xp32 + 4gb
- Thread starter Davros
- Start date
Bouncing Zabaglione Bros.
Legend
You shouldn't have problems. I've got a 4 gig XP32 system that works fine bar only seeing 3.25 gigs of ram. The extra ram is there for the future when I upgrade to a 64 bit OS, and it's more economical in the long run (DDR2 ram prices are going up now, and using only 2x2 gig ram sticks generally gives you better memory speeds than 4x1 gig at some later date). The extra 1.25 gigs over a 2 gig system comes in very useful most of the time.
Some people have the occasional issue with PAE extension drivers (Microsoft's attempt at dealing with the 4 gig limit on 32 bit systems where driver location is concerned), but I've never seen them.
Some people have the occasional issue with PAE extension drivers (Microsoft's attempt at dealing with the 4 gig limit on 32 bit systems where driver location is concerned), but I've never seen them.
I bnought 2x2gb sticks my bios reports 3072mb
really you'd think that the computer would use some sort of interleaving across the two channels.
really you'd think that the computer would use some sort of interleaving across the two channels.
Last edited by a moderator:
I just upgraded my Dell Inspiron E1505 from 2 x 1Gb to 2 x 2Gb on Vista32 and the performance change has been measurable. I get something like 3372 available for use...
Not entirely true, it depends on the density of the chips in use and the chipset involved. A double-density pair of sticks will go faster than four double-density sticks, but four single-density sticks will perform faster than a pair of double-density sticks and faster than a pair of single-density sticks on pretty much every Intel-based northbridge setup. I can't speak for AMD systems though...
These issues are near-entirely related to Windows 2000; PAE is force-enabled on all WinXP SP2 machines that have more than one logical processor installed. Thus, if you've got a P4 with hyperthreading or ANY of the Core 2 Duo lineup (or newer), you're already using PAE if you've got XPSP2 or later. It was a requirement of the Execute Bit Disable support...
Bouncing Zabaglione Bros.
Legend
Update your BIOS?
It does, but with four slots filled, most motherboards need to up voltage and reduce speed to get the timing stable. So two slots filled with larger ram modules are often faster than four smaller ram modules.
Bouncing Zabaglione Bros.
Legend
You'll find most motherboards will automatically slacken the memory timings with all four slots filled. You can check easily by just looking at the difference in the auto BIOS settings between two and four slots (though some motherboards do this silently). My P45 board does, which is why I went for 2x2 rather than 4x1.
... but that doesn't mean that everyone writes their drivers to be properly compatible with PAE. It's probably a bit of a moot issue now that XP has been around for so long, but I thought I'd mention it for completeness, though as I said, I've never seen any issues myself.
Still have to disagree with you...
The best link I can find at this moment is here: http://www.anandtech.com/showdoc.aspx?i=1839
Even in "auto" mode, four sticks of dual-sided DIMMs performed best, and two double-sided compared equally to four single-sided sticks. There are some other overclocking concerns linked to the timings you talk about, but not at 'auto' settings at stock clocks.
The best link I can find at this moment is here: http://www.anandtech.com/showdoc.aspx?i=1839
Even in "auto" mode, four sticks of dual-sided DIMMs performed best, and two double-sided compared equally to four single-sided sticks. There are some other overclocking concerns linked to the timings you talk about, but not at 'auto' settings at stock clocks.
Yeah I think because of that article I did get 1gb x 4 sticks IIRC.
Bouncing Zabaglione Bros.
Legend
You can check it yourself. Put 2 dimms in and look to see what the BIOS has set the timings to. Put four dimms in, and see the BIOS has slackened the timings all by itself. You can even look in the back of most motherboard manuals, and they will tell you that populating all four dimm slots will cause timings to go down.
