

I was under the impression that an LL would take something like a month to complete on modest hardware, but this seems to be inconsistent in regards to the exponent I'm currently testing (M46241791 & M46241807).

I have a stock clocked E5200 running 24/7 running in Windows 7 x64, and it has taken 62 days (according to primenet) to do 79% of both LL tests. I have the two exponents running on a single core each.

The average itn. time seems to be around 0.280 sec, although this can go up to 330 and down to 250.

I have a feeling that my GPU folding@home client, which also runs 24/7 on this pc, may be stealing some CPU cycles, but I wouldn't have though it would make prime95 seem so slow.

Is it normal for this rig to take this long when working on 2 exponents?

Any advice/knowledge would be appreciated :D

sdbardwick · 2009-04-01, 09:05

You are getting horrible performance from that E5200. Usual suspects are:
1. F@H client (even though it is GPU focused - stop it and run a benchmark)
2. Misassigned worker threads
3. Excessive heat causing thermal throttling of CPU, due to misaligned HSF or excessive dust blocking airflow (dust is unlikely due to age of the box)

What version of the P95 software are you running (if it is less than 25.9, obtain the current version)

For comparison, here is the benchmark results page (it can take a while to load); I'd disregard the extremely high and extremely low times for each processor.

Kevin · 2009-04-01, 19:24

I think it's probably F@H. I tried using the GPU client on my Q6600, and it killed Prime95's productivity.

cheesehead · 2009-04-02, 01:48

hj47,

GIMPS software is designed to run at lowest priority so as not to cause delays in your other system work. That means that whenever your system has _any_ other task to perform, that task will have priority over prime95/mprime.

GIMPS is intended to make use of your system's otherwise-idle cycles; running at the lowest priority accomplishes that.

Quote:

Originally Posted by hj47

I was under the impression that an LL would take something like a month to complete on modest hardware, but this seems to be inconsistent in regards to the exponent I'm currently testing (M46241791 & M46241807).

That prediction presumes that your other system tasks do not take a significant amount of time to run (at higher priority than prime95). Any other compute-intensive task, such as some other distributed computing projects or video games, will run at the expense of taking time from prime95.

If you want to change that, assign a higher priority to prime95. Warning: such a change might interfere with overall system responsiveness.

RMAC9.5 · 2009-04-02, 06:10

hj47,
I currently run F@H and Prime95 on 2 PCs, so I may be able to provide some insight. You didn't share what kind of video card you have but I will hazard a guess that it is an ATI card. If so, that is most likely where the problem lies. This is not a knock on ATI cards. I use them in most of my PCs and prefer them over Nvidia cards. The F@H GPU client relies on the CPU to keep it fed with data. ATI cards currently need a lot more CPU cycles than Nvidia cards do to stay fed, but the F@H help forums indicate that this ATI problem will hopefully be solved in the next few months with new drivers from AMD.
Use Windows Task Manager to look at the CPU usage for Prime95.exe and FahCore_11.exe. On my dual core 2 GHz Opteron 170, the F@H client is using about 1/2 of 1 CPU and the iteration times for my M2311xxxx DCs range between .072 and .200 per sec.

joblack · 2009-04-02, 20:51

Quote:

Originally Posted by RMAC9.5

hj47,
I currently run F@H and Prime95 on 2 PCs, so I may be able to provide some insight. You didn't share what kind of video card you have but I will hazard a guess that it is an ATI card. If so, that is most likely where the problem lies. This is not a knock on ATI cards. I use them in most of my PCs and prefer them over Nvidia cards. The F@H GPU client relies on the CPU to keep it fed with data. ATI cards currently need a lot more CPU cycles than Nvidia cards do to stay fed, but the F@H help forums indicate that this ATI problem will hopefully be solved in the next few months with new drivers from AMD.
Use Windows Task Manager to look at the CPU usage for Prime95.exe and FahCore_11.exe. On my dual core 2 GHz Opteron 170, the F@H client is using about 1/2 of 1 CPU and the iteration times for my M2311xxxx DCs range between .072 and .200 per sec.

Ati tried to sneak in their Distributed project. It is enabled by default - so some uneducated user would just press 'next' and would have this 'spyware' on the computer.

hj47 · 2009-04-03, 11:50

Quote:

Originally Posted by joblack

Ati tried to sneak in their Distributed project. It is enabled by default - so some uneducated user would just press 'next' and would have this 'spyware' on the computer.

AFAIK ATI does not enable F@H by default. I don't know why you are comparing this to spyware :( ????

