• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Optimum E6600 OC'ing Question(s)

M

Markdek

[H]ard|Gawd
Joined
Jan 5, 2007
Messages
1,471
I'll have to admit it, I'm a modest overclocker. I just build a new system with the E6600 and a 965 motherboard (DS3). My RAM is 4-4-4-12 pc2 6400 (800) My modest goal is/was:

3.2 with the E6600 and to maximize the bandwidth of the 800mhz RAM.

I'm now stable at 3.2 @ 400 fsb x 8 (hurray), with, Modest increases in vdimm, vcore, mch. This give me a 1:1 with the RAM at 4-4-4-12.

Now, here's the question, based on my readings of the "strap" latency issues I've read about for the 965 chipset.

Is this the best settup for my goals for the E6600 (3.2) and RAM (1:1 @ 800 rating). Or, should I have used the "9" multiplier at some other combination of settings to get to the SAME Place, thus avoiding some "imposed" latency by the chipset.

The memory performance is good (read, write and copy), but the RAM latency per Everest is 74 ns; this is good, but is it OPTIMUM and is there some "strap crap" going on here???? Seems I'd also want to avoid the "imposed" 965 chipset overclocking (9/8 x FSB = MCH speed).


I.E., should I have gotten to 3.2 via 356 x 9, putting the RAM at 714, and possibly more overclocking, bringing the RAM up to 800 (but 3.6 is too high for my cooling/9700 Zalman), AND/OR, just use a divider other than 1:1.

Thanks,

Mark
 
The FSB strap is at 401, anyways I've found reducing the multiplier then maxing out the FSB works quite well as atm I'm running 7x450 w/ a 1:1 ram ratio and couldn't be happier :)
 
Thanks.

So is the 401 strap for all 965 boards, or just my DS3 [sorry, just noticed you've got the S3]. If this is the case, and I don't plan on extreme overclocking, then I'm set.

Else, I'd want to overclock upwards of 440-470 to see some speed improvements.

I'm currently checking out jumping up to 3.6 but I think it's going to be too hot.

Mark
 
dekruyter said:
Thanks.

So is the 401 strap for all 965 boards, or just my DS3 [sorry, just noticed you've got the S3]. If this is the case, and I don't plan on extreme overclocking, then I'm set.

Else, I'd want to overclock upwards of 440-470 to see some speed improvements.

I'm currently checking out jumping up to 3.6 but I think it's going to be too hot.

Mark
Click to expand...

to my knowledge yes
 
tisb0b said:
to my knowledge yes
Click to expand...

NO, some boards hit this strap @ 400, 402, 410, i've even seen 411. If I was you, keep using the 9x multi (dont drop down) and raise the FSB as much as possible. I bet if you run a 1MB PI run using the 9x multi and let say 356FSB (3.204GHz) VS 8 x 400 3.20GHz that the 9x multi will be faster. Try it! PLUS @ 400FSB your OCing your NB at its Very limit, if in fact your strap changes @ 401 I woould look into seeing how the performance is @ 425FSB and up.
 
Thanks,

I think you're right. I've been looking at Sandra memory performance (latency, etc.) and the chipset speed and it looks like the mch/965/northbridge is overclocked beyond my set FSB; ie., I set it at 400, and the NB is at 450. Latency numbers are high and stability is not 100% there.

I'm going to go down/up to 9x, lower FSB, and check memory performance and stability.

thxs.
 
Sorry, one more question/comment.

If i did move up to 425, and UP, then I would definitely have to use the lower mulipliers; maybe even have to drop to 7x,

because for now, I want to stay at 3.2.

Now, you see the situation.

AND, as you suggested (NEWLS1), I did try moving down to 8x and 356 and all specs dropped (memory and cpu).

I also move up to 425 and specs were still below 400 specs. So, I'd have to move up past 425, probably like 455, to get same specs as 8x at 400.

So, so far, 8x at 400 is actually seeming good, despite pushing the motherboard. Atleast I'm not pushing the E6600.

What do you think now; now that I've tested these? I'm really interested because I have a little knowledge, but not much practical experience.