Personally owning a Gigabyte GA-X38-DS4 and four sticks of Micron D9GMH memory, I can say with 100% certainty that absolutely no timings change from 1 stick, 2 sticks, 3 sticks or all 4 sticks installed on my system. And as for overclocking, all four sticks are rated at 4-4-4-12 at 800Mhz speeds, and yet all four installed will do 1200Mhz at 5-5-5-18 with a bit of extra voltage. Removing two sticks will get me all the way to a whopping 1225mhz, which is pretty much worthless in the grand clocking scheme of things.
So, if your system suffers from altered (relaxed?) timings with two vs. four sticks installed, I'd suggest looking into other issues such as non-matching DIMMs or the like.
It's certainly possible that some "lesser" boards out there may do such things, but I'd be quite surprised if any of the 865 and later Intel chipsets did it automatically.
To be a bit slanted / biased, this is something I'd expect to see from a cheaper NForce chipset, but only because I've had prior NF chipsets do stupid stuff somewhat like this to me in the far distant past. But I'm talking NF2 days, so even that's a bit of a stretch.
Edit:
Out of fairness to BZB, there are other timings than the 4-4-4-12, such as tRd and some others that I can't immediately think of. Not all boards expose these items, sometimes you need a software application to see them. However, my Gigabyte board exposes something like 12 different settings that all manage RAM and/or the northbridge and how it interfaces with RAM. But all twelve are visible in "auto" mode, and none of them change when memory slot qty changes.
To be a bit slanted / biased, this is something I'd expect to see from a cheaper NForce chipset, but only because I've had prior NF chipsets do stupid stuff somewhat like this to me in the far distant past. But I'm talking NF2 days, so even that's a bit of a stretch.
Edit:
Out of fairness to BZB, there are other timings than the 4-4-4-12, such as tRd and some others that I can't immediately think of. Not all boards expose these items, sometimes you need a software application to see them. However, my Gigabyte board exposes something like 12 different settings that all manage RAM and/or the northbridge and how it interfaces with RAM. But all twelve are visible in "auto" mode, and none of them change when memory slot qty changes.
I've only seen this behavior happen on the AMD Platform (Socket 939) with an NForce 4 chipset in relations to the command mode timing going from 1T to 2T. My Intel P965 and P35 boards do not exhibit this behavior.
Wow, so maybe I was actually right with my "cheapass nforce chipset" bias 
Not exactly a fair statement to NForce chipsets I realize; I haven't touched one in many years. I guess at least that explains where this discrepancy comes from though.
Not exactly a fair statement to NForce chipsets I realize; I haven't touched one in many years. I guess at least that explains where this discrepancy comes from though.
Crap I havent mucked around with the command mode stuff in a while or remember what the 4-4-4-12 timings really mean but I have to go home and check it out if I am in 1T or 2T mode...I do have the Nforce based chipset after all, nforce7 but an Nforce nonetheless...
Well, somewhat back to the original topic...
For those machines that may only have two DIMM slots (like my Inspiron laptop), a pair of double-sided DIMMs will certainly give you better performance than a pair of single-sided DIMMs.
Nevertheless, the performance difference between multiples of single and/or double-sided DIMMs was still relatively small (in terms of total percentage gain / loss), so I wouldn't worry too much about it unless you're one of those folks looking to squeeze out the very last drop of performance from your rig. In which case, you're probably already well aware of this to begin with.
For those machines that may only have two DIMM slots (like my Inspiron laptop), a pair of double-sided DIMMs will certainly give you better performance than a pair of single-sided DIMMs.
Nevertheless, the performance difference between multiples of single and/or double-sided DIMMs was still relatively small (in terms of total percentage gain / loss), so I wouldn't worry too much about it unless you're one of those folks looking to squeeze out the very last drop of performance from your rig. In which case, you're probably already well aware of this to begin with.
Blazkowicz
Legend
yes, it's tied to Execute Bit Disable, also called the NX bit (so if I want to nitpick, don't forget one-logical-processor systems such as AMD CPUs, VIA CPUs, core 2 celeron, and don't forget old SMP systems).
Thanksfully you don't really have to worry about driver compatibility in that mode :
http://technet.microsoft.com/en-us/library/bb457155.aspx
Similar threads
