Corsair 875P Memory Benchmark Report

[eva2000]

Interesting http://www.dvhardware.net/article.php?sid=1643

[ThugsRook]

you can see it coming now from far off in the distance....
you can hear it brewing as it gets closer and closer....
you can already smell it as im typing this up......



BS! :na:

[eva2000]

Originally posted by ThugsRook
you can see it coming now from far off in the distance....
you can hear it brewing as it gets closer and closer....
you can already smell it as im typing this up......



BS! :na: ROTFLMAO

[ThugsRook]

ok lets see here....
laming out your mem is the best way right?
well here's what happens when you "lame out" a system @ 1:1...


* P4 2.53b @ 3.44ghz 181fsb 362ddr
* P4P800 (springdale)
* bios v1008.003
* Kingston HyperX 3000 2x512mb
* GF4600


> 3DMark2001 <
16218 = Turbo / 2.0-2-2-5
15933 = Turbo / 2.5-3-3-7
15663 = Turbo / 2.5-4-4-8
15593 = Auto / 2.5-4-4-8
15384 = Standard / 2.5-4-4-8

TOTAL = -834 points



so now youre gonna tell me you can "make up" that 834 3dmarks by OCing your cpu further? :wack:

why bother?

[Hawk]

I saw that article ealry yesterday and decided not to post a link. . .


And also, for the record, I AGREE. . . . BS. . . . hahahahaha! ! ! !

[Mikki]

I'm also in agreement....:rolleyes: :wave: :p

[pointreyes]

Originally posted by Mikki
I'm also in agreement....:rolleyes: :wave: :p

But, but, it's from Corsair-why, it's got to be true. Why they probably even used that NaDa software for the tests. :p

[pastorjay]

:wave: <----PJ raises hand to join the consensous of the group...

[oldfart]

This benchmarking exercise allowed us to draw three significant conclusions. First, it confirmed for us that in current 875P (Canterwood) based systems, memory clock speed significantly outweighs memory latency settings as the primary performance factor in the memory subsystem. More simply put, you are better off clocking the memory as fast as it will go (assuming the CPU/FSB will go that fast) at the expense of looser latency settings, as opposed to cranking the latency settings as tight as they will go (as is the case with Nforce2, for instance).
Um....NO

Another Corsair PC3700 reveiw (http://www.3dxtreme.org/Corsair_xms3700_twinx_p1.shtml)

[roystarman]

Originally posted by oldfart
Um....NO

Another Corsair PC3700 reveiw (http://www.3dxtreme.org/Corsair_xms3700_twinx_p1.shtml)

I have the 3200LL Corsair memory and able to mildly oc my 2.8P4 to 3.015 (215 MHz FSB). I ahd to loosen thememroy latency to 2.5,7,3,3 to get this to work. However it does turn in about 5100 on the Sandra Bandwidth test Buffered. It also runs the MEMTEST86 and Prime 95 for at least 8 hours (one night each).

I've been reading a lot about all the newer memory and the latency vs FSB arguments. From what I can read in the articles about the 3700 and 4000 memory, unless you plan on really exceeding the 250FSB range then the looser memory latencies are going to cost you some performance at the lower FSBs.

With my 3200LL at stock it turns in about 4800 to 4900 on Sandra and I'm still debating if the extra heat in the CPU is worth the 200 point gain. I'm using a Spark 7 for now and it keeps the temp under 60 as measured by the IC7-G. It does seem to make Homeworld run a little smoother at the OC vs stock. So for now I'll keep it that way.

So now I'm really going out on a ledge here to make this statement. For those who will be Overclocking to the maximimum extent possible then buying the 3700 or 4000 is a worthwhile investment. For those like me who are more conservative and actually want thier CPUs and MOBO's to last a few years then, a lower speed memory with tighter latencies maybe in order. I guess I may not be real bleedinedge material but there is so much great stuff in here.

I'm not talking about Asynchronous operation. It appears that even at 200 Mhz FSB there is enough memroy bandwidth to allow a CPU to really fly in a lot of real world applications. Is this true or am I missing something?

So what we really need is some unlocked P4s for those real madmen!.

[ThugsRook]

thats a very good question roystarman ~ and you are correct that a ES chip is needed to truely answer this question.
however i think we could easily narrow it down w/o one....

1st off, p4 temps skyrocket after 3.4ghz so lets just use that number as an example.

in my case im a little too slow ~ only running 382ddr but i am using all the GA options (street racer c2622 nothing disabled) and only need 2.5v to run there.

async is not the way to go ~ it all runs best @ 1:1 with all the goodies turned on.

the 800fsb chip guys are running too fast ~ 500+ddr in as lame a mode as they need to actually still run and using 2.8+v to do it.

where would the best speed lie?
id say its wherever your ram can max out with ALL the goodies still on.
in my case that would be in the 410-424ddr area.

so my best 3.4ghz chip would be a 16x212 or a 17x200 chip.
of course these chips dont exactly exist ~ but they would yield the best possible performance out of a 3.4ghz system.

HTH :-)

[oldfart]

roystarman, dont use SiSoft mem to measure your system performance. That bench has nothing at all to do with real world performance. Bench something you use. Games, encoding, archiving..whatever.


I dont see the 3700 - 4000 mem as a good investment for anyone. Yeah, you can run a higher DDR speed, but the slow timings you have to use negate any benefit you get from doing so. You spend a lot of $$ for little if any gain. Many people have seen a decrease in performance running it. Read that article I linked to.