Stupid old Mobo...HELP!?

[Bulova]

I'm thoroughly confused as to how anyone could make a design that works the way this seems to.

The main computer at home is a patchwork PC system with an ASUS A7V8X-X motherboard and AMD XP2500+ processor (though I'm unsure if it's a Barton or not...I think it's running at 333MHz FSB). I'm currently running WinXP(SP2) and have 1GB of RAM in one DDR stick. (Yes, DDR. It's old, like I said in the subject.) But when I'm running multi-Forefox windows (with multi-tabs, of course) my virtual memory blows up to as much as 1.8GB!

The Mobo has three DDR slots, so I figured I could just find a couple more DDR sticks and take care of the problem.

But the docs indicate something weird: "Max 3GB unbuffered PC2100/1600 non-ECC DDR SDRAM (Note: PC3200 maximum to 2 banks only. PC2700 maximum to 4 banks only)"

So unless my English is as poor as those ASUS Techs from Korea or someplace (and my Tech is suffering as well), what I get from this is that in order to actually put 3GB RAM, I have to slow down the memory bus to 266MHz from as fast as 400MHz (with a maximum of one double-sided, or two single-sided PC3200 sticks) or 333MHz (with two DS or one SS PC2700 stick). And then, the BIOS makes some reference to only running the memory at 333MHz with a processor that runs at that rate.

I know this can't have been a "real" hardware limitation and must just be something that could have been fixed in the BIOS...if ASUS had chosen to program the fix into a BIOS update at all. I mean, can it? Is there some intrinsic difficulty with getting three mem-slots to operate at the max speed of the bus? Or was ASUS just being a pain (in the AS)?

Paul

[Bulova]

There's good news and bad news.

The good: I wasn't confused. The board is exactly as the document described. (How I would describe the design engineer is another matter.)

The bad: Staying with the PC2700 memory as required by the BIOS to go with the AMD XP2500+ processor, I could get a second stick of 1G in, but when I tried the third, it would fail POST. On the other hand, even just that second Gig of RAM has the Virtual Memory size down to a manageable 768MB.

Anybody wanna buy a used motherboard?

[klz_fc]

If I'm not mistaken, the doc are probably referring to regular 32-bit Windows/Linux versions. 32-bit OSes can only support up to 3 gigs of RAM due to overhead and this is especially true if the video memory is not built in but shared with the main memory.

A couple of quick links from searches:

http://answers.yahoo.com/question/index?qid=20080115132745AAvqOtO
http://www.vistaclues.com/reader-question-maximum-memory-in-32-bit-windows-vista/
http://forums.cnet.com/5208-6142_102-0.html?forumID=5&threadID=114117&messageID=1296823

[Bulova]

If I'm not mistaken, the doc are probably referring to regular 32-bit Windows/Linux versions. 32-bit OSes can only support up to 3 gigs of RAM due to overhead and this is especially true if the video memory is not built in but shared with the main memory.

A couple of quick links from searches:

http://answers.yahoo.com/question/index?qid=20080115132745AAvqOtO
http://www.vistaclues.com/reader-question-maximum-memory-in-32-bit-windows-vista/
http://forums.cnet.com/5208-6142_102-0.html?forumID=5&threadID=114117&messageID=1296823

It fails POST before Windows starts to load. The docs to which I referred is the motherboard manual, making the issue a hardware, not software spec. And it does really seem that the bank-limitation is based on the memory bus speed it registers on initialization from the SPD spec of the memory, influenced by the processor spec, as well.

Seems to me like an asinine way to design a board and its BIOS, but there it is. Two Gigs DDR RAM work on the board at 333MHz, but it all chokes (that is, fails POST with a circular reset) when I put in a third. The doc says a third stick (and a max of 3 Gigs RAM) work if the bus speed is 266MHz (PC2100), but my current processor (AMD XP2500+) likes the bus rate to be 333MHz (PC2700). And so it goes.

Thanks for the link info, though...I'll check it out.

P

[Socket]

I have the same board and processor, but I run 3 512MB PC3200 chips.  I considered trying to max it out, but figured I shouldn't bother with such an old machine.  It overclocks to a 3200+ nicely though and HL2 Episode 1 runs almost cleanly...

