The answer of the age-old question: 4x512mb or 2x1024mb?

Proof is not some synthetic benchmark to me. I dont rely on them at all, I use actual games. Sismark and others show lage drop in score going from 1T to 2T, games do not however. At least from the limited numbers I have seen.

Is he at 6, or higher?
 
fallguy said:
Proof is not some synthetic benchmark to me. I dont rely on them at all, I use actual games. Sismark and others show lage drop in score going from 1T to 2T, games do not however. At least from the limited numbers I have seen.

Is he at 6, or higher?
Click to expand...

If the synthetics, which are really sensitive to timings and such, show 3-3-3 1T at DDR500 to be better than 2-2-2 2T at DDR400, what makes you think games will show the reverse?
 
250mhz
11x cpu multi
200mhz (1:1)


3-3-3, 1T
3-3-3-1T.PNG



2.5-2-2, 2T
2.5-2-2-2T.PNG



to summarize, superPI, an extrememly memory performance sensitive program, shows 3-3-3 at 1T to be slightly better than 2.5-2-2 at 2T, given equivilent ram speeds.
if you were to compare 250mhz, 3-3-3, 1T to 200mhz, 2-2-2, 2T, the former would definitly be better, but still not a noticable amount in most, if not all, real world programs
 
robberbaron said:
If the synthetics, which are really sensitive to timings and such, show 3-3-3 1T at DDR500 to be better than 2-2-2 2T at DDR400, what makes you think games will show the reverse?
Click to expand...

I didnt say it would show the reverse. In Sandra, it shows 1T being much, much faster than 2T. However, in games, the difference is very small, just a few frames, out of a hundered. Again, thats from some limited benches on various forums. I have yet to see an "official" review of sorts between the two.

The pic above doesnt tell me which is faster. As I stated, my ram does Cas2 2-2-6. Not Cas2.5 2-2-10. Which may, or may not make up the difference from the 1T to 2T loss.
 
Dont get upset becuase I like proof for something Im being told. Its a simple question, that has not been answered yet. Is 2x1gig at Cas2.5 3-3-10-1T faster in games, or is 4x512 Cas2 2-2-6-2T faster. Nothing in this thread has answered that.

What is that you're doing the ram bench with? Ill do mine, and compare if to yours.

edit, Ill be down for a bit. Got my new PSU in, gonna rewire the case.
 
fallguy said:
Dont get upset becuase I like proof for something Im being told. Its a simple question, that has not been answered yet. Is 2x1gig at Cas2.5 3-3-10-1T faster in games, or is 4x512 Cas2 2-2-6-2T faster. Nothing in this thread has answered that.

What is that you're doing the ram bench with? Ill do mine, and compare if to yours.

edit, Ill be down for a bit. Got my new PSU in, gonna rewire the case.
Click to expand...

He's running ScienceMark.

(cf)Eclipse said:
oh boy.. sometimes i want to give up
Click to expand...


Don't worry about it. I see value in your work and that's what counts am i rite?
 
I never said I didnt find "value" in his work. I just didnt like his tude he got, because I wanted proof. He and I both had the same question, then for some reason he flipped. Here are my results:

2T.jpg


Unless I am doing something way differently than he is, my Cas2 2-2-6-2T results are much better than his Cas3-3-3-10-1T results. 5417 vs. 3376. I forgot to do it before I took a ss, but with the settings above, I got 30s in SuperPi.

edit, just noticed I had 128 bit in controller mode from A64 tweaker, his was 64 bit. Whats the diff?

Just checked the tracking, my ram is on the truck.. smooth BF2 here I come!
 
fallguy said:
edit, just noticed I had 128 bit in controller mode from A64 tweaker, his was 64 bit. Whats the diff?
Click to expand...
i have single channel.. you have dual ;)

and that was my 'quick proof'. it might not be enough, but i was hoping you could extrapolate the results and consider how superPI is effected compared to other programs, and see that there is next to no difference :D
 
(cf)Eclipse said:
and that was my 'quick proof'. it might not be enough, but i was hoping you could extrapolate the results and consider how superPI is effected compared to other programs, and see that there is next to no difference :D
Click to expand...

How was it proof? You didnt show the speeds I was wondering about. In fact, you were wondering the same questions as me, on the first page.

robberbaron said:
Eclipse proved that the 1 gig sticks will beat out 2T with 2-2-2. Crucal Ballistix are the best for latency at high densities, beating out 4x512 just because 2T tends to be more severe than going from 2-2-2 to 3-3-3.
Click to expand...

As I stated above, nothing was proven. You were not at 2-2-2-6-2T. Not even close really.
 
Just installed a second 2x512 sticks giving me 4x512mb Crucial sticks running at 200mhz dual channel. Can't belive its still dual channel. Thought it wouldnt work like that . I have been getting strange lockups. Any ideas?
 
