Unsolved
This post is more than 5 years old
3 Posts
0
83329
August 15th, 2006 17:00
Precision 690 slow
I've recently bought a Dell Precision Workstation 690 with the following configuration:
- Dual Xeon 5130 (2GHz, FSB 1333MHz, 4MB)
- 2GB DDR2 533 FBD Memory (2 x 1GB)
- 2 x 146GB (15k rpm) SAS HARD DISK (wiht Raid 0 on integrated Dell SAS Controller 5/iR)
- 16X DVD+/-RW Drive
- 16X DVD-ROM Drive
- ATI Fire GL V7200 256 MB
- Windows XP Professional SP2
My problem is that the system seems to be more slow than it should be also in simple task like explore windows folders in local resorces.
I tried to reinstall OS but this has not fixed the problem.
I hope to recive some advice and/or info to try solving my problem.
Thank you in advance.
Gibi
No Events found!
Chipstone
525 Posts
0
August 15th, 2006 21:00
The technology is available at the processor level, but is not recognized or even utilized at the software level.
The new generation of dual core Xeon’s (5000 & 5100 series) are engineered with features not supported by the current Windows operating systems, to include the X64 versions.
One of the great benefits of past Xeons & Windows was the NetBurst & Hyper threading technology: Windows XP was compiled to fully utilize Hyper threading, creating virtual CPU’s.
The Xeon 5000 series was left with hyper threading, but not the 5100 series:
The 5000 series L2 caches is independently accessed by each dual core (2x2MB).
The 5100’s aggregate 4MB cache is shared by the dual cores:
Windows XP HAL will require a major patch, maybe with the release of SP3 in order to recognize, and hopefully utilize more of the benefits dual-cores with shares L2 cache.
Vista’s re-designed kernel will dynamically recognize the various permutations contained in the current, and future Xeons dual, and multi cores processors.
We have clients that purchased the 690 systems with the 5100 & 5000 Xeon series, and found the 5000 series more responsive under Windows XP. I believe partly due to the legacy hyper threading technology left on the 5000, and L2 cache independent to each dual core.
Windows Server 2003 SP2 (x32 & x64), configured as a workstation operating system, have produced the best performance to date when tested with these new Xeon dual-cores.
HDTechnical
29 Posts
0
August 16th, 2006 16:00
gibi72
3 Posts
0
August 16th, 2006 20:00
HDTechnical
29 Posts
0
August 16th, 2006 21:00
Chipstone
525 Posts
0
August 16th, 2006 21:00
RightMark RM Clock Utility 2.1
http://cpu.rightmark.org/download.shtml
This latest edition contains enhancements to create power management profiles based on dual-cores CPU technology: The Dual Cores section allows you to set the software to reset the CPU to the Max, Average or Sum of the aggregate dual cores speed.
The fix described on Microsoft 896256 article is realy a hard coded registry value, and does not address several issues with dual cores: Try the RM Clock Utility 2.1 with various settings and see if you find a combination that creates sustainable improvement. The software can load up at boot, or manually selected.
Unless you have a PCI-X SAS RAID card, you are better off utilizing the SAS interface on the motherboard. If PERC PCI RAID cards are now SAS compliant make sure they are also PCI-X compliant. PERC Card usually transfers at the old 66 MHz speed interface; severely handicapping any benefits of RAIDS with SAS or SCSI. The PCI-X bus is capable of 133 MHz transfer rates.
Any SCSI/SAS RAID setup will need to be configured with a PCI-X, otherwise you will only see aggregate transfer rates of 65-70MB/sec. PCI-X will yield, depending on the RAID configuration between 120-160MB/sec.
Also check with the Raid management software any play with the write-thru and write-back commands to determine the best transfer rates.
gibi72
3 Posts
0
August 16th, 2006 21:00
HDTechnical
29 Posts
0
August 16th, 2006 22:00
ntheno
44 Posts
0
August 22nd, 2006 16:00
Make sure you go into the RAID management software and make sure that write caching is turned on for each Hard drive to get the best performance. For RAID 1 the default cache settings are off which will result in lower performance (but better data integrity).
HDTechnical
29 Posts
0
August 22nd, 2006 20:00
ntheno
44 Posts
0
August 23rd, 2006 14:00
From Start->Programs open Dell SAS RAID Storage Manager->StartupUI. Login with your current windows login name and password. Select the Logical tab in the first window and then select the virtual disk to highlight it. Then select the operations tab on the right hand window. Then toggle the Set Virtual Disk Properties button and select Enable on the Disk Cache Policy. If you don't have an array defined you can select the Physical disk tab, highlight a drive and then select the Operations tab and select the Enable Write Cache button.
HDTechnical
29 Posts
0
August 25th, 2006 00:00
YannickM
11 Posts
0
August 25th, 2006 08:00
Launching some benchmark like Sisandra, I found the problem was coming from network bandwith, being half on what I had on other machine...Tried to look for bios or chipset upgrade, but didn't found any.
Anyone experienced same issues with memory ? We had several 690 in my company and general feeling is that this machine is slower than it should. Note that we are using XP64.
wotnocomputer
8 Posts
0
September 5th, 2006 23:00
Chipstone
525 Posts
0
September 6th, 2006 04:00
Several caveats will impede performance on some applications, and other will fair better.
Microsoft will need to recompile existing Windows Operating systems to take full advantage of dual/multi cores CPU's functionality:
The operating system can display dual-cores, but that's just eye candy with current compilers:
Vista will have this functionality right out the gate.
Some if not all the current batch of dual-cores Xeons, offered for the Dell 690, are designed without the Hyperthreading capability: Windows operating systems were compiled to take full advantage of this feature.
The operating system and software will take a performance hit if compiled for HT and these Xeons installed.
X32 & x64 Operating Systems & Software will need to be re-compiled to take full advantage of any dual multi-cores capabilities:
Intel’s new C+++ 9.1.030 compiler was tailor made to take full advantage of multi-cores systems, a Fortran compiler is out as well, and Microsoft X64 is either testing or shipping a re-compiled version of the x64; enhanced for multi-cores systems.
I'm doing finite element analysis and was really surprised to see the 690 being 30% slower than a 670.
You are taking a hit in performance if you have the Xeon dual-cores series from 5110-5160, none have HT.
Remember, a 32bit application will not run faster in a 64 bit environment. Only true native x64 applications can fully utilize x64 systems with dual-cores, and only when re-compiled for dual & multi cores CPU's.
Other factors, in combination, can degrade performance as well:
Having the Execute Disable Bit Functionality, of the CPU and software enabled, in conjunction with antivirus program that utilizes heuristic scanning options can really slow down some application processes.
YannickM
11 Posts
0
September 25th, 2006 08:00
Issue was that I had only 2 DImms of RAM, and not 4 Dimms, this was messing up the memory performance.
It tooks 1 month for Dell to resolve that by replacing the 2 Dimms by 4 Dimms, even if I pointed them to it right at the start.
But performance are still average for this machine, not much better than a 3.4 GHz, and far less than a 670 with a processor of 3.73 GHz.
I will test a 5160 and see...
I don't understand where the problem is coming from, note the application I'm testing doesn't use much the disks during analysis, except to write results file from time to time.