memory timings vs. memory speed

CDsDontBurn · Jan 13, 2005, 03:53 AM

my question is, what do you guys think is better: memory timings, or memory speed?

The other day, i ran PCmark04 and got a score of 58xx (in my sig). Well, i've got 1Gb Corsair XMS DDR400, and when i ran that benchmark i had my timings @ 2-3-3-6 1T 5:4 ratio and my cpu running at 3.9Ghz. Well, my friend ran PCmark04 on his system and he has 1Gb Corsair XMSpro DDR500 running @ 2.5-4-3-5 1T 1:1 ratio and his CPU running at 3.6Ghz. He ran PCmark04 and his system w/his specs scored about 150pts. less than my system. The only reason why my system got a higher score than his was because of my 300Mhz higher OC.

Here is my friend's score

I've always come to believe that tighter memory timings are better, but if running looser timings and a 1:1 ratio is better, i'll start saving up money for some new RAM.

Iridium · Jan 13, 2005, 09:32 AM

From what I have read at the different forums, the only time that the really loose timings start help your performance with a 1:1 ratio, is when you start pushing the upper limits of the FSB, like around the 270 FSB mark. You are in the same boat as me, you have a 3.2ghz processor, although your Prescott will clock higher than my Northy, you are running on air which will limit the max speed you will be able to clock it at, and are already close to the upper limits of the 3.2e that I have seen on other forums. The best bet is to get as high of a FSB with as tight of timings that you can find....

So basically if you can find some lower latency higher speed DDR like your buddy has, you would be better off. From what I have been told running at 1:1 is always a big plus. This is a great website for tech viewpoints and drivers etc. and I am all for tryinig to answer the question "in the family" here at DH but there are other forums that have more "specific" overclocking discussions for the different kinds of tech in your PC broke into categories. Personally I am not sure how much more room you have left in your CPU if you continue to cool with air. I would do some looking around the web and see what some other people have been saying about the same issue. Although I am sure there will be some good answers and viewpoints brought to the table here too.

~~Edit: How high can you run up your current ram at the loosest timings at the 1:1 ratio?

Vampyromaniac · Jan 13, 2005, 01:10 PM

Quote:

Originally Posted by CDsDontBurn

I've always come to believe that tighter memory timings are better, but if running looser timings and a 1:1 ratio is better, i'll start saving up money for some new RAM.

Is 150 "points" in a synthetic benchmark really worth it?

CDsDontBurn · Jan 13, 2005, 01:16 PM

Quote:

Originally Posted by Iridium

~~Edit: How high can you run up your current ram at the loosest timings at the 1:1 ratio?

i can't get too high on 1:1 ratio w/loose timings. Around 205~208FSB on a 1:1 ratio before my system becomes unstable.

Quote:

Originally Posted by Vampyromaniac

Is 150 "points" in a synthetic benchmark really worth it?

although a synthetic benchmark, it also reflects on actual gaming performance. the higher the score in a synthetic benchmark, the more FPS one will have while playing a game.

Vampyromaniac · Jan 13, 2005, 01:55 PM

Quote:

Originally Posted by CDsDontBurn

although a synthetic benchmark, it also reflects on actual gaming performance. the higher the score in a synthetic benchmark, the more FPS one will have while playing a game.

True, but by that logic, you'd more appropriately test with 3DMark rather than PCMark

Also, gains shown in synthetics are almost always more than gains in the real-world. So a 2 or 3% gain in PCMark, may be 1 or 2% in real-world . . . so even if you're optimistic, 103fps vs. 100fps . . . again, I ask: is it worth it?

Iridium · Jan 13, 2005, 01:58 PM

Quote:

Originally Posted by CDsDontBurn

i can't get too high on 1:1 ratio w/loose timings. Around 205~208FSB on a 1:1 ratio before my system becomes unstable.

Hmm that seems pretty low for ddr 400 ram...I can get my Kingston valueram ddr 333 to run at 220FSB at a 1:1 ratio on my board at 2.5/3/3/6...

I don't know what to say about that. Perhaps you can run memtest to see if they are faulty, it would appear so to me. If even at loose timings you can't get your PC3200 ram to run at only 5~8mhz higher than stock FSB. I could see 5~8 at tight timings but when you loosen it up, I would think they would be able to shoot up around 225-235 pretty easy.