Definitely thanks for your input; I need it.
 
newls1 said:
NO, some boards hit this strap @ 400, 402, 410, i've even seen 411. If I was you, keep using the 9x multi (dont drop down) and raise the FSB as much as possible. I bet if you run a 1MB PI run using the 9x multi and let say 356FSB (3.204GHz) VS 8 x 400 3.20GHz that the 9x multi will be faster. Try it! PLUS @ 400FSB your OCing your NB at its Very limit, if in fact your strap changes @ 401 I woould look into seeing how the performance is @ 425FSB and up.
Click to expand...

well my superpi run is faster at 3.15ghz(450FSB) over 3.2ghz(400fsb) which can be attributed to the ram running 100mhz faster.
 
Well, that's clear now; LOL


If I keep the 9X it's: 356 x 9 = 3.2 - Newls1's vote, keep the high Multi.

If I keep the 8X it's: 400 x 8 = 3.2 - Tisbob's vote, I think (higher FSB/ram).

Keep in mind, I don't have ultra fast ram to play with (800 / 4-4-4-12, barely). It can't be pushed much.

Anyone want to break the tie.

Since both end up at 3.2, and the 8x option results in higher FSB and appears to indicate better Ram throughput in Sandra, then is the 8x option it?

I think it would be nice/better if I had the ram (1000+) and/or the E6400/E6300 that seems to have more options to play with.
 
Stay at 9x356 Its better than 400x8. I have already benchmarked both scenerios and 400x8 just doesnt cut it the way 360x9 does.
 
well why not do a couple of superPI runs as the 8x multi does allow you to keep your ram at DDR800 whereas with the 9x multi your ram will only be running at 712mhz.
 
Ok, we're close here.

So Infospace, did you keep the ram at 712 / 1:1 and still see an improvement at 360 x 9.

See that's what I would have to do because any other divider would put it too far past 800.
 
Ok guys/girls.

SuperPi results:

8 x 400 = 3.2

Everest latency 74.2 ns
Everest read 7042 mb/s
SuperPi 7m 34s

9 x 360 - 3.24

Everest latency 81.5 ns
Everest read 6493 mb/s
SuperPi 7m 41s

Not a crushing defeat, but the 8 x 400 = 3.2 has it. Sandra also shows better memory numbers at 400 fsb (irrespective of multiplier)

Now the next question, if I move up to 425 at 8x, and beyond, will I see the sudden drop in ram throughput caused by the northbridge strap change that everyone is talking about.

More tests are in order. Future tests would involve moving toward 3.4, which I'm not sure I would stay at anyway. I think this would involve alot more boots (voltage checks), and I need a break!!
 
dekruyter said:
Ok guys/girls.

SuperPi results:

8 x 400 = 3.2

Everest latency 74.2 ns
Everest read 7042 mb/s
SuperPi 7m 34s

9 x 360 - 3.24

Everest latency 81.5 ns
Everest read 6493 mb/s
SuperPi 7m 41s

Not a crushing defeat, but the 8 x 400 = 3.2 has it. Sandra also shows better memory numbers at 400 fsb (irrespective of multiplier)

Now the next question, if I move up to 425 at 8x, and beyond, will I see the sudden drop in ram throughput caused by the northbridge strap change that everyone is talking about.

More tests are in order. Future tests would involve moving toward 3.4, which I'm not sure I would stay at anyway. I think this would involve alot more boots (voltage checks), and I need a break!!
Click to expand...

btw what is your time for a 2m SuperPI run at 400FSB? because I get 18.250 and I'm curious to see what a similarly clocked conroe gets :)
 
tisb0b said:
btw what is your time for a 2m SuperPI run at 400FSB? because I get 18.250 and I'm curious to see what a similarly clocked conroe gets :)
Click to expand...


I ran a bigger M to show the difference more clearly.

I got distracted and wandered up to 3.42 (428 x 8) and memory overclocked. I had it stable at a Vcore of just over 1.375 - not bad.

But, low and behold it wouldn't cold boot.

So I'm back to my standard specs for now. I suspect that Vdrop and/or mch voltage could have caused the black screen / no boot. I'll work on this (maybe) later. I'm not real anxious as Windows did get temporarily corrupted and almost didn't boot after return to old/stable settings. But obviously, I'm ok now. I hear Rev. 2 of the DS3 solved stability issues at higher overclocks - this 3.2 may be THE stable limit for me.
 
