Corsair 3500 XMS Or Corsair 3700 XMS for OC'ing

F1

Member
May 3, 2003
46
0
0
I am going to buy a p4 2.6 800 and abit ic7 mb very soon. I was planning on getting Corsair 3500 XMS DDR RAM but with the recent memory increases Corsair 3500 XMS is about the same price as Corsair 3700 XMS. Im planning to overclock my system to a nice stable level (nothing too extreme). What RAM should i go for seeing the above are basically the same price? Will one allow for better mid range overclocking?

Thanks.

 

zainali

Golden Member
Jun 18, 2003
1,687
0
76
from what i have heard/read those two and 3200LL are the same chips just tested at different timings. u might wanna try ur luck with 3200LL which most prolly will run at 3700 speeds with relaxed timings.
 

F1

Member
May 3, 2003
46
0
0
I remember hearing about many people have memory issues with the Abit Ic7 and 3200LL, so i thought id stay clear of the 3200LL.


Thanks
 

Fallengod

Diamond Member
Jul 2, 2001
5,908
19
81
Are you gonna get 1gb? If so you could try the stuff at bestbuy. Bestbuy has kingston 3500 512mb stick for $100 after a $20 rebate.
 

bgeh

Platinum Member
Nov 16, 2001
2,946
0
0
i'll get the 3700-the timigs may be a little low, but i think you should be able to tighten them without a lot of problems
 
Apr 17, 2003
37,622
0
76
Originally posted by: zainali
from what i have heard/read those two and 3200LL are the same chips just tested at different timings. u might wanna try ur luck with 3200LL which most prolly will run at 3700 speeds with relaxed timings.

you would need horrible loose timings to accomplish that, like 3-4-4-8
 

borgmang

Senior member
Jun 27, 2003
335
0
0
I just went through 2 weeks of trying to decide on memory. You cant go wrong at this point with Kingston HyperX 3500. Fairly decent price and they peform well. I would also highly consider the 2.8C if I were you.
 

F1

Member
May 3, 2003
46
0
0
Originally posted by: borgmang
I just went through 2 weeks of trying to decide on memory. You cant go wrong at this point with Kingston HyperX 3500. Fairly decent price and they peform well. I would also highly consider the 2.8C if I were you.

Is it really worth paying the extra $50US for the 2.8 over the 2.6 when planning on overclocking my system?

thanks.
 

borgmang

Senior member
Jun 27, 2003
335
0
0
Originally posted by: F1
Originally posted by: borgmang
I just went through 2 weeks of trying to decide on memory. You cant go wrong at this point with Kingston HyperX 3500. Fairly decent price and they peform well. I would also highly consider the 2.8C if I were you.

Is it really worth paying the extra $50US for the 2.8 over the 2.6 when planning on overclocking my system?

thanks.

It's up to you, but you can OC both the 2.6C and the 2.8C about the same or you might get more headroom with less stress on the system with the 2.8C. It really depends how good the CPU is that you get. They bin sort the CPU's and sometimes you can actually get lucky and get one that has more potential. I personally think that the 2.8C is the one to get. Just lthinking ogically it is the one in the middle of the 800C processors - 2.4, 2.6, (2.8), 3.0, 3.2. Up to you - Good luck.
 

tazdevl

Golden Member
Mar 1, 2000
1,651
0
0
Timings mean squat if you're running the RAM on a Canterwood based board with PAT enabled.
 

oldfart

Lifer
Dec 2, 1999
10,207
0
0
Originally posted by: tazdevl
Timings mean squat if you're running the RAM on a Canterwood based board with PAT enabled.
That is simply not true. I don't know who started that rumor, but it is false. How do I know? I tested it myself. So have others and they came up with the same. Timings DO matter. Running a high DDR speed with some of the pathetic timings I've seen is not as good as you might think.

Here is a post I put up not too long ago on my testing.

**********************************8

I ran some benches on my system for another forum and thought it might be interesting for some here. I didn't run every combo, but enough to show how memory settings can affect system performance.

3DMark was a PITA. I cant get consistent enough results to draw any conclusion from it. When I saw a bench that didn't look right, I ran it again. It seemed I could get a different bench each time. I didn't have that problem with any other bench. All the other tests were very consistent. I pretty much throw the 3DMark results out as invalid.

