Big difference between 250mhz at 1:1 and 250mhz at 5:4?

[oldfart]

3.3 GHz with a stock HS/fan will work, but you will be throttling at load. Not a good idea. There is a reason that the there is one stock HS/F up to 2.8 GHz and a larger copper core for 3.0 - 3.2 GHz chips. A larger one is needed. I run a Swiftech MCX4000 and a Panaflo M1A. Even with it, temps can get into the mid 60C range. With a stock HS/F you will easily go over 70C an will throttle.

As far as mem goes, the OCZ can not run tight timings. It is also difficult to find memory that will run 2-2-2-6 these days. Most mfgrs have gone to the high DDR speed high latency design. The best for tight timings right now is the Mushkin level II PC3500 stuff, and it is not cheap.

 

[FPSguy]

I agree that 3.3GHz is pushing it with stock cooling, though I think it is do-able (and I intend to try it as soon as my Antec TrueBlue 480 arrives ...). At 3GHz, though, my P4 2.4C seems very comfortable, and my load temps are in the 60's.

 

[hurrikaane]

"As far as mem goes, the OCZ can not run tight timings. It is also difficult to find memory that will run 2-2-2-6 these days"

Yeah I have seen that also. From the tests it seems Corsair's 3200 pro/LL runs at the lowest timings (seen 2-2-2-5 at 250mhz).

 

[FPSguy]

Originally posted by: hurrikaane
The OCZ I was referring to was the 500mhz version, which can't really be run at 5:4 since that'd be beyond the capability of the board (312mhz). I like the idea of running the 400/433 mhz stuff at 5:4, since that gives you 250/270 (thus 3.3-3.5ghz). That's pretty ideal, given than the Corsair has been shown to be able to run in real-world tests at 2-2-2-5.

Just in case it's not clear, you can run the OCZ below its top speed. Thus, you can run it at 5:4, but you probably will not be able to run it at its top speed at 5:4. You would just be throttling it back a bit and probably tightening the timings. Whether that costs more than lower-speed memory with tight timings is another question.

 

[hurrikaane]

Yeah I knew that, but it'd make no sense to spend 400 bucks on the OCZ 4000 and then throttle it back to 3200/3500 speed 😉

 

[hurrikaane]

I checked out that Mushkin Level 2 3500 stuff, $315 for 1 GB of it at newegg. The Corsair 3200LL stuff is $285. So only a $30 difference for memory that'll run at 2-2-2 and 33mhz higher. Not bad really.

 

[orion7144]

Originally posted by: hurrikaane
"As far as mem goes, the OCZ can not run tight timings. It is also difficult to find memory that will run 2-2-2-6 these days"

Yeah I have seen that also. From the tests it seems Corsair's 3200 pro/LL runs at the lowest timings (seen 2-2-2-5 at 250mhz).

My OCZ PC3500 EL runs at its rated speed and CAS rating of 2-2-2-6 so yes it is good memory. It is the PC4000 Gold that has the slower timings. However with the GOLD at 275fsb 1:1 and the PC3500 @ 275fsb 5:4 the memory scores are higher for the Gold so as I have stated before if you can get the fsb up and run memory at 1:1 then that will make up for the timings.

Also I have 2 of my 2.4C at 3.4gig with the stock HSF and the load temps never get above 61C and that is what is reported on the Abit boards (we all know they read high).

 

[Slammy1]

Well, I still argue the best bang for the buck is a 2.4C and PC3500 memory, but the best performer is a 3.2C with PC3500. I'm running HyperX 3500, the BH-5 version, and it's 280FSB 5:4 3.363GHz 2-3-3-7 24/7 (well, when it's on which is a lot). I do like the OCZ, but with binning of higher speed chips I don't see good mem o/c as people used to get. The problem with 2.6's is they are more variable in their ability to o/c than the 2.4s. Key is thermal control, The problem with the stock HSF is the thermal tape. Well worth an extra $35 for a better solution. I max 50-51C for my set up, case cooling helps especially off the Northbridge. I've heard the newer IC7 can run over 300 FSB.

 

[oldfart]

Originally posted by: orion7144

Originally posted by: hurrikaane
"As far as mem goes, the OCZ can not run tight timings. It is also difficult to find memory that will run 2-2-2-6 these days"

Yeah I have seen that also. From the tests it seems Corsair's 3200 pro/LL runs at the lowest timings (seen 2-2-2-5 at 250mhz).

My OCZ PC3500 EL runs at its rated speed and CAS rating of 2-2-2-6 so yes it is good memory. It is the PC4000 Gold that has the slower timings. However with the GOLD at 275fsb 1:1 and the PC3500 @ 275fsb 5:4 the memory scores are higher for the Gold so as I have stated before if you can get the fsb up and run memory at 1:1 then that will make up for the timings.

