New System perfomance - Very poor! Help! Dual Intel E5440

Message boards : Number crunching : New System perfomance - Very poor! Help! Dual Intel E5440
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile Todd Hebert
Volunteer tester
Avatar

Send message
Joined: 16 Jun 00
Posts: 648
Credit: 228,292,957
RAC: 0
United States
Message 738953 - Posted: 14 Apr 2008, 14:41:25 UTC

Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.

The CPU benchmarks are extremely high but my work unit performance is garbage.

My benchmarks are: x8
Measured floating point speed 3091.85 million ops/sec
Measured integer speed 8906.6 million ops/sec

But a single wu is taking on average 8 hours to complete and significantly longer than my Intel Quad Q6600. Most of my other systems are perfoming better than this system. And that includes a dual socket/dual core Dell server that uses Intel 5130 CPU's.

What is wrong? Please help!


ID: 738953 · Report as offensive
Fred W
Volunteer tester

Send message
Joined: 13 Jun 99
Posts: 2524
Credit: 11,954,210
RAC: 0
United Kingdom
Message 738960 - Posted: 14 Apr 2008, 14:58:37 UTC - in response to Message 738953.  

Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.

The CPU benchmarks are extremely high but my work unit performance is garbage.

My benchmarks are: x8
Measured floating point speed 3091.85 million ops/sec
Measured integer speed 8906.6 million ops/sec

But a single wu is taking on average 8 hours to complete and significantly longer than my Intel Quad Q6600. Most of my other systems are perfoming better than this system. And that includes a dual socket/dual core Dell server that uses Intel 5130 CPU's.

What is wrong? Please help!


First point: Your Q6600 is running an optimised App. Which Opp App are you using for the new box?

And Richard Haselgrove has experience of running 4 / 8 core hosts. I'm hoping he might be able to pass an opinion when he happens along.

F.
ID: 738960 · Report as offensive
Profile mr.kjellen
Volunteer tester
Avatar

Send message
Joined: 4 Jan 01
Posts: 195
Credit: 71,324,196
RAC: 0
Sweden
Message 738965 - Posted: 14 Apr 2008, 15:28:16 UTC

Hi,
Too bad. I was looking to getting one of these as well. The few user reports I've been able to find suggests the bios isn't quite up to snuff. Especially memory performance lacks, apparently.

Let's just hope ASUS takes care of the issues in a bios update.

/Anton
ID: 738965 · Report as offensive
Profile Todd Hebert
Volunteer tester
Avatar

Send message
Joined: 16 Jun 00
Posts: 648
Credit: 228,292,957
RAC: 0
United States
Message 738967 - Posted: 14 Apr 2008, 15:37:32 UTC - in response to Message 738960.  

Thanks for your quick reply to my post.

To answer the question as to the Op App that is being used- both systems are using the same SSE3 EM64T app from Crunch3r. No graphics are being employed and both systems run Bonic as a service. I did try the app from Lunatic.kswn.net as well - with the same results.

The task manager app lists the following file as being running (KWSN_2.4V_SSE3_MB.exe) No special priotity has been given to either system to run the process at anything other than normal.

Hopefully the community can come up with some suggestions. This is a serious system and should be performing much better.
Many thanks!

Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.

The CPU benchmarks are extremely high but my work unit performance is garbage.

My benchmarks are: x8
Measured floating point speed 3091.85 million ops/sec
Measured integer speed 8906.6 million ops/sec

But a single wu is taking on average 8 hours to complete and significantly longer than my Intel Quad Q6600. Most of my other systems are perfoming better than this system. And that includes a dual socket/dual core Dell server that uses Intel 5130 CPU's.

What is wrong? Please help!


First point: Your Q6600 is running an optimised App. Which Opp App are you using for the new box?

And Richard Haselgrove has experience of running 4 / 8 core hosts. I'm hoping he might be able to pass an opinion when he happens along.

F.


ID: 738967 · Report as offensive
Profile Neil Walker
Volunteer tester
Avatar

Send message
Joined: 23 May 99
Posts: 288
Credit: 18,101,056
RAC: 0
United Kingdom
Message 738976 - Posted: 14 Apr 2008, 15:50:44 UTC - in response to Message 738953.  

Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.