A 1/2% change in a 17000 score is 85 points. That is well with in the normal variance in a benchmark.

TEST SETUP
Abit IC7 | BIOS 14 | 2.6C @ 3.34 | 257 FSB | 2 x 512 XMS 3200C2 | 5:4 | DDR412 | GA settings = auto

Note1: ctiaw shows PAT as "Fully Enabled"
Note2: CPUz and ctiaw show cas3 BIOS setting as cas 2.5 actual

3Dmark2K1
2.0/5/2/2 = 17987
2.5/5/2/2 = 17983
3.0/5/2/2 = 17816/17899/17876 (1st run/2nd run/reboot - retest)
2.0/7/3/3 = 17858
2.5/7/3/3 = 17760
3.0/7/3/3 = 17756
2.0/8/4/4 = 17414/17545 (1st run / reboot retest)
2.5/8/4/4 = 17569/17602 (1st run / reboot retest)
3.0/8/4/4 = 17500


Q3A
2.0/5/2/2 = 402
2.5/5/2/2 = 400
3.0/5/2/2 = 400
2.0/7/3/3 = 392.8
2.5/7/3/3 = 391.2
3.0/7/3/3 = 391.8
2.0/8/4/4 = 385.9
2.5/8/4/4 = 382.5
3.0/8/4/4 = 383.3

UT2K3 Flyby/Bot
2.0/5/2/2 = 237.94/92.1
2.5/5/2/2 = 237.50/90.09
3.0/5/2/2 = 237.54/91.45
2.0/7/3/3 = 235.4/89.95
2.5/7/3/3 = 235.22/89.80
3.0/7/3/3 = 235.33/89.54
2.0/8/4/4 = 233.16/85.48
2.5/8/4/4 = 233/87.56
3.0/8/4/4 = 232.95/86.45

TMPGEnc (lower is better)
2.0/5/2/2 = 4:23
2.5/5/2/2 = 4:29
3.0/5/2/2 = 4:28
2.0/7/3/3 = 4:31
2.5/7/3/3 = 4:32
3.0/7/3/3 = 4:32
2.0/8/4/4 = 4:34
2.5/8/4/4 = 4:36
3.0/8/4/4 = 4:36
 

tazdevl

Golden Member
Mar 1, 2000
1,651
0
0
If you're a benchmarking freak and that's all you do, fine it makes a marginal difference. However from an end-user perspective it means squat.

http://www.techseekers.net/modules.php?name=Reviews&rop=showcontent&id=65&page=2

The biggest challenge at the moment is finding memory that will perform consistently well with Canterwood boards. What works well for one person, doesn't for another. You have the 3200LLs, I had them, ended up RMAing, wouldn't even run spec. Tried Buffalo tech, wouldn 't post after initial BIOS configuration. Waiting on TwinX 3700, we shall see. If this doesn't work, chances are it's the motherboard, but I don't think that's the case.

It seems that the days of being able to smack whatever you want into a Intel based motherboard are long gone.
 

oldfart

Lifer
Dec 2, 1999
10,207
0
0
Actually, I'm not a benchmarking freak. You will note none of those lame Sandra mem benches were done. These benches were done for a site that has those types. The same ones who claim that you MUST run 1:1 ratio even at lousy timings and even sacrifice CPU speed. People are buying $$ ram to run 1:1 @ slow timings and getting no benefit from it. My purpose of those benches was to show that you can run a 5:4 ratio with tight timings instead of 1:1 with slow timings and get ~ the same performance for a lot less $$.

BTW, they are not 3200LLs. They are the 3200C2 regular. Bought them used off the FS/T forum from two different people. They are not even a matched set!. i got pretty lucky with them.
 

tazdevl

Golden Member
Mar 1, 2000
1,651
0
0
Never said you were a benchmarking freak.

My point was that with current memory technology, you can buy low latency modules and hope they are able to hit higher FSBs or pay a bit extra for modules that will run @ higher FSBs with looser timings. As the bus speed gets higher, latency has a lesser impact on performance. The question is, what's the best strategy? I'd rather not be limited by my memory.

It's a good thing we'll be seeing DDRII in early 04.

You did get lucky on those modules.

