P4 Platform, What matters more...bandwidth or timings?

I'm trying to find good ram for my P4 system, P4 2.4C SL6WF week 10, Abit IC7-G (newer version), and a Fortron 530 PSU.

Basically, i'm going to overclock it all and i'm hearing mixed issues about the P4 platform, some people say bandwidth is all that matters on a P4 system, not timings. Yet i hear others say timings are more of an issue, well i'm sure if it were possible to have the best of both worlds at once, that would be the answer, but i need someone to clear this up for me.

I personally need 1GB (2x 512's) and i dont really want to shell out 3 bills for Mushkin Level 2 Black...I'm also hearing that the KHX PC3000 is now CH-5 (what i was going to get). Man i just wish i could have gotten some of those sweet Buffalo BH-5 PC3700's when they were around ( *glares at Larva in a mischeivious way* ), and now i gotta "hunt" for good ram lol.

So would someone gimme some ideas here reguarding the P4 platform and overclocking its ram?

BTW: if i were to actually get the courage to buy the Mushkin, which would be better? PC3200 or PC3500?? the PC3500's are cheaper right now on newegg, whats all that about?

checkout this thread it might help u http://www.ocforums.com/showthread.php?s=&threadid=259352

flapperhead said:

checkout this thread it might help u http://www.ocforums.com/showthread.php?s=&threadid=259352

Click to expand...

I read that post, its clearly boasting bandwidth over timings, more specifically more bandwidth first tigher timings second on a P4 system. If this is absolutely true and then RAM such as pc4000 is in fact better for most systems, can you suggest some good pc4000 that will get close to 300FSB? thats my target area, 290-300FSB.

I noticed you were talking about the Kosuma DDR being Adata, is this true? here's the linky i found.
http://www.komusa.com/komusa/51ddrpc50hy.html

Also, is KHX PC4000 any good? rated at 3-4-4-8 does anyone think this stuff is worth it? Cause i can get a whole GIG of this KHX PC4000 for $200 at work lol, if i can spend $200 on 1GB of P4000 and it be nice ram for overclocking i'll do it in a heartbeat but thats prolly the only name brand ram we carry also. Newegg linky here:
http://www.newegg.com/app/ViewProductDesc.asp?description=20-144-134&depa=1

http://www.pcstats.com/articleview.cfm?articleID=873

Perhaps that will answer your question

i had both adata pc4000 and mushkin pc3500 lvl2 black

all of my sandra scores would be higher on the tighter timing mushkin, but 3dmark scores remained the same, with little variation on either side

right now im at 267fsb with the 5:4 ratio which is like 214fsb which is underclocking it

i believe i can push the mushkin far, but not as far as pc4000 that is guaranteed to work at ddr500

is KHX PC4000 any good? rated at 3-4-4-8 does anyone think this stuff is worth it? Cause i can get a whole GIG of this KHX PC4000 for $200 at work lol, if i can spend $200 on 1GB of P4000 and it be nice ram for overclocking i'll do it in a heartbeat but thats prolly the only name brand ram we carry also.

Click to expand...

anyone used KHX PC4000 before?

if u want to go hi fsb 1-1 i would suggest something like ocz4200,something with adata chips,corsair pc4400 komusa/pc4200 with the adata upgrade...

What matters more...bandwidth or timings?

Click to expand...

Not all P4 set ups are condusive to a high FSB. There are a couple of factors involved. The multiplier of the cpu, the dividers available, and the potential of the cpu. Looking at the cpu database the one thing that sticks out is the average of the northwood cpu's. No matter what the rated speed, most seem to hit in th 3.4-3.6 range.

This is where the multipliers come in. Take the potential maximum oc (let's say 3.5 or 3500) and divide that by the fixed multipler rate of your given P4. A 2.4c has a 12x multiplier, 3500 divided by 12=291, 3500/15= 233(P4 3.0). This would be your potential fsb.

Now with the 3.0 cpu the memory choice gets easy. You just need ram that will do 233 with tight timings(a no brainer). The 2.4 with a potential 290+ bus is a problem. Currently nobody makes DDR600 so 1:1 is difficult at best. 1:1 with tight timings is not going to happen. Here come the memory dividers. A FSB of 290 using a 5:4 divider gives a 232 bus speed for the ram. Back to the 233 with tight timings is a no brainer. Now here's the big difference in performance. The 2.4 has a system bus of 290x4=1160(quad pumped), the 3.0 is only 233x4=932. Both running the same memory with the same timings. Now you've got the best of both worlds, a high fsb(1160) with tight timings. You currently can't get tight timings with memory above about pc3500. This why I chose pc3500 that would do 2-2-2-5 at spec and does 2-3-2-5 at 290+ fsb, 5:4.

Grandpa Dan said:

Not all P4 set ups are condusive to a high FSB. There are a couple of factors involved. The multiplier of the cpu, the dividers available, and the potential of the cpu. Looking at the cpu database the one thing that sticks out is the average of the northwood cpu's. No matter what the rated speed, most seem to hit in th 3.4-3.6 range.

This is where the multipliers come in. Take the potential maximum oc (let's say 3.5 or 3500) and divide that by the fixed multipler rate of your given P4. A 2.4c has a 12x multiplier, 3500 divided by 12=291, 3500/15= 233(P4 3.0). This would be your potential fsb.