Quite honestly, even running with Linux, I consider 512MB RAM/core to be the absolute minumum. It makes no sense to spend that amount of money on motherboard and CPUs and then starve them of RAM. The poor thing is probably spending all it's time swapping to disc. :(





Be lucky

Neil



ID: 738976 · Report as offensive
Profile Todd Hebert
Volunteer tester
Avatar

Send message
Joined: 16 Jun 00
Posts: 648
Credit: 228,292,957
RAC: 0
United States
Message 738982 - Posted: 14 Apr 2008, 15:56:02 UTC - in response to Message 738976.  

Later today it will have an additional 4GB added to the system. I have to wait for UPS to roll through with my package. These DIMM's are just what I had laying around. The one box that had my ram in it didn't ship from the same location and it didn't arrive.

I agree, a system like this deserves plenty of ram and in two weeks I'll be adding another 4GB-I would have done it now-but I have to pay my taxes :(

Todd

Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.


Quite honestly, even running with Linux, I consider 512MB RAM/core to be the absolute minumum. It makes no sense to spend that amount of money on motherboard and CPUs and then starve them of RAM. The poor thing is probably spending all it's time swapping to disc. :(






ID: 738982 · Report as offensive
Profile zoom3+1=4
Volunteer tester
Avatar

Send message
Joined: 30 Nov 03
Posts: 65738
Credit: 55,293,173
RAC: 49
United States
Message 738983 - Posted: 14 Apr 2008, 15:56:06 UTC - in response to Message 738967.  

Thanks for your quick reply to my post.

To answer the question as to the Op App that is being used- both systems are using the same SSE3 EM64T app from Crunch3r. No graphics are being employed and both systems run Boinc as a service. I did try the app from Lunatic.kswn.net as well - with the same results.

The task manager app lists the following file as being running (KWSN_2.4V_SSE3_MB.exe) No special priority has been given to either system to run the process at anything other than normal.

Hopefully the community can come up with some suggestions. This is a serious system and should be performing much better.
Many thanks!

Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.

The CPU benchmarks are extremely high but my work unit performance is garbage.

My benchmarks are: x8
Measured floating point speed 3091.85 million ops/sec
Measured integer speed 8906.6 million ops/sec

But a single wu is taking on average 8 hours to complete and significantly longer than my Intel Quad Q6600. Most of my other systems are performing better than this system. And that includes a dual socket/dual core Dell server that uses Intel 5130 CPU's.

What is wrong? Please help!


First point: Your Q6600 is running an optimized App. Which Opp App are you using for the new box?

And Richard Haselgrove has experience of running 4 / 8 core hosts. I'm hoping he might be able to pass an opinion when he happens along.

F.


You could with those cpus switch to the SSSE3 2.4V Seti app as from what I've been able to find online, It supports SSSE3 just fine and I think It's faster on this cpu than SSE3 would be, I run SSSE3 on My 3-Quads of course.

Good Luck.
The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's
ID: 738983 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 738999 - Posted: 14 Apr 2008, 16:39:47 UTC - in response to Message 738976.  

Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.

Quite honestly, even running with Linux, I consider 512MB RAM/core to be the absolute minumum. It makes no sense to spend that amount of money on motherboard and CPUs and then starve them of RAM. The poor thing is probably spending all it's time swapping to disc. :(

A single instance of the S@H app 2.4V 32 bit builds uses less than 30 MiB, the 64 bit builds somewhat more, perhaps 50MiB. I think if the Task Manager commit charge were examined on that system while running 8 S@H WUs, it would be under 512 MiB unless some XP background tasks are using far more than they should.
                                                               Joe
ID: 738999 · Report as offensive
Profile Todd Hebert
Volunteer tester
Avatar

Send message
Joined: 16 Jun 00
Posts: 648
Credit: 228,292,957
RAC: 0
United States
Message 739006 - Posted: 14 Apr 2008, 16:54:20 UTC - in response to Message 738999.  

Joe (et al)
The instances of the KWSN app use around 34 Megs each so the load in terms of system memory are not that significant. So for the 8 tasks that adds up to 272 MB.
The commit charge on the system is around 650MB with a limit of 5GB when considering the swap file is available-but not used significantly.