Ok, I'm bad at following posts so I have questions. If I were to install 4x512mb single-sided RAM, would that allow me to run at 400mhz? What would the performance hit be like? And would I still be able to run Dual Channel?

I read somewhere on here that Dual Channel gets diabled when theres more than 2 dimms, yet my mobo manual says I'm able to use Dual Channel with all slots filled. Would appreciate it if someone cleared all this up for me.
 
rohcky said:
Ok, I'm bad at following posts so I have questions. If I were to install 4x512mb single-sided RAM, would that allow me to run at 400mhz? What would the performance hit be like? And would I still be able to run Dual Channel?

I read somewhere on here that Dual Channel gets diabled when theres more than 2 dimms, yet my mobo manual says I'm able to use Dual Channel with all slots filled. Would appreciate it if someone cleared all this up for me.
Click to expand...

What core?
 
It seems like the performance difference is so slight between 4x512 and 2x1024 that the latter would be the clear winner for future upgradability.

Everyone seems to think that the Ballistix memory is the best performer for 2x1024. But it's so damn expensive. :( Someone asked about OCZ performance memory (the one with 2-3-2-5 timing, not the 3-3-3-7). Is that a good value and best way to go for ~$250?

BTW, a friend told me about PDP XBLK model that has the Samsung TCCD memory chips, saying their much cheaper than the competition but still have the good chips. (But he was referring to their 2x512, which I assume is also applicable to their 2x1024.) Any thoughts?

Thanks.
 
rohcky said:
Winchester
Click to expand...

Gee thanks for getting a core that was made during the transition between stupid and good memory controllers! :p

Alright, winchester has a good chance of doing it. I've messed with two Winchesters in my day and I was able to get them both to DDR400 2T with 4 dimms.
 
I pulled the trigger on 2x1 GB OCZ 2-3-2-5 Platinum RAM. I would've been better off hitting a bar and getting a few beers. It would've been much cheaper than this RAM. :D
 
robberbaron said:
Gee thanks for getting a core that was made during the transition between stupid and good memory controllers! :p
Click to expand...


Yea... I was stupid. My friend asked me for tips on building a pc. That got me excited and I decided to build myself a new one (recently moved out and left old one with my 'rents). So it was a spur of the moment thing and I didn't research the new cores. Ended up with the Winchester about a week before the Venice came out... *sigh*
 
rohcky said:
Yea... I was stupid. My friend asked me for tips on building a pc. That got me excited and I decided to build myself a new one (recently moved out and left old one with my 'rents). So it was a spur of the moment thing and I didn't research the new cores. Ended up with the Winchester about a week before the Venice came out... *sigh*
Click to expand...


Nah, Winchester is cool. Literally. It runs so much cooler than most A64's (in my experience.)

And, if you got a CBBHD, you're set!
 
(cf)Eclipse said:
if he got it a week before venice, a cbbid is likely :(
Click to expand...

I looked at a boxed 3200+ today at work and it had a BI. I looked at the stepping and saw CBBID and thought "sweet!" Then I rememebered a half second later that CBBID isn't "sweet." CBBHD... its a shame those days are gone.
 
ok,
Now I'm just TOTALLY confused :confused:

My rig: Venice 3000+, DFI NF4 SLI-DR
CURRENT RAM: 2 x512 Corsair XPERT3200
2.5-3-3-8-1T @ 256mhz, 2.8 Volts

XPERT timings:
# Timing: 2-3-3-6
Voltage: 2.75V
# Model #: TWINXP1024-3200C2

My BF2 was taking 1.5Gigs Peak Mem, so I want to go to 2 Gigs. BUT, I want to keep my OC on my Venice (currently at 2561 mhz).

I also have in my posession, 2 MORE sticks of XPERT for a 512x 4 combo, and I also have 2 x 1GB OCZ Platinum w/ these specs to try as well:
# Cas Latency: 2
Heat Spreader: Yes
Manufacturer Warranty: Lifetime
Timing: 2-3-2-5
Voltage: 2.6V
# Model #: OCZ4002048ELDCPE-K

So, before I start pulling my hair out and making mad changes to my DFI BIOS, in your humble opinions, which setup should I try to maximize with my rig? 4x512 Corsair or 2 x 1GB OCZ?

I'll sell the XPERT and RMA the other XPERT if I need to, or keep the XPERT and RMA the 2GB OCZ. I just want my BF2 stable without killing my OC or mem bandwidth too bad.