Have you made any adjustments to your "Game Accelerator" in the "Advanced Chipset Features" menu of your BIOS? I have read that recommended settings for overclocks are Auto/Normal/Auto/Disabled/Disabled for improved overclocked system stability with your board. Along with a couple of other tweaks.

Iridium · Jan 13, 2005, 02:05 PM

Quote:

Originally Posted by Vampyromaniac

True, but by that logic, you'd more appropriately test with 3DMark rather than PCMark

Also, gains shown in synthetics are almost always more than gains in the real-world. So a 2 or 3% gain in PCMark, may be 1 or 2% in real-world . . . so even if you're optimistic, 103fps vs. 100fps . . . again, I ask: is it worth it?

I fall victim to the benchmarks myself. Is it worth it, probably not. But if your pockets are deep enough...

I myself think that he will show some serious real world gains if he can get his ram to run at 1:1 ratio with his CPU at 244 FSB

vs the 195 FSB he is running at now, I would think there would be more than a 3~4% gain with an increase of around 50mhz on his memory. The fact that his ram doesn't run much over stock FSB speeds at loose timings with Corsair XMS makes me think something is up with his ram, or settings somewhere...

CDsDontBurn · Jan 13, 2005, 02:07 PM

Quote:

Originally Posted by Vampyromaniac

True, but by that logic, you'd more appropriately test with 3DMark rather than PCMark

Also, gains shown in synthetics are almost always more than gains in the real-world. So a 2 or 3% gain in PCMark, may be 1 or 2% in real-world . . . so even if you're optimistic, 103fps vs. 100fps . . . again, I ask: is it worth it?

well, i'm not saying that from 100FPS to 103 FPS is worth it, i'm talking about overall performance. Framerates will always vary when playing a game. When framerates drop significantly, having those extra 3,4, or even 5 frames higher will be the difference between playable, and choppy.

Quote:

Originally Posted by Iridium

Hmm that seems pretty low for ddr 400 ram...I can get my Kingston valueram ddr 333 to run at 220FSB at a 1:1 ratio on my board at 2.5/3/3/6...

I don't know what to say about that. Perhaps you can run memtest to see if they are faulty, it would appear so to me. If even at loose timings you can't get your PC3200 ram to run at only 5~8mhz higher than stock FSB. I could see 5~8 at tight timings but when you loosen it up, I would think they would be able to shoot up around 225-235 pretty easy.

Have you made any adjustments to your "Game Accelerator" in the "Advanced Chipset Features" menu of your BIOS? I have read that recommended settings for overclocks are Auto/Normal/Auto/Disabled/Disabled for improved overclocked system stability with your board. Along with a couple of other tweaks.

well, i've been playing around in my BIOS, but i don't know how loose is loose. i've ran my timings at 2.5-3-3-7 and was able to get my FSB not too much higher than the 8Mhz i said earlier. But i don't remember if i increased my voltage or not. I've gotta test this out again.

Iridium · Jan 13, 2005, 02:10 PM

I would say 3/4/4/8 would be considered "loose". Try running like that and bump that voltage up and run at the 1:1 ratio and run a couple of benchies like that and see what happens. I am not sure what voltage you are at now, but I would say those sticks can handle 2.7~2.8 easy. I think 2.8 is all the higher our IC7-G max II's go. As for the settings I told you about for the GAT Auto/Normal/Auto/Disabled/Disabled try those, they seem to help quite a few people other than me achieve a higher stable FSB speed. Also, make sure that you have your AGP/PCI set at "Fixed" and it should be set at 66/33, if not, set it at 66. Try those out and see what happens..

MiDKnYtE · Jan 13, 2005, 02:43 PM

Here's a good read on timings versus bandwidth:
http://www.pcstats.com/articleview.cfm?articleID=873

I find that I get better performance when I OC at 1:1. That's not such a big deal with Athlons, but P4s really like to be synced up.

