09-02-2013 02:09 PM
Purchased a used XW6400 about a year ago and it was used rarely. Upgraded the processors to a pair of E5345's, graphics card to a NVIDIA GTS450. Now using the machine pretty much daily as a home PC. Its used for a few hours a day. Primarily browsing, spreadsheets. etc.
Total Ram was 6GB now upgraded to 8GB with matching Samsung sticks.
Getting the 927 FSB error F1 reboot errors only maybe once a week. Machine is set to go to standby which it recovers from no problems, but randomly and at random times - only once a week it will hang for a second then reboot. I don't get a BSOD just power cycles to the F1 Bios Error screen. Latest BIOS, running W7 ULTIMATE 64. Latest NVIDIA drivers. Install of W7 wasn't a problem but I didn't reinstall after upgrading the processors. My motherboard is supposedly compatible with the E5345's once I read Scott Harrison's posts and my model and version matched his. When I had two Elpida 2GB's and Micron 1GB sticks in the system I would also get the error as well. From his posts he said that using matched Samsungs eliminated his non-fatal FSB errors but in my case it does not.
I've done the usual tests of MEMTEST86 - running for several hours. Ran FurMark and other burn tests for several hours all without faults, crashes or errors. To me if it passes these tests the processor, graphics card, memory or motherboard should theorectically be sound since all the components were pretty stressed. The system is attached to a UPS with voltage regulator/compensator so the power to the system is clean. Power supply to the PCI-E should be oK but its not been bench tested.
I've read that there could be problems with the E5345's and this motherboard but supposedly its been tested on my version without any problems. I have a 442029-001
I have a spare motherboard 442029-001 Version OT which I can swap in and a very old Quado FX570 card which I can put back into the system to see if its the processors/motherboard or the graphics card that is incompatible. I also have another GTS450 running in a XW8600 which runs flawlessly which I also can swap between the two machines. I also have the original 5150's that came with the machine which as a last resort I can put back in but I'd prefer the quads.
So does it sound like the graphics card my be incompatible with the motherboard or is something else at fault - hard drive, WIndows Installation, etc. Since the error isn't something that I can force to replicate its going to be difficult to narrow down the problem or at least it will take several weeks of removing a component and seeing what happens.
Any thoughts would be appreciated. Since its not my primary machine I can start from scratch - new Windows 7 install, etc.
09-02-2013 05:48 PM
These things can be frustrating...... I think you're doing all the right things maybe for the wrong reasons. The E5345's inability to run in a subset of xw6400 motherboards generates a "922 - Fatal error on front side bus" and crash. You're getting a 927 non-fatal error, and applying some of my 922 solutions for the 927 error code. They are quite different.
Here's a couple of ideas...... go the the "Search HP Forum" on this forum page, near top left, and type in "927". You'll find this thread and 3 others. Then, launch google and search for "927 nonfatal error hp". You'll find more good stuff.
There are a bunch of hp related hits there that are worth reading, and things seem to point to the video card. Make sure you update your xw6400's BIOS to the latest before doing anything (it is 2.38, from December 2010). Has that been done already? I just posted on the safest way to do that earlier today.
There also are firmware updates for alternative HP video cards you might want to use, instead, in that xw6400, and I'd recommend changing to a different card as your second step, even temporarily. I have found that HP firmware updaters virtually always work with the Dell and PNY branded same cards, also, and make sure you do such an update if you choose to try such a card. Post back if you can't find one for a quadro card you might choose..... I have the SP numbers for many of those, but they should be pretty easy to find.
I'm liking the Quadro FX3700 or FX3800 for single slot use, and these are reasonable, used, on eBay. We're also running a EVGA GTX 650 TI Boost here without trouble, on our xw8400. That's 134W max. Your current card runs at 106W max, and should be happy if it is given extra power and plugged into the top PCIe x16 slot. The extra power should come from the PCIe x16 supplemental power cable that comes out of the HP power supply for that purpose, not some adapters. The FX3700 runs 78W max; the FX3800 runs 108W max. The fan on the FX3700 runs faster/louder than the FX3800, so I like that one for fast quiet builds.