Bottom line is there is plenty of memory available even with 8 tasks running. At this point there is not many processes running in the background. Now there is 41, and 8 of those are BONIC and the application. This system doesn't have much in the way of apps installed yet.
Todd

Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.

Quite honestly, even running with Linux, I consider 512MB RAM/core to be the absolute minumum. It makes no sense to spend that amount of money on motherboard and CPUs and then starve them of RAM. The poor thing is probably spending all it's time swapping to disc. :(

A single instance of the S@H app 2.4V 32 bit builds uses less than 30 MiB, the 64 bit builds somewhat more, perhaps 50MiB. I think if the Task Manager commit charge were examined on that system while running 8 S@H WUs, it would be under 512 MiB unless some XP background tasks are using far more than they should.
                                                               Joe


ID: 739006 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14650
Credit: 200,643,578
RAC: 874
United Kingdom
Message 739020 - Posted: 14 Apr 2008, 17:30:34 UTC - in response to Message 738960.  

Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.

The CPU benchmarks are extremely high but my work unit performance is garbage.

My benchmarks are: x8
Measured floating point speed 3091.85 million ops/sec
Measured integer speed 8906.6 million ops/sec

But a single wu is taking on average 8 hours to complete and significantly longer than my Intel Quad Q6600. Most of my other systems are perfoming better than this system. And that includes a dual socket/dual core Dell server that uses Intel 5130 CPU's.

What is wrong? Please help!


First point: Your Q6600 is running an optimised App. Which Opp App are you using for the new box?

And Richard Haselgrove has experience of running 4 / 8 core hosts. I'm hoping he might be able to pass an opinion when he happens along.

F.

Sorry, I was wrestling with a gigabyte motherboard which keeps resetting and BSODing for no apparent reason - most commonly, BSOD D1 (driver error) NVNRM.SYS (onboard NVidia network). Updated the motherboard drivers (inclucing network) and it lost all connectivity. Ugh.

Anyway, you were asking about 8-cores. First, while you're still learning the system, upgrading things etc., I'd downsize your cache. A lot. Much less embarassing when/if things go wrong! Eventually I found a recent working set (round about offset 300/500), and yes, they look extremely slow.

When I started with my dual E5320s, I was quickly told that the channel mode of the FB DIMMs was crucial. CPU-Z doesn't seem to pick it up well, but my Dell BIOS reports it as quad channel, now I have 4 matching DIMMs. You've got 4 x 512MB (to start with), but to get quad channel speed, you have to use the right slots - some seem to use [1,2,3,4] [5,6,7,8], and others [1,3,5,7] [2,4,6,8].

But having said that, I only speeded up 25% when switching from dual to quad channel memory - your speeds seem much further below par than that. And with XP_64, I'm afraid I'm out of my depth: but I think I would be double-checking that I'd got the right 64-bit motherboard device drivers installed.
ID: 739020 · Report as offensive
Profile Todd Hebert
Volunteer tester
Avatar

Send message
Joined: 16 Jun 00
Posts: 648
Credit: 228,292,957
RAC: 0
United States
Message 739031 - Posted: 14 Apr 2008, 18:00:33 UTC - in response to Message 739020.  

I am familar with the Quad memory bus layout of this type of system and that is the way that it is configured. Hence the current reason for using 4x512 dimms. I wanted to rule out the memory as being the bottleneck.

I have ran several test with SiSoft Sandra and the memory benchmark is right where it should be when compared to other systems of this variety.

Granted the channel size is only 512mb, but the performance is better than the reference systems that are included as a baseline.

The drivers for this system are the most currently available from Intel for the chipset and Nvidia (Video Card) and Ciprico (Raid card-PCI-E x4-running RAID 0)

From the outside it is difficult to understand why this is happening. All other benchmark utilities show my performance to be top notch, and operation with the limited number of apps installed are really fast. It just seems to be an issue with BONIC.

Thanks for the feedback
Todd
Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.

The CPU benchmarks are extremely high but my work unit performance is garbage.

My benchmarks are: x8
Measured floating point speed 3091.85 million ops/sec
Measured integer speed 8906.6 million ops/sec

But a single wu is taking on average 8 hours to complete and significantly longer than my Intel Quad Q6600. Most of my other systems are perfoming better than this system. And that includes a dual socket/dual core Dell server that uses Intel 5130 CPU's.