Anyhoo, I am running a 9600GSO and that's why I am confused, because my CPU usage is always at or below 1%.

I will have to troubleshoot this weekend with the suggestions provided.

Cheers :D

starrynte · 2009-04-08, 17:56

Also formerly Digsby (IM client) had this "research module" enabled by default last I checked...(see http://blog.digsby.com/archives/68)
To disable: Help -> Support Digsby, then disable "Help Digsby Conduct Research"

joblack · 2009-04-19, 17:30

Quote:

Originally Posted by hj47

AFAIK ATI does not enable F@H by default. I don't know why you are comparing this to spyware :( ????

Anyhoo, I am running a 9600GSO and that's why I am confused, because my CPU usage is always at or below 1%.
Cheers :D

They enabled it by default in my driver version. Additionally I have seen over 15 % processor use on my Q6600.

I think if you have it as a default setting the user has the right to know what this software does. ATI didn't explain what it exactly does. Additionally the graphics card is leeching a lot of extra power so ATI is tricking the users in paying their needs and demands.

2009-04-01, 07:44	#1
hj47 Oct 2008 2⁶ Posts	Frustratingly long LL :( Hey again, So I set up a new rig for 24/7 LL testing 2 months ago, but it seems as though the LL test I am working on is taking forever. I was under the impression that an LL would take something like a month to complete on modest hardware, but this seems to be inconsistent in regards to the exponent I'm currently testing (M46241791 & M46241807). I have a stock clocked E5200 running 24/7 running in Windows 7 x64, and it has taken 62 days (according to primenet) to do 79% of both LL tests. I have the two exponents running on a single core each. The average itn. time seems to be around 0.280 sec, although this can go up to 330 and down to 250. I have a feeling that my GPU folding@home client, which also runs 24/7 on this pc, may be stealing some CPU cycles, but I wouldn't have though it would make prime95 seem so slow. Is it normal for this rig to take this long when working on 2 exponents? Any advice/knowledge would be appreciated :D

2009-04-01, 09:05	#2
sdbardwick Aug 2002 North San Diego County 5×137 Posts	You are getting horrible performance from that E5200. Usual suspects are: 1. F@H client (even though it is GPU focused - stop it and run a benchmark) 2. Misassigned worker threads 3. Excessive heat causing thermal throttling of CPU, due to misaligned HSF or excessive dust blocking airflow (dust is unlikely due to age of the box) What version of the P95 software are you running (if it is less than 25.9, obtain the current version) For comparison, here is the benchmark results page (it can take a while to load); I'd disregard the extremely high and extremely low times for each processor. Last fiddled with by sdbardwick on 2009-04-01 at 09:06

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Frustratingly frequent forum font fluctuations?	WraithX	Forum Feedback	11	2014-10-18 13:36
How long is too long?	ThomRuley	Msieve	3	2013-11-30 04:52
A frustratingly large number on offer	fivemack	Aliquot Sequences	7	2012-01-14 02:39
Using long long's in Mingw with 32-bit Windows XP	grandpascorpion	Programming	7	2009-10-04 12:13
I think it's gonna be a long, long time	panic	Hardware	9	2009-09-11 05:11

2009-04-01, 19:24	#3
Kevin Aug 2002 Ann Arbor, MI 433 Posts	I think it's probably F@H. I tried using the GPU client on my Q6600, and it killed Prime95's productivity.

2009-04-02, 06:10	#5
RMAC9.5 Jun 2003 3²·17 Posts	hj47, I currently run F@H and Prime95 on 2 PCs, so I may be able to provide some insight. You didn't share what kind of video card you have but I will hazard a guess that it is an ATI card. If so, that is most likely where the problem lies. This is not a knock on ATI cards. I use them in most of my PCs and prefer them over Nvidia cards. The F@H GPU client relies on the CPU to keep it fed with data. ATI cards currently need a lot more CPU cycles than Nvidia cards do to stay fed, but the F@H help forums indicate that this ATI problem will hopefully be solved in the next few months with new drivers from AMD. Use Windows Task Manager to look at the CPU usage for Prime95.exe and FahCore_11.exe. On my dual core 2 GHz Opteron 170, the F@H client is using about 1/2 of 1 CPU and the iteration times for my M2311xxxx DCs range between .072 and .200 per sec.

2009-04-08, 17:56	#8
starrynte Oct 2008 California 236₁₀ Posts	Also formerly Digsby (IM client) had this "research module" enabled by default last I checked...(see http://blog.digsby.com/archives/68) To disable: Help -> Support Digsby, then disable "Help Digsby Conduct Research"