geoduck

Hi, I just noticed your post and I have been having the same problem on the Expert board running 64-bit Linux. I don't have the problem when running 32-bit Windows XP (even with the memory hole enabled), but of course that doesn't address all of the memory. I was just about to install the eval version of 64-bit XP to test it but I think you've saved me the trouble. So have you found any sort of solution (other than install a separate NIC)?

It sounds like from this thread at AMD forums that memory remapping to get around all of the device addresses in the 3.5-4GB range is a function of the memory controller on the CPU (rev E supports it) but that the BIOS and chipset has to actually allow the remapping. I looked at a number of other RS482-based boards and as far as I saw none of them had a memory remapping feature in their BIOS. I've asked at Rebels Haven if there is a modded BIOS for the RS482 that will allow this. In the meantime I'll play around with the PCI/PCIE options to see if I can get anywhere, and I'll PM Happy_Games or RGone to get their input.

These are the replies I received from DFI tech support (I hope it's ok I'm reposting these here): This doesn't really answer why the NF4-SLI chipset gives access to all 4GB and the RS482 doesn't as presumably they are both 64-bit designs since they are for the AMD64 architecture. Then when I asked whether it was a chipset limitation or a BIOS problem I received: This doesn't seem to be a definitive answer regarding chipset vs BIOS and whether it can be fixed for the RS482 chipset as it is for the NF4-SLI chipset. How would I go about asking Happy_Games or RGone specifically for help on this matter? It would be great if they could add anything to the discussion that DFI tech support hasn't, like perhaps if there are any modded BIOS's for the board that might fix the problem? I have Tmod's latest BIOS CD and I didn't find any RS482 BIOS that indicated it fixed this issue. I'm planning to reconfigure my machines tomorrow night so any info that would keep me from having to do that would be greatly appreciated.

Thanks for your input. This actually has nothing to do with Windows since no Windows CD has been within 10 feet of this machine. I've tried 64-bit linux and a bootable Memtest CD (which handles all 4GB on my other machine) and they both show 3.5GB, as does the BIOS. So it appears to be BIOS related since it shows up in the initial memory detection, and unfortunately DFI doesn't seem to want to do anything about it. Don't get me wrong, DFI technical support was great in replying promptly to my support request with an explanation of why I only see 3.5GB, but they didn't offer a solution other than to try a different motherboard and suggest I should be using a workstation class motherboard (which I probably should be, but I'm trying to do this on a limited budget). I'm not going to push them because I don't have the time to wait while they mess with the BIOS. As I stated before, my Lanparty UT NF4 SLI Expert board has an option in BIOS to enable 'Memory Hope for PCI MMIO' and this allows it to see all 4GB of installed memory, but the RS482 doesn't have any such option (although it has one entitled 'Memory Hole' but this seems to be something entirely different.) My Gigabyte NF4-SLI based board also has an option to enable use of 4GB so I'm just going to switch my Opteron 165 and 4GB of memory to it and put my A64 3000+ with 2GB in the RS482.

Thanks for your responses. I've been reading through the thread and just received a response from DFI technical support that says basically the same thing. I get the same ~3.5GB whether I look in BIOS, boot 64-bit linux, or run Memtest from a CD. The thing is that I have another machine running a Lanparty UT NF4 SLI Expert mobo with 4GB and it sees all of the memory (after enabling the 'Memory Hole for PCI MMIO' option in the Advanced Chipset Features. A similar feature in the RS482 BIOS appears to be unrelated to running 4GB.). So is this a limitation of the RS482 chipset or of the current BIOS for the machine? Should I look for another NF4 based motherboard? I have yet another machine running with 2GB using a Gigabyte GA-K8NXP-SLI board that from what I've read in other forums will support the full 4GB, so perhaps I will do some swapping and put the 4GB and Opteron 165 in that machine and run the Athlon 64 3000+ with 2GB in the RS482 board. I think the RS482 will overclock better than the Gigabyte so I was hoping to run the 165 in it and hit a high overclock. Oh well. Well, I've been to a lot of places in B.C. because I'm from Seattle originally, but I've never been to Westbank. Anyway, the problem is that I can't get the BIOS to see the 4GB with this board. I run 64-bit linux so the Win32 limitation doesn't effect me but in BIOS and when I boot linux or run Memtest from a CD it only shows around 3.5GB of memory. This also brings up a question about the specifications for the board. In the manual and on the DFI website it says that this board 'Supports up to 4GB system memory', which I take to mean that you should be able to see all 4GB. Call me crazy.

It's a quantum chemistry program called Molpro. The memory requirements vary according to what molecule is being calculated and what level of theory is being used. It just happens that my latest project needs a minimum of 3.8GB and takes about one week on my current machine to calculate the energy for a single geometry. I plan on calculating many geometries so it would be nice if I could have two machines working on it at once. Thanks for your interest.