This all should be fixable, and please post back with a solution that you may discover to help others.
09-02-2013 06:57 PM
I've gotten a 922 Fatal Error before. I just checked my paper log. If was a 922 Fatal Error Northbridge B3 Err Coherency Violation Error. That error only occurred once and the mismatched memory was installed. I've been runing the matched Samsungs for about 5 days.
Looking at the Windows logs there have been a few Nvidia crashes that recovered. I think I'll do a video driver cleanse and reinstall and see how that goes. The GTS450 is a decent card - runs quiet and cool - and I'd love to keep it in this machine.
I'm assuming that if I had hardware faults I would crash under stress - like running Prime95 and say a 3d mark at the same time or individually. So far the machine is rock solid running those tests. Whenever it crashes the machine is only running IE V10. Nothing is started in the background except for Charter Internet's Security Suite.
09-02-2013 11:52 PM
Agree that it is not a stress-induced issue. Probably a driver or memory problem. I am only running HP hologram label memory, and in xw6400 the sweet spot seems to be 4x2 for 8 total. You get a "quad channel" result in BIOS, and I have 4GB sticks available but really can't see a performance boost for my needs. I have one xw6600 running 16GB...... no big deal found.
Agree on the video card..... attached are the results from this xw8400. Very nice video scores from the newer nVidia cards, and they don't have to be Quadro as you'll see. When you load the drivers fresh you only need the default checked single option under custom.... that will get you the driver and the control panel. You can add back in later. Check the clean install box on that page if it is available to you.
I can attach my latest xw6400 BIOS replicated setup for you if you think a BIOS issue might be part of the problem, for you to try.
By the way, the processor/memory scores represent two of those X processors, and if the drive on this box (or the xw6400) was an Intel G2 or G3 SSD the drive score would have been 7.6-7.8. The G2 ones seem to be a bit faster, getting 7.7-7.8 consistently on these xw workstations.
09-03-2013 05:08 AM
I think the memory should be ok - ran MEMTest86 for hours without a glitch - also it was swapped out with a different set and still got the same error. Now running matched Samsungs 8Gb total 2GBX4. Was using Nvidia's Beta Driver - now clean install WhQL. If errors return I'll swap out the GTS450 from my XW8600 to see if the cards memory or the card itself is bad. Even when doing any of the tests you never fully test all the card's ram. I'm pretty certain the processor/motherboard is ok unless I have a bad PCI-E (video) slot which is giving the error.
Since my crashes only occur once over a period of a week with the machine left on constantly it may be just pure luck that I can fix it since the error is impossible to create/replicate.
I do have one other question - do you recall the part number for the front PCI-card guide for the XW6400 that you can attach a fan? I tried looking on PartSurfer and eBay and can't seem to find a match for the Xw6400.
09-03-2013 02:21 PM
Here's what you need: 349571-001 and -002. Do a search on that first part number in this forum and you'll find some pictures that I've posted on the install method in the past. The fan has to be 92mm, and its thickness 25mm. It's a bit of a hassle to mount the fan in the black plastic holder, and be sure to mount it so it blows inward.
Those holders are pretty hard to find now, and not cheap. I personally like the Noctua 92x92x25 mm non-PWM case fan (NF-B9) for this purpose, and I've been using the included Noctua Ultra Low Noise Adapter attached to the lead, and from that to the bottom front motherboard fan header (both a LNA and ULNA come with the fan). These are virtually silent, and last a very long time. You shave off the very small lip on the yellow wire side of the ULNA's 3 wire plug end, and then it will slide properly onto the 4-pin motherboard header.