I once dealt with a motherboard at work that wouldn't even power on with 4 1GB chips, but needed 3 in the right arrangement of the 4 slots.  Yes, hardware mfrs are that dumb.

[Bulova]

Some more info:

Seems that when I had 3x1G sticks installed, the POST problem was that it would actually recognize 3200+K of installed RAM. Somewhat more than the expected 3072K that 3 sticks of 1G should have delivered. And when POST counted to 3212M or thereabouts, the screen would go mostly green, aside from some artifacts where changing text would have been. And hitting the del key would still successfully get into the BIOS setup program.

So, a theory emerged: with less than 3G installed RAM, other component ROM and RAM (particularly dedicated video RAM) would be mapped in without conflict, but with 3GB installed, the memory controller would insert such RAM as part of system memory, mapping it higher than the 3G max noted in the Mobo doc, and choke trying to test it.

Then, looking in my old stash, I found one DDR stick marked 256M PC2100/266 MHz and one unmarked stick. I installed the UNMARKED stick and it is also 256M. I believe that memory is now overclocked to PC2700/333MHz but the system is stable so far with 2.25GB RAM installed and all the DDR slots occupied.

Aside from conflicting with the 128M of dedicated video RAM on my ATI Radeon 9600 card, I have no other theories. Clearly it had less to do with banks filled, and more to do with total memory. Still seems like a brain-damaged way to program a memory controller.

edit: I looked this over, and saw the end of the first, and last three paragraphs slashed-out. So I subbed "hitting the del key" where I had previously referred to it in a way that the forum interpreted it as a del code. I guess "< del >" is the same as "[ s ]" (sans the spaces inside the brackets).

update: I've also since upgraded the processor to an XP3000+; now every so often at times when Task Manager used to go all green, it now peaks at about 88%, so I do have some processing improvement.

conclusion: I "need" a new computer with Core 2 Quad.

[Socket]

My 2500+ is overclocked to a 3000+.  Oddly enough I've come to the exact same conclusion ;-)

[Bulova]

My 2500+ is overclocked to a 3000+.  Oddly enough I've come to the exact same conclusion ;-)

What did you find to be the safest way to OC the processor? Being more specific, was it by popping the internal multiplier, or by popping the external bus rate?

Earlier in the thread, you said you had the same mobo/proc overclocked to 3200+. Why did you ramp down to 3000+? What's the nominal clock speed of the 3200 (the 2500 is about 1.8GHz, the 3000 about 2.1Ghz)?

Do you think I could tweak the current 3000+ proc rate and OC it to about a 3600+? Should I boost the CPU voltage to get it to work? Will that blow out the PC2700DDR memory? And....what about Naomi?

For the answers to these and other questions ("Is anyone else still reading this thread?"), tune in tomorrow for the next episode of "Love...Of Chair Desktop (Computing)"

P

PS: Bonus points for anyone who remembers the source of, "Love...Of Chair. The story that asks the question, 'Can a boy from a small chair in a big room find happiness as...' (my most strongly remembered conclusion to this line is '...a stand-out at a sit-in?')"

[T_S_Kimball]

Regarding the Core 2 Quad, look here:  http://sungak.net/weblog/computer_specs

My current PC and links to the various posts regarding it.

--TSK

[Socket]

Actually, I had to double check my settings.  I had clocked it back down because it was unstable, but it was just really dusty and overheating in there.  Current settings:
External Frequency: 200 MHz
Proc. Multiplier: 13.0 x
Performance: Aggressive

The stock frequency settings on the mobo were 133 or 166 IIRC, so I matched it to the memory speed (PC 3200).  I also bumped the multiplier, but I left the voltage alone.  The bigger bumps come from the multiplier settings, as the math would suggest, but much higher than I have it and I would need to up the voltage to stabilize it.  At 2.6 GHz I already have a huge solid copper heatsink with Arctic Silver paste and an 80mm fan.

I'd say match your memory speed for external frequency and bump up the multi one number at a time.  After each bump, run something really proc intensive, check your temps, artifacts on the screen and such and if it's stable, bump it up again.  Push it till it won't boot right and pull it back a notch, then up the voltage a tiny bit and see if it stabilizes.  Repeat till you don't feel comfortable pushing it any further.  Keep it cool and you should be able to hit a 3600+.