THANK YOU GUYS!!!
 
typhoon43 said:
ok,
Now I'm just TOTALLY confused :confused:

My rig: Venice 3000+, DFI NF4 SLI-DR
CURRENT RAM: 2 x512 Corsair XPERT3200
2.5-3-3-8-1T @ 256mhz, 2.8 Volts

XPERT timings:
# Timing: 2-3-3-6
Voltage: 2.75V
# Model #: TWINXP1024-3200C2

My BF2 was taking 1.5Gigs Peak Mem, so I want to go to 2 Gigs. BUT, I want to keep my OC on my Venice (currently at 2561 mhz).

I also have in my posession, 2 MORE sticks of XPERT for a 512x 4 combo, and I also have 2 x 1GB OCZ Platinum w/ these specs to try as well:
# Cas Latency: 2
Heat Spreader: Yes
Manufacturer Warranty: Lifetime
Timing: 2-3-2-5
Voltage: 2.6V
# Model #: OCZ4002048ELDCPE-K

So, before I start pulling my hair out and making mad changes to my DFI BIOS, in your humble opinions, which setup should I try to maximize with my rig? 4x512 Corsair or 2 x 1GB OCZ?

I'll sell the XPERT and RMA the other XPERT if I need to, or keep the XPERT and RMA the 2GB OCZ. I just want my BF2 stable without killing my OC or mem bandwidth too bad.

THANK YOU GUYS!!!
Click to expand...

I am afraid you are going to have to sacrifice something: ram size or ram bandwidth,
You can go with either 4x512 or 2x1G and the results would be similar, and mostly will depend on what kind of ram you are putting in. 4 sticks of 512 would overclock to, lets say, half of what you normally do with 2 (to about 230 I would guess). 2x1G sticks don't normally overclock as well as 512 sticks. You might get lucky, and your ram might be the bomb, and you could do 255, but there are no guarantees. The only 2x1G sticks that will do 255 consistently are Ballistix, they are also 400 bucks for a pair. I just ordered 2x1G of Mushkin HP Blue. I will see how they do (some people report these to go up to 260 or so). So, the reality of the situation is, you MOST LIKELY (BUT NOT 100%) have to drop to a higher multiplier and run your ram at 233.
HOWEVER, I would strongly recommend going with 2G of ram. I've been running 4x512 for a year and I can't be happier even with a slightly slower ram. The beauty of 2G is that you can turn off virtual memory(paging file to zero size) and everything suddenly becomes super smooth. Desktop, game loading, playing games. Just tested with and without virtual mem in doom3 and hl2, since I just built this new box. The difference was amazing. Constant loading WITH virtual memory must not have been so apparent before I went to 2G, but now I can't stand it. WITHOUT virtual memory, things were smooth as glass playing inside those games. Of course, with or without virtual memory, loading of games is ridiculously faster. I remember how everyone complained loading times when Vampire Bloodlines came out. People were saying 1 - 2 min level load times, and I was sitting with a happy 10-15 sec. It was like that with every game that had long loads for other people.
Of course, preferences differ for different people.

Disclaimer: I've only tested one set of 4x512 sticks with 2G and 1G configurations on I875p and Nforce4 SLI platforms. The rest of the information on possible overclocks comes from countless hours of reading other people's posts on 2G setups in this forum, ocforums and xtremesystems forums.
 
^ That brings up an interesting question. For those who have 2 GB of RAM, is it recommended (or even safe) to completely disable the PC's use of virtual memory by setting paging file size to 0 MB as the intial and maximum size?
 
From what I've read, it is not reccomended to disable the paging.

The NT5 (2000, XP, 2003) Paging is very efficient, and does a pretty good job never tossing anything to disk when there's more headroom in the RAM.

Also, I seem to remember that with NT5 Paging, there is always a page file dynamically created, even if you've disabled it, but I may be mixing memories by this point.
 
I haven't done a lot of reading on this subject, but I HAVE done extensive testing.
About a year worth of testing, in fact. In this time, I'm yet to run out of physical memory with 2G of ram and zero paging file size. By mistake, I have even managed to run 3 games at the same time (sacred underworld, quake3 gtv, sof2 online play) once with no bad consequences.
At the same time, when virtual memory was enabled, there was definite HDD thrashing even with 2G of ram in most game I have tried. So, while microsoft can do all the shouting and arm waving about the efficiency of their virtual memory, it is an undeniable fact that it is not efficient.
Of course, I have been running games with max quality settings and high resolutions (1280x or 1600x) as I've been using x800xt pe, and now upgraded to 7800gtx sli. For those with lower end video cards, the HDD thrashing may not be as severe, as the poor souls in question would be forced to use lower quality/resolution settings in games.
It is disappointing that people who don't have any experience running 2GB setups always make it their business to step in and educate the rest of us on "potential dangers" of running a computer without virtual memory. Look fellas, I use my computer for web browsing, email, word editing , powerpoint presentations, and last but not least, gaming. I do not do any professional work that requires more than 2G of ram such as scientific computations or 3D design and modeling. As MY OWN EXPERIENCE has taught me, 2G of memory has been more than sufficient for what I do with my computer. And while some of us may need more in a particular situation, I still have to advocate the use of 2GB of physical ram (with or without virtual memory), as the benefits of having more ram is significant to justify the cost (monetary or lower bandwidth for overclockers).