There is a "ghetto mod" that I've used with these particular fans, also with the ULNA. I can post more exact details if you're interested. This works great with the xw6200, xw6400, xw6600 as they all have the same front case perforation pattern. You likely also could use the same idea with the xw8400 and xw8600. For some odd reason HP uses an 80x80x25mm fan in the front fan holder for those. The pic below actually shows one of those Noctua fans mounted in the xw6x00's 349571-001, and the positioning of where the 3 machine screws will go through to match up with the fan mount holes...... no drilling needed this way. You can see that things still get great airflow the ghetto way:
You can trim off about 3/32 of the tip of the black plastic post pointed to below so it can't touch the fan blades.... the shoulders of the post rest against the outside of the metal frame hole, also pointed to. This trim job may not be necessary, and you can use a nail clipper for that if you need to. Do that only if things were to rub when you put the front face back on and fire it up.
You want to add on a fan grill on the inside face, as you can see to protect internal cables. The fans alway blow towards their labels on the hub, so point that label inwards when you mount it. The exact size of the machine screws and washers is important in that you get just enough wiggle room to mount the fan using 3 of the 4 mount holes as you can see. I can get that info, but my recall is that they're #6 wedge head stainless 1.5" long machine screws, with #10 fender washers so the wedge can nestle deep, and Ny-lock nuts (don't over tighten) with a #8 washer beneath the nut. All stainless from Ace, but stainess is not necessary.
Works great, and very cheap. Again, that NF-B9 non-PMW fan with the ULNA is the key... not too fast and not too slow this way. PWM version of that will run too slow even without an adapter.... I have tried it. And here's how it looks from the inside:
09-04-2013 05:32 AM
Thanks for the info. You are right the price for that piece of plastic is pretty high - $30 and your method will work fine.
I've been using Artic Cooling F8 CO PWM on my memory and the performance is excellent and its dead quiet. It does regulate properly and the airflow is excellent. The original fan does produce a ton of movement but the noise factor is just too high - high pitched whine. May replace the rears as I have the rare cage to fit the proper standard width fans instead of those thick 38mm guys. I tried the Noctua on the memory but I think the airflow is significantly better with the Arctic
Reinstalled Windows 7 this morning and I think I know what may be causing the problem. When I installed the machine in its current location I used the Intel Chipset Updated from Intel and installed their INF/CHIPSET which I suspect may be causing havoc with my system. When it was running as a test machine it had all the Windows updates, graphic card updates and was running fine. Had issues with going to sleep properly so I installed their chipset update. Turns out the problem was a Microsoft USB Mouse/keyboard adapter that was really misbehaving. Looked at the log yesterday and it did an unexpected boot yesterday and had some issues with the disk controller. Recalling that I did the update and if something was not right with that it would cause the problems I'm experiencing. Will see if things improve over the next few days after the reinstall. If this doesn't do it there is something really wacky with the hardware - either motherboard, PS or HD.
Will let you know - crossing my fingers.
09-04-2013 12:30 PM
There is the xw6600 memory fan cooling shroud that uses 92x92x25mm fan to know about..... I have converted the xw6400 type over to that on a number of my xw6400 builds because it runs quite a bit more quietly with more air flow ( but only if I can find that part cheap). It is 446342-001.... that is directly imprinted on the plastic. You'll see mistakes on some of the shroud numbers on eBay, but this is the correct one for that. Here's an example:
Those fit fine on a xw6400 also, and are made for 92x92x25mm fans. They usually came with the HP PWM NMB-MAT 92mm fans 3610RL-04W-B56, and the plug end is a standard type PWM white connector with the usual 3 total orientation ridges that we generally see these days on consumer grade PWM fans. It has a "FB1" modifier, as a result, on the HP fan label, on the hub. That is a very good fan for this purpose and fan holder.
The case fans (in contrast to the memory fan) for the xw6600 are the same NMB-MAT model number, but with a "FB3" modifier on the hub fan label (indicating the plug has only two edge orientation ridges and no ridge in the middle). The motherboard PMW headers for the two paired rear case fans and the single front case fans are identical on both the xw6600 and the xw6400, and thus you can take a FB3 version and it will fit on your front fan header, and you preserve PWM modulation of that fan. I still like the Noctua, but that NMB-MAT fan is pretty quiet, and it would let a person who is going to really ramp up a xw workstation for a particular project get that benefit via temporarily going into BIOS and setting all the fan speeds up simultaneously.
I use the FB1 version of that fan in my xw6600 memory fan shrouds whether it is on a xw6400 or xw6600 because I want the better cooling that the 92mm version of the fan gives with lower noise than the 80x80x25mm fan that fits in the xw6400's memory cooling shroud, and I also want to preserve the option of ramping up (via BIOS setting) the memory fan's RPM's if needed for a special high intensity project.
I agree on not using that generic Intel chipset driver loader...... On my clean W7Pro 64-bit builds I use what the W7 DVD loads, then update the Audio and Network drivers using the universal installers from the Z600 workstations that I've posted about here in the past. Those work great. The only added change is putting on the updated driver for the "RAID + AHCI" BIOS setting, also posted about here.
I use the latest nVidia drivers, and note that old or incorrect video drivers can make your sleep not work properly in W7, etc.
Finally there are the BIOS settings.... let me know if you want my latest via a HP replicated setup file.
03-02-2017 11:29 PM
I'm having the same problem with a E5345 processor:
Workstation boots fine under Linux with 4*4GB RAM.
When I run Prime95 with Maximum heat I get after few minutes:
922 Fatal Error on Front Side Bus 0
F1Err: Request/Address parity error
=> 922 is a Fatal error
PCB FAB REV: 1.09
Xeon E5345 SLAEJ (G0)
Is the motherboard Rev 1.09 known to work?
BTW: FMB-060 motherboard is also used on xw6600 as far as I saw it on Google image search. Is the xw6600 Motherboard just a new revision or what are the know differences? Any upgrade possible?
03-03-2017 05:48 AM - edited 03-03-2017 05:58 AM
In your situation your motherboard is showing incompatibility with the E5345. You are using the right version of sSpec code (SLAEJ). In this situation I just shift over to two E5160 dual core processors..... less cores but faster clock speed, so not a bad tradeoff. That is a lot easier than swapping out the motherboard.
Two E5160 processors will actually "feel" a bit faster than 2 E5345.
The xw6600 is a quite different motherboard, and its backplane is physically quite different from the xw6400. No, you cannot swap in a xw6600 motherboard to a xw6400 case. I would always buy a xw6600 over the xw6400 now. It runs a newer generation of processors and has 8 instead of 4 memory slots. It has a better memory controller and no similar issue to the E5345 in the xw6400. Same memory, so money saved there. Still a sweet workstation..... use Intel 300GB 320 series SATA generation II SSD.
Two of each:
E5160 for xw6400..... you want SLAG9 sSpec code.
E5450 for xw6600.... you want SLBBM sSpec code (fastest best official quad core).
X5270 for xw6600.... you want SLBAQ sSpec code (fastest best official dual core).
I recently swapped out two E5450 for two X5270 in a workstation I test on and found the higher clock speeds outweighed the loss of cores by doing that. Many programs don't need 8 total cores, but all programs like the higher clock speeds. The perceived benefit is not that big, however.
Currently for the best bang for my buck I'd buy a version 2 Z400 (for 1 processor box) or a version 2 Z600 and build up from there. Oh, and the Kingston Predator M.2 AHCI-based-controller PCIe SSD is working very well still on my souped up Z600. You just need the right storage controller driver, which I discovered and posted about here. That is a very fast workstation (and I have Z640s at work to compare to). It has been rock solid.
These still are SATA generation II workstations, and thus I still find that Intel 320 series SSD to be the best choice for any build on those Z400/600 workstations or the xw series.