What is wrong? Please help!


First point: Your Q6600 is running an optimised App. Which Opp App are you using for the new box?

And Richard Haselgrove has experience of running 4 / 8 core hosts. I'm hoping he might be able to pass an opinion when he happens along.

F.

Sorry, I was wrestling with a gigabyte motherboard which keeps resetting and BSODing for no apparent reason - most commonly, BSOD D1 (driver error) NVNRM.SYS (onboard NVidia network). Updated the motherboard drivers (inclucing network) and it lost all connectivity. Ugh.

Anyway, you were asking about 8-cores. First, while you're still learning the system, upgrading things etc., I'd downsize your cache. A lot. Much less embarassing when/if things go wrong! Eventually I found a recent working set (round about offset 300/500), and yes, they look extremely slow.

When I started with my dual E5320s, I was quickly told that the channel mode of the FB DIMMs was crucial. CPU-Z doesn't seem to pick it up well, but my Dell BIOS reports it as quad channel, now I have 4 matching DIMMs. You've got 4 x 512MB (to start with), but to get quad channel speed, you have to use the right slots - some seem to use [1,2,3,4] [5,6,7,8], and others [1,3,5,7] [2,4,6,8].

But having said that, I only speeded up 25% when switching from dual to quad channel memory - your speeds seem much further below par than that. And with XP_64, I'm afraid I'm out of my depth: but I think I would be double-checking that I'd got the right 64-bit motherboard device drivers installed.


ID: 739031 · 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 739034 - Posted: 14 Apr 2008, 18:07:33 UTC - in response to Message 738976.  

Hello, I just built a new system that consists of a Dual Socket Asus Z7S WS (Intel 5400) that is populated with Dual Intel Quad Cores E5440's running at 3.06Ghz with 2GB RAM (4x512 FB-DIMM's) Running XP Pro x64.


Quite honestly, even running with Linux, I consider 512MB RAM/core to be the absolute minumum. It makes no sense to spend that amount of money on motherboard and CPUs and then starve them of RAM. The poor thing is probably spending all it's time swapping to disc. :(





OT
Hey Neil, Fellow class mate, Good to see you up and running again...Thanks for letting me get 1st place for a bit....You already have me looking at more upgrades...LOL
Official Abuser of Boinc Buttons...
And no good credit hound!
ID: 739034 · Report as offensive
HPew

Send message
Joined: 10 Nov 07
Posts: 8
Credit: 232,258
RAC: 0
United Kingdom
Message 744420 - Posted: 26 Apr 2008, 10:52:48 UTC

I must report the same problem as Todd Hebert, although on a lower scale.

I have just picked this PC up, second hand. Two Xeon E5320s (BSEL'd to 2.3 GHz) Tyan 5000 board, 4GB RAM (4 x 1 FB, all DIMM slots populated).

Running Debian, BOINC benchmarks of 2255/9420. Time taken to complete a SETI WU around four hours.

My comparison is a Xeon 3050 (benchmarks 2090/4856) with 1 GB of RAM which takes around 2 hours 15 minutes on a SETI WU.

I'm not impressed.
ID: 744420 · Report as offensive
Profile Bryan Wallace
Volunteer tester

Send message
Joined: 28 Jul 00
Posts: 22
Credit: 52,559,173
RAC: 0
United States
Message 745344 - Posted: 28 Apr 2008, 3:09:57 UTC

todd, it appears that you still haven't gotten your issue figured out. try posting a thread in the DC section at Anandtech. Rattledagger has helped me out many times with intractable problems such as this!
ID: 745344 · Report as offensive
Profile mr.kjellen
Volunteer tester
Avatar

Send message
Joined: 4 Jan 01
Posts: 195
Credit: 71,324,196
RAC: 0
Sweden
Message 745356 - Posted: 28 Apr 2008, 6:42:33 UTC

The consensus out there on other fora seems to be that there is an option missing in the ASUS bios, (And one that could be the source of your problems as well with the Tyan board Kosme,) called Disable Snoop Filter. Not sure what it does really, but it needs to be cecked. (Snoop filter disabled)
(Didn't read it but someone over at realworldtech.com seems to be in the know.)

You could see THIS thread for the Asus Z7S-WS board.