 

oldfart

Lifer
Dec 2, 1999
10,207
0
0
The question is, what's the best strategy?
For me, get a low multiplier CPU (I prefer 2.6C over 2.4C), run a high FSB and a 5:4 ratio with tight timings. A high mult CPU and expensive ram @ 1:1 is not much faster, but costs a bunch more.
 

tazdevl

Golden Member
Mar 1, 2000
1,651
0
0
Originally posted by: oldfart
The question is, what's the best strategy?
For me, get a low multiplier CPU (I prefer 2.6C over 2.4C), run a high FSB and a 5:4 ratio with tight timings. A high mult CPU and expensive ram @ 1:1 is not much faster, but costs a bunch more.


Same line of thinking except about the RAM.

We'll see. If the 3700 doesn't work, I'll have to figure something else out.
 

tazdevl

Golden Member
Mar 1, 2000
1,651
0
0
Dinna know yet. Had problems with the last 2 types of RAM. Though I'm shooting for 3.5GHZ.

Also have a Microcool (7V modded) on the NB.

If this doesn't work, P4C800E is going back and I'm picking up a IC7 and some OCZ Gold.

I used to love Asus, but after their KT266A (went through 3) and now (I think) this one, I'm going back to Abit. My KR7A133R was the best board I've owned.
 

oldfart

Lifer
Dec 2, 1999
10,207
0
0
2.6C @ 3.5 GHz = 270 FSB. Thats DDR 540 @ 1:1. Not easy to do. You will have to run some VERY slow timings. Also, most ram that can do that is 256 Meg sticks. I dont think it is possible with 512 Meg sticks. 5:4 would be DDR 432 which is not hard at all even with reasonably good timings and 512 Meg sticks.

Asus is overrated IMHO. They have made some very good boards and quite a few duds as well although I think the P4C800 is pretty good. I get whatever is best at the time. My last mobos have been (from most recent): Abit, Albatron, Epox, Gigabyte, Asus, Asus, Abit.

Why dont you run your mem @ slowest possible speed and see where the CPU will go first, even if you have to run 1 stick of mem as a test? That is the way I always do it.
 

Allen7

Golden Member
Apr 10, 2000
1,589
0
0
Friend of mine bought 2x Corsair 3700 256MB sticks and his system could not boot at 466 rated speed. Switching to 2x OCZ 3700EL 256MB and his system runs like a champ, 2.8 -> 3.2 rock solid.
 

tazdevl

Golden Member
Mar 1, 2000
1,651
0
0
Thanks for the advice, I know what I'm doing. My point was that I have yet to even get up @ stock speeds due to either RAM or motherboard issues. I literally can't even get beyond the initial BIOS configuration and post. That has me leaning towards memory since all peripherals are seen in BIOS OK, voltages are in spec.

Already pulled everything unnecessary out and tried to boot, pulled the motherboard out of the case and tried to boot... so it isn't a short.

I might just try 4 DIMMS of 256MB OCZ Gold if the TwinX 3700 doesn't work. Canterwood does seem to do OK with all 4 DIMMs populated.

I wonder if it's a SIMM/DIMM issue with the Canterwood chipset. That would be a shame if a lot of folks are limited by the size of the RAM modules.

Allen, a 2.8 @ 3.2 is a pretty feeble OC unless he's running 1:1. He could probably go higher.
 

Nebor

Lifer
Jun 24, 2003
29,582
12
76
I'm running a 3.0 @ a 217mhz FSB. Keeping the 1:1 ratio... I don't remember the exact number, but it's about 3.3ghz. I don't remember what timings are.... I think 2.5C. I haven't tried taking it higher. Sandra says that I'm performing worse than an Intel 875 with pc3200CL2. Says it makes 5007mb\s... and I get 4915. But my 3dmark scores are 15.5k.... so... you be the judge.
 

Allen7

Golden Member
Apr 10, 2000
1,589
0
0
Allen, a 2.8 @ 3.2 is a pretty feeble OC unless he's running 1:1. He could probably go higher.

I think it's not bad in a Shuttle SB61G2 :D and he's running 1:1.

OCZ may cost you more but if hasn't failed me like the Corsair 3700EL (I used to be a Corsair believer). Check the support page of Corsair and you'll see many people are having problems with their 3700EL.

Allen