Also I have 2 of my 2.4C at 3.4gig with the stock HSF and the load temps never get above 61C and that is what is reported on the Abit boards (we all know they read high).

Synthetic mem scores will be higher. Real apps will not. Real world apps will run faster with tight timings. This is why people who sell the high DDR speed high latency memory put so much focus on programs like SiSoft instead of real applications. There has been plenty of testing done that proves that 5:4 with tight timings outperforms 1:1 with slow timings.

 

[orion7144]

Originally posted by: oldfart

Originally posted by: orion7144

Originally posted by: hurrikaane
"As far as mem goes, the OCZ can not run tight timings. It is also difficult to find memory that will run 2-2-2-6 these days"

Yeah I have seen that also. From the tests it seems Corsair's 3200 pro/LL runs at the lowest timings (seen 2-2-2-5 at 250mhz).

My OCZ PC3500 EL runs at its rated speed and CAS rating of 2-2-2-6 so yes it is good memory. It is the PC4000 Gold that has the slower timings. However with the GOLD at 275fsb 1:1 and the PC3500 @ 275fsb 5:4 the memory scores are higher for the Gold so as I have stated before if you can get the fsb up and run memory at 1:1 then that will make up for the timings.

Also I have 2 of my 2.4C at 3.4gig with the stock HSF and the load temps never get above 61C and that is what is reported on the Abit boards (we all know they read high).

Synthetic mem scores will be higher. Real apps will not. Real world apps will run faster with tight timings. This is why people who sell the high DDR speed high latency memory put so much focus on programs like SiSoft instead of real applications. There has been plenty of testing done that proves that 5:4 with tight timings outperforms 1:1 with slow timings.

But, from what I have seen, most of those real life benchmarks that you see on the web do not take FSB into account. Example PC3500 at a lower overall FSB but with the better timmings then they run the same FSB with the slower timings. My testing using SmartPar, winrar, and various programs to compres video have all shown the same thing that I stated in my last post. We are talking <1-2%performace swing on any of those apps at the same FSB hoever with the higher FSB 1:1 that increases to 3-9% depending on how much difference there is in the FSB.

 

[oldfart]

Yes they do.

I can link you to a bunch if tests real uses have done. The tests look something like this:

250 FSB
1:1 ratio
DDR400
3-3-3-8 timings

Vs

250 FSB
5:4 ratio
DDR500
2-2-2-6 timings

The second system is faster. Same FSB. Same CPU speed. Same mobo. The faster timings outperform the higher DDR speed.

 

[Slammy1]

Originally posted by: oldfart
250 FSB
1:1 ratio
DDR400
3-3-3-8 timings

Wow, that's some killer PC3200 😉.

But, but, the unbuffered benchmarks are so much better... I agree. I think it makes some difference, but it's not so much that better timings or a higher o/c cannot compensate. Maybe it's just a Northbridge quality issue, or BIOS problems, but there are cases where it will give some difference. Just not worth the added costs.

 

[oldfart]

Well, obviously, the DDR500 setup is not PC3200 ram.

The main point is running 1:1 with slow timings is not a great way to go. 5:4 with fast timings will get you the same and often better performance.

If people would stop wasting their time running useless synthetic SiSoft benches that in no way have anything to do with real world performance and ran tests on actual applications, this would be very apparent.

 

[Slammy1]

^Agreed. I don't think the synthetic benches are meaningless, just almost meaningless. Definitely not meaningful in the sense of what others try to prove when they post them.

 

[SuPrEIVIE]

Great info! is Buffalo PC3700 ram any good to do 250MHZ at 5:4 w/tight timings its 200$for 2 sticks!(i heard they are BH-5 chips...)

 

[Shimmishim]

i'm running 12 x 292 so i had to get somethign better than pc3200 even to run with a 5:4

so i'm running geil gold dragon pc3700 @ 2.5-4-3-7

i could run the 3:2 ratio but i dunno if that would be worth it...



kind of makes me wish now that i had gotten a 2.8 or even 3.0C

 

[FPSguy]

Originally posted by: Shimmishim
i'm running 12 x 292 so i had to get somethign better than pc3200 even to run with a 5:4 so i'm running geil gold dragon pc3700 @ 2.5-4-3-7 i could run the 3:2 ratio but i dunno if that would be worth it... kind of makes me wish now that i had gotten a 2.8 or even 3.0C

IMO what you need most is a new video card. That Ti4200 isn't doing justice to your 3.5GHz P4. 🙂