Someone over at WCG is running into trouble as well.

/Anton
ID: 745356 · Report as offensive
HPew

Send message
Joined: 10 Nov 07
Posts: 8
Credit: 232,258
RAC: 0
United Kingdom
Message 745791 - Posted: 29 Apr 2008, 14:52:59 UTC - in response to Message 745356.  

Thanks for your efforts mr.kjellen.

But I checked my board's bios and it has snoop filter set to disabled already, so I am going to *reverse* your advice and enable it to see what happens.

The computer has a number of other problems but I will report how the change effects performance.
ID: 745791 · Report as offensive
HPew

Send message
Joined: 10 Nov 07
Posts: 8
Credit: 232,258
RAC: 0
United Kingdom
Message 746105 - Posted: 30 Apr 2008, 15:04:26 UTC - in response to Message 745791.  

After wrestling with some non-BOINC-related computer problems I ran two sets of eight WUs, one with snoop filter enabled then one disabled.

With snoop filter enabled all eight WUs took around 3hrs 45mins each. All March '08 WUs.

With snoop filter disabled the eight WU times ranged from 3hrs 15mins to 3hrs 55mins. These were a mix of WUs from Dec '06, Jan '07, and Feb '08.

So, ceteris paribus, the problem with my computer's BOINC performance is not related, or only marginally related (in the order of less than 5%?), to the snoop filter settings.
ID: 746105 · Report as offensive
Profile Bryan Wallace
Volunteer tester

Send message
Joined: 28 Jul 00
Posts: 22
Credit: 52,559,173
RAC: 0
United States
Message 746583 - Posted: 1 May 2008, 18:50:18 UTC - in response to Message 746105.  

After wrestling with some non-BOINC-related computer problems I ran two sets of eight WUs, one with snoop filter enabled then one disabled.

With snoop filter enabled all eight WUs took around 3hrs 45mins each. All March '08 WUs.

With snoop filter disabled the eight WU times ranged from 3hrs 15mins to 3hrs 55mins. These were a mix of WUs from Dec '06, Jan '07, and Feb '08.

So, ceteris paribus, the problem with my computer's BOINC performance is not related, or only marginally related (in the order of less than 5%?), to the snoop filter settings.


the time to complete a workunit doesn't mean as much as the credit from the workunit. for example, a better description would be: "Complete eight WU's, all of which were 54-55 pts, in 3hrs 45min, then completed 8 more WU's with similar pt value in 3hrs 5min to 3hrs 55 min".
ID: 746583 · Report as offensive
Ingleside
Volunteer developer

Send message
Joined: 4 Feb 03
Posts: 1546
Credit: 15,832,022
RAC: 13
Norway
Message 746629 - Posted: 1 May 2008, 20:27:54 UTC
Last modified: 1 May 2008, 20:28:42 UTC

Hmm, there was a thread over at Rosetta resently of someone having mediocre performance with dual E5430 and another Asus-board... Atleast in this instance, fooling-around with the various BIOS-memory-settings seemed to improve things somewhat...

But, since already has high memory-benchmarks, it's not certain this will improve anything in this instance...

In any case, it's still an idea to look-through all the various BIOS-settings, and make sure there's not an obvious bad setting like "disable cpu-cache". Also make sure "Speedstep" is disabled, and check on any other power-saving-settings. Also check power-saving-settings in Windows.

Oh, and if haven't done it already, check that the system isn't slowing-down due to overheating...


Hmm, just to rule-out a memory-bottleneck, is it the same dreadful performance if you switches to running only 1 seti-instance instead of 8 at a time?
"I make so many mistakes. But then just think of all the mistakes I don't make, although I might."
ID: 746629 · Report as offensive
Profile mr.kjellen
Volunteer tester
Avatar

Send message
Joined: 4 Jan 01
Posts: 195
Credit: 71,324,196
RAC: 0
Sweden
Message 748205 - Posted: 4 May 2008, 20:09:48 UTC

Seems there is a new Bios available for the Z7S-WS at the ASUS FTP site.

Take that for a spin.

/Anton
ID: 748205 · Report as offensive
1 · 2 · Next

Message boards : Number crunching : New System perfomance - Very poor! Help! Dual Intel E5440


 
©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.