Hmm cant understand the rigidity in keeping the cas 4 timings as it has been shown on P965 boards memory timings have a very small impact on performance. Do me a favor ;)

set your ram to 5 5 5 12 or 5 5 5 15, keep memory multiplier at 2:1 (1:1 actually) cpu multiplier at highest (default) and run the fsb up until coretemp program shows 65C running Orthos. Then bench it.
 
BillParrish said:
Hmm cant understand the rigidity in keeping the cas 4 timings as it has been shown on P965 boards memory timings have a very small impact on performance. Do me a favor ;)

set your ram to 5 5 5 12 or 5 5 5 15, keep memory multiplier at 2:1 (1:1 actually) cpu multiplier at highest (default) and run the fsb up until coretemp program shows 65C running Orthos. Then bench it.
Click to expand...

Actually, I didn't keep the lower timings; I used 5-5-5-15 with the ram overclocked from 800 (norm) to 856. So it probably wasn't the ram as this should have been ok. I'm still learning the DS3, but it's not as easy with the e6600, as some say, to move past 3.2 and on up - so it seems.

And Bill, I'd love to keep moving the FSB overclocking up, but it doesn't doo much good good to do so much testing, if the machine won't boot from a cold start. That seems to be the biggest problem - some assurance of a cold boot after working so hard. Now if its the voltages, I've been conservative here - many use the "normal setting" where the bios picks the volts and moves it right up - 1.5v mch, etc. I think I'll try this next.

M
 
Auto voltage seems to be the way to go.

I'm now at 3.44, and Orthos temps are generally less than 60c in TAT.

I wish there was a program that could tell me the actual volages, like mch, on the ds3 so then I could tweek down these auto voltages to get my lowest temps.
 
dekruyter said:
Auto voltage seems to be the way to go.

I'm now at 3.44, and Orthos temps are generally less than 60c in TAT.

I wish there was a program that could tell me the actual volages, like mch, on the ds3 so then I could tweek down these auto voltages to get my lowest temps.
Click to expand...

upping the MCH will actually improve stability as I have mine on +0.2v and its stable yet on +0.1v it isn't.
 
Boy, I kept wavering back and forth on the Vmch to keep the following stable:

400 x 8 =3.2, with Vcore at 1.36875

After many tests, I'm keeping the Vmch at +0.2; it did seem like Orthos pooped out when I used my standard +0.1 Vmch. I'll be watching the mch temp now. With very intensive gaming and video, the motherboard temp (I'm assuming the mch) gets up to 55c; this seems ok as others say the northbridge allways runs hot.

Mark
 
Well, I'm a happy camper. Something let go in this bios while playing around, and I'm very stable and fast at:

400 x 8 = 3.2; STOCK voltage = 1.325, mch = +0.2, and the temps are increadibly great. I'm happy with this and going to use it for some time now while I enjoy my new build - first one.

Silverstone tj09 (black w/ window)
Thermaltake 850 watt
Zalman 9700 AT
DS3 rev 1
G.Skill 2gb; 800 4-4-4-12; 1:1
Sound Blaster Fatal1ty Extreme gamer
8800 gts !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
etc., etc., etc., oh, and a nice 22" Samsung

AND, when I get back in the mood; all I'll need to do is up the multiplier to 9x, adjust voltage and pray for 3.6 with air/Z. 9700 AT

Sandra and Everest scores looking good. And sorry Gigabyte, you didn't make this that easy! Shame on you.
 
dekruyter said:
Well, I'm a happy camper. Something let go in this bios while playing around, and I'm very stable and fast at:

400 x 8 = 3.2; STOCK voltage = 1.325, mch = +0.2, and the temps are increadibly great. I'm happy with this and going to use it for some time now while I enjoy my new build - first one.

Silverstone tj09 (black w/ window)
Thermaltake 850 watt
Zalman 9700 AT
DS3 rev 1
G.Skill 2gb; 800 4-4-4-12; 1:1
Sound Blaster Fatal1ty Extreme gamer
8800 gts !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
etc., etc., etc., oh, and a nice 22" Samsung

AND, when I get back in the mood; all I'll need to do is up the multiplier to 9x, adjust voltage and pray for 3.6 with air/Z. 9700 AT

Sandra and Everest scores looking good. And sorry Gigabyte, you didn't make this that easy! Shame on you.
Click to expand...

awesome, glad your happy with what you have....
 
You must log in or register to reply here.
Back
Top