Vampyromaniac · Jan 13, 2005, 03:10 PM

I'll be interested to know how it turns out either way. Let us know, bud

CDsDontBurn · Jan 14, 2005, 12:42 AM

Quote:

Originally Posted by MiDKnYtE

Here's a good read on timings versus bandwidth:
http://www.pcstats.com/articleview.cfm?articleID=873

I find that I get better performance when I OC at 1:1. That's not such a big deal with Athlons, but P4s really like to be synced up.

of course it doesn't matter on Athlon 64s, the memory controller is on the CPU itself. Running a 1:1 ratio on p4 or even AthlonXP, it's always been best to run it syncronus vs. asyncronus (sp), that i know. Just wanting to know which is better, the lower timings and memory speeds @ 5:4 ratio, or higher timings and higher memory speeds @ 1:1 ratio.

MiDKnYtE · Jan 14, 2005, 02:27 AM

They tell you in the link I posted.

Quote:

On average, the system with the memory running at 400 MHz (5:4 memory divider enabled) with aggressive memory timings performed 2-3% faster than the system using high speed memory with loose timings.

They use very loose timings at 1:1 versus very tight settings at 5:4. You'll have to just bench to see which of your settings work best for you.

CDsDontBurn · Jan 14, 2005, 04:29 AM

Quote:

Originally Posted by MiDKnYtE

They tell you in the link I posted.

They use very loose timings at 1:1 versus very tight settings at 5:4. You'll have to just bench to see which of your settings work best for you.

i want my friend to let me borrow his DDR500 for a couple days to see how much better/worse it will benchmark and play games w/it compared to my DDR400. If it's a noticable difference, i'll go ahead and upgrade, but if it's a very minimal difference, i'll probably just upgrade to 2Gb of ram instead

.

Iridium · Jan 14, 2005, 09:05 AM

With your processor I doubt you'll ever hit the "high" FSB speeds anyways. The 3.2e that you have at 3.9ghz is a very good overclock, I don't think that you will be getting much higher than the 244FSB you are running at now with air cooling.

Have you tried the options I have mentioned about system stability settings in your bios with your current ram?

CDsDontBurn · Jan 14, 2005, 11:40 AM

Quote:

Originally Posted by Iridium

With your processor I doubt you'll ever hit the "high" FSB speeds anyways. The 3.2e that you have at 3.9ghz is a very good overclock, I don't think that you will be getting much higher than the 244FSB you are running at now with air cooling.

Have you tried the options I have mentioned about system stability settings in your bios with your current ram?

it's only a 700Mhz overclock. I know it's pretty good, but i hear of people hitting 800Mhz no problems and running stable. My chip runs fine @ 3.9Ghz, but once i take it to 4Ghz (800Mhz OC/250FSB), it starts to run unstable. I have the sp-94, which to me is THE best air cooling solution ever made for the p4 478.

I haven't yet tried to test out the settings you told me because i've been busy these past few days w/work and family and what not. I'll be able to try out your suggestions in the next couple of days (probably on sunday) cuz i'll have some free time to mess around w/my system.

mike2h · Jan 14, 2005, 11:57 AM

how high can you get with slightly looser timings & 1-1 ratio whith the mem you have? if you already posted, sorry.

CDsDontBurn · Jan 15, 2005, 02:14 AM

Quote:

Originally Posted by mike2h

how high can you get with slightly looser timings & 1-1 ratio whith the mem you have? if you already posted, sorry.

well, i was running my RAM at 2.5-3-3-8 1T, 2.8v, 1:1 ratio, and 210 FSB just now and playing HL2 for about 20min.

I got a memory error, and windows had to terminate the program i was running (which was HL2). So i would say is that it wasn't running stable. For me a stable OC is running my hardware OCed, and run smooth all the way smooth until i decide to stop playing, not when my computer decides to stop for me.

CDsDontBurn · Jan 15, 2005, 02:45 AM

well, it happened again. this time, i took a screen shot of the memory error just so you guys know what i'm talking about.

mike2h · Jan 15, 2005, 12:03 PM

try 3/3/3/8. that .5 makes little if any diff but it might help stability. after that, try 3/4/3/8 & so on.
i agree with you. do not get having an oc that causes you problems.
been running my 2.4c at 3.035(253 fsb) for a year now, 1:1 w 3/4/4/5 timings.
fast & stable.