Now with the 3.0 cpu the memory choice gets easy. You just need ram that will do 233 with tight timings(a no brainer). The 2.4 with a potential 290+ bus is a problem. Currently nobody makes DDR600 so 1:1 is difficult at best. 1:1 with tight timings is not going to happen. Here come the memory dividers. A FSB of 290 using a 5:4 divider gives a 232 bus speed for the ram. Back to the 233 with tight timings is a no brainer. Now here's the big difference in performance. The 2.4 has a system bus of 290x4=1160(quad pumped), the 3.0 is only 233x4=932. Both running the same memory with the same timings. Now you've got the best of both worlds, a high fsb(1160) with tight timings. You currently can't get tight timings with memory above about pc3500. This why I chose pc3500 that would do 2-2-2-5 at spec and does 2-3-2-5 at 290+ fsb, 5:4.

Click to expand...

GEEZ! It sure is a benefit to have a "math proffessor" that's an Overclocking Guru too! and to think, I was contemplating the Ai7 just for the "software version". I think I can speak for most of us around here when I say " Your screen name sure fits like a glove"

Grandpa Dan said:

Not all P4 set ups are condusive to a high FSB. There are a couple of factors involved. The multiplier of the cpu, the dividers available, and the potential of the cpu. Looking at the cpu database the one thing that sticks out is the average of the northwood cpu's. No matter what the rated speed, most seem to hit in th 3.4-3.6 range.

This is where the multipliers come in. Take the potential maximum oc (let's say 3.5 or 3500) and divide that by the fixed multipler rate of your given P4. A 2.4c has a 12x multiplier, 3500 divided by 12=291, 3500/15= 233(P4 3.0). This would be your potential fsb.

Now with the 3.0 cpu the memory choice gets easy. You just need ram that will do 233 with tight timings(a no brainer). The 2.4 with a potential 290+ bus is a problem. Currently nobody makes DDR600 so 1:1 is difficult at best. 1:1 with tight timings is not going to happen. Here come the memory dividers. A FSB of 290 using a 5:4 divider gives a 232 bus speed for the ram. Back to the 233 with tight timings is a no brainer. Now here's the big difference in performance. The 2.4 has a system bus of 290x4=1160(quad pumped), the 3.0 is only 233x4=932. Both running the same memory with the same timings. Now you've got the best of both worlds, a high fsb(1160) with tight timings. You currently can't get tight timings with memory above about pc3500. This why I chose pc3500 that would do 2-2-2-5 at spec and does 2-3-2-5 at 290+ fsb, 5:4.

Click to expand...

Sure, but what's better... 290 1:1 @ 3,4,4,8, or 290 5:4 @ 2,3,2,5? That's the real question here.

kestrel said:

Sure, but what's better... 290 1:1 @ 3,4,4,8, or 290 5:4 @ 2,3,2,5? That's the real question here.

Click to expand...

Thats what I want to know!

I chose pc3500 that would do 2-2-2-5 at spec and does 2-3-2-5 at 290+ fsb, 5:4.

Click to expand...

Why would I do this if I didn't believe it was faster?

I'm not running 1:1 @ 290, but here's a benchie running 5:4 at 290.


Find somebody doing 1:1 @ 290 and compare.

And your question actually should read: "FSB or Timings". the two combined is what gives bandwidth.

Well here is 292 1:1 3-4-4-8 timings



While this may look impressive it's not realy that good. I had PAT enabled while your board didn't. You would have also beaten me if we ran this unbufferer or used almost any other benchmark.

Using sandra to test speed is one thing but for perfomance you need to use other real benchmarks

Steve

mrspec3 said:

Well here is 292 1:1 3-4-4-8 timings

I had PAT enabled while your board didn't.
Using sandra to test speed is one thing but for perfomance you need to use other real benchmarks

Steve

Click to expand...

Yeah, I only get partial pat. What does ctiaw show on your sys?

Those are about right PAT vs. non PAT for % increase.

It shows full pat. I did the bios hack on my p4p800 that lets you run 1:1 with full pat

Sorry, I meant the "memory read bandwidth" numbers.

GD

Well I didn't run it at that speed to check but at 275 it gave me 6739.1MBytes/s

At those settings I'm high 6500's low 6600's. At 275 is that with the same timings?

This question has been asked many times. If you search the forum you'll see several posts. Most are probably a couple of months ago. The question is basically, which is faster. High fsb at 1/1 using loose timmings versus 5/4 using tight timings. If you run Sandra buffered the 1/1 loose timmings will win every time, but using other benchmarks such as 2001SE the tight timmings are a small amount faster. One of the review sites on the net actually tested this and found the same results. If you run a search far enough back you should be able to find it.

pkrew said:

This question has been asked many times. If you search the forum you'll see several posts. Most are probably a couple of months ago. The question is basically, which is faster. High fsb at 1/1 using loose timmings versus 5/4 using tight timings. If you run Sandra buffered the 1/1 loose timmings will win every time, but using other benchmarks such as 2001SE the tight timmings are a small amount faster. One of the review sites on the net actually tested this and found the same results. If you run a search far enough back you should be able to find it.

Click to expand...

Or look at the first post I made in this thread

And yes it was same timings at 275