New DDR2 Ram Upgrade

H

haxxorboi

Limp Gawd
Joined
Aug 21, 2007
Messages
159
I'm currently running a Gigabyte P35C-DS3R with an e8400@4.0GHz with 2GB OCZ 1066MHz ram and just got an EVGA 260 and I'm starting to find bottlenecks in the system and overclocking. The OCZ ram is running at 2.24V (+.3 v, +.4 bumps it to 2.4V which is outside my comfort and warranty) and honestly it is really hot, so I'm looking to upgrade...

My plans are either go with low voltage DDR2 or some DDR3, going with 2x2GB kits. If the price isn't too horrible I may get two kits of DDR2, but I can only get 1 DDR3 kit.

What's a better buy, 4GB of DDR3-1333 C7 or 4GB of DDR2-1066/1100 C5? I'm overclocked right now with (445x9) but can probably push it to a bit higher (e8400 E0 step) but I don't really know if I want higher frequency or lower latency. Also, I have no idea how to compare DDR2 and 3 as I've heard they're operate differently and thus direct comparison is difficult.

The two kits I was eying were:
DDR2
DDR3

Suggestions/Comments gladly appreciated! Thanks!
 
Latency matters less to DDR3 than it does with DDR2. DDR3 will only give you a marginal boost in performance however it's important to note that they have a ton of low voltage kits (1.65v and below) that will decrease the thermals in your case and possibly increase your overclock. I have an i7 with a ton of goofy issues, (mostly because I'm new to the platform), so don't take my word for it.
Right now DDR2 is significantly cheaper than DDR3, but the gap is closing...(slow but steady).

Both kits look nice.
 
If the bus speed is only running at 445, then why are you wanting to upgrade the RAM? (edit: other than to have more RAM) If you run the RAM at a 1:1 ratio, you won't hit 1066 till the FSB hits 533.

And running the RAM higher than the FSB on a core2 system is going to get you very little in terms of RAM throughput.
 
cyclone3d said:
If the bus speed is only running at 445, then why are you wanting to upgrade the RAM? (edit: other than to have more RAM) If you run the RAM at a 1:1 ratio, you won't hit 1066 till the FSB hits 533.

And running the RAM higher than the FSB on a core2 system is going to get you very little in terms of RAM throughput.
Click to expand...

Couple reasons:
1. I want more RAM, like you mentioned
2. The RAM I have now are little toasters and I can't keep my fingertip on them for over 2 seconds when stressing them
3. I've already RMAed this ram set 2 times. I think the temp is a large factor to them dying.
4. My CPU can go well above 450 I suspect, but the ram craps out after 1068. It simply doesn't have enough voltage to overclock and I'm not giving it more otherwise I'll void the warranty and likely fry them.

I also disagree with the 1:1 ratio benefits. I've run several tests above 455x9 with the 1:1 ratio and my system benchmarks are a decent amount lower than when setup in 445x9 and using the 2.4 memory strap. I'm not talking a 1% improvement, I'm talking 5+% improvements, which to me seems pretty decent.

I suppose I could drop the multi to 8/7.5 and run the FSB at 500/533, but my experience in overclocking tells me the voltage required to reach this would be