P.S. At the office, I use a computer with 512M of ram, and that's sufficient for what I do for work.
 
This is an interesting dialog, gentlemen, and one that I plan to research a little and also experiment w/ my system ( a lot), much like how acrh2 has done.

I know in the past, I've read that one should set the initial paging size and the maximum paging size to be the same value. I think we would all agree on that, which would minimize the disk fragmentation done by virtual memory. Also, I read that the paging size should be set to three times the physical memory. That kind of confused me since I figured if one had more physical RAM, it would obviate the need for virtual memory.

Now w/ my 2 GB of physical RAM, if I follow past convention and triple that amount for virtual memory, I'd have to "waste" 6 GB of hard drive space for memory. That ain't gonna happen. So I'll try 0 MB like acrh2 and then maybe try 1 or 2 GB of virtual memory and see what kind of performance I observe. I'd also be interested in others' opinion as well.
 
so confusing. But for Intel, would i want 2x1024 or 4x512? i know 1T and 2T doesnt make much of a difference but dual channel does in intel. Would it run in dual channel with 4x512?
 
AznAnarchy99 said:
so confusing. But for Intel, would i want 2x1024 or 4x512? i know 1T and 2T doesnt make much of a difference but dual channel does in intel. Would it run in dual channel with 4x512?
Click to expand...
Yes, provided two 2x512 pairs can run dual channel.
 
beowulf7 said:
This is an interesting dialog, gentlemen, and one that I plan to research a little and also experiment w/ my system ( a lot), much like how acrh2 has done.

I know in the past, I've read that one should set the initial paging size and the maximum paging size to be the same value. I think we would all agree on that, which would minimize the disk fragmentation done by virtual memory. Also, I read that the paging size should be set to three times the physical memory. That kind of confused me since I figured if one had more physical RAM, it would obviate the need for virtual memory.

Now w/ my 2 GB of physical RAM, if I follow past convention and triple that amount for virtual memory, I'd have to "waste" 6 GB of hard drive space for memory. That ain't gonna happen. So I'll try 0 MB like acrh2 and then maybe try 1 or 2 GB of virtual memory and see what kind of performance I observe. I'd also be interested in others' opinion as well.
Click to expand...

That whole "your page file should be x times as much as your RAM" thing is bunk. With so much RAM in systems these days, it's about the total size of your RAM and page file. If you've got 2gb of RAM, you're gonna have PLENTY to spare. (unless you're running 2 instances of BF2... lol) Just make sure that the total of your RAM and page file is higher than the MAX memory used at one time in your system. It's that simple. So... with 2gb of RAM, you should be able to get away with a very small page file or none at all. I've read that some apps hate not having a page file, but I have not seen a problem with zero page file personally. I'm a gamer/internet surfer/e-mailer and I don't do multimedia stuff on my PC. (just to give you an idea where I'm comin' from)
 
WhyYouLoveMe said:
That whole "your page file should be x times as much as your RAM" thing is bunk. With so much RAM in systems these days, it's about the total size of your RAM and page file. If you've got 2gb of RAM, you're gonna have PLENTY to spare. (unless you're running 2 instances of BF2... lol) Just make sure that the total of your RAM and page file is higher than the MAX memory used at one time in your system. It's that simple. So... with 2gb of RAM, you should be able to get away with a very small page file or none at all. I've read that some apps hate not having a page file, but I have not seen a problem with zero page file personally. I'm a gamer/internet surfer/e-mailer and I don't do multimedia stuff on my PC. (just to give you an idea where I'm comin' from)
Click to expand...

If by multimedia stuff you meant something like reencoding full length dvd with DivxToDvd, Tmpgenc, or Virtualdubmod, then 2 GB physical + zero virtual memory works for such things just fine.
 
acrh2 said:
If by multimedia stuff you meant something like reencoding full length dvd with DivxToDvd, Tmpgenc, or Virtualdubmod, then 2 GB physical + zero virtual memory works for such things just fine.
Click to expand...

Yeah, that's what I meant. Then zero it is, baby!! :D
 
^ Great, I'm going to try it out w/ 0 virtual memory, at least initially! :) I'm going to use Windows XP Pro x64 that I have, or if I notice driver problems, then fall back Windows 2000. I figure either way, it should work since both of those OSs have pretty good memory management vs. crap like Windows 9x/ME/NT.
 
Good question, but I can't answer as I have never pulled the heat spreaders and they are comfortably in my SN25P.
 
You must log in or register to reply here.
Back
Top