I'm having trouble getting my new system to recognize the full 4GB of installed memory. I haven't installed an operating system yet but in the BIOS it shows the total memory as 3505152 K. This doesn't change whether I have the 'memory hole' enabled or disabled. I booted 64-bit linux from a DVD and it only sees 3.5GB as well. The only mention I've found of this problem here is in this link. I'm running an application that requires 3.8GB of memory so I really need the full 4GB. I've had no problems running a similar setup in my older machine with a Lanparty Expert motherboard. Thanks for any help!

Flashed the beta BIOS (3/29) and everything seems good, and in fact it has solved my video issues described in: http://www.dfi-street.com/forum/showthread.php?t=47450 http://www.dfi-street.com/forum/showthread.php?t=46989

This issue has been resolved with an update to the latest beta BIOS (3/29) for the Expert board.

In an act of desperation before I went out and bought an nvidia-based video card I flashed the lastest beta BIOS (3/29) for the Expert and the problem has been magically solved. Of course, there was a moment of terror when the machine kept rebooting before the BIOS screen, but this was traced to the old PCI video card I was using which suddenly became unusable with the new BIOS. Once I reinstalled the X550 it worked fine. Hope I don't need the old PCI card again!

I was told by Diamond technical support that this video card comes in two versions, one with a cooling fan and one with a big heatsink, and I needed to swap the one I had (fan) for the other (heatsink). I RMAd the board and received the new heatsink version today but it is behaving exactly the same way. I'm not booting into Windows (I don't even have it installed yet) so I don't think it is a driver problem. Diamond technical support previously told me it was a problem with my system, that I didn't have the right chipset driver or something. I'm just trying to run memtest from BIOS so does it need a chipset driver for that? It also does the same thing when I try to boot from a bootable floppy or a bootable CD. I had an old PCI video card installed and it worked fine but I don't want to live with that card forever! Am I missing something in BIOS? I even set the PCI express frequency to 105 MHz as suggested in http://www.dfi-street.com/forum/showpost.p...75&postcount=15 Any suggestions (other than simply getting a different card, which is my final option)?

RGone, thanks for the great reply. Yes, it seems the definitions of common words take on new meaning in the computer industry. How true. I've encountered this far too many times myself. Must go to sleep now...

Yes, they were all open to me but I chose the DFI expert board for specific features it had that the others didn't. I made the mistake of assuming that they supported Opterons because the other companies did, but I'm not saying I made a mistake in buying the DFI board because so far I'm very happy with it and in all likely hood the issue of support will never rear its ugly head for me except in this thread. Actually, It seems to me that good companies listen to their customers and respond. Ignoring consumers has put a lot of companies out of business. It's still unclear to me how AMD can wield so much power over DFI when DFI is an independent company. I mean AMD can rant all it wants about reserving Opterons for servers and workstations but how does that affect DFI? I might just be ignorant of the ways of business, but can someone clear this up? I don't want to make anyone mad here because DFI-Street has been a tremendous resource that I intend to use (unsupported or not) repeatedly as I set up my first DFI system, so please take my comments as curiosity-driven and not just inane ranting (which they very well may be because I've been up all night and now the sun is coming up).

I know that DFI does not support the use of Opterons in their desktop motherboards but my question is: why not? The official line from the DFI FAQ is 'AMD does not recommend using the Opteron or other " mobile" CPU on any desktop motherboards' and so DFI does not support their use, either. The Unofficial Opteron 939 Thread states 'THIS NON-SUPPORT STATUS HAS BEEN MANDATED BY AMD'. Why then, do Asus, Gigabyte, Foxconn, and MSI all include Opterons on their lists of supported CPUs for their desktop boards? Why does AMD control DFI's support and not these others? As I was gathering the pieces for my system I was trying to decide between an Asus board and a DFI board. I knew that Asus supported Opterons and I assumed (my mistake) that DFI would as well. I haven't encountered any problems with my Opteron yet, but it is very disturbing that DFI does not support the processor that probably half their customers are using. Could it be that this significantly reduces the cost of support for DFI? If they can dismiss support out-of-hand for half their customers it has to save them a lot of money. Sorry to be so cynical but DFI's reasons for not supporting Opterons just don't ring true in light of the other manufacturers' support. I don't want to start anything bad here, but it seems a discussion of this lack of support is in order. If DFI really is the board for enthusiasts they should offer at least the same level of support most other major board makers do for enthusiasts' favorite CPU.

I apologize if this is too simple or obvious, but I had a similar problem and it turned out to be a short between the back of the motherboard and the case. There was an extra motherboard mounting standoff in my case that didn't line up with a mounting hole and because it was pressed in and not easy to remove I simply covered it with electical tape. Unfortunately, a sharp protrusion on the back of the motherboard poked a hole in the tape and caused a short, leaving me with a board that showed no signs of life when I tried to turn it on in the case but worked fine out of the case. I finally made a homemade puller and pulled the extra standoff out of the case and now it works in the case (except for the weird video problem described here). Does your board work out of the case now, after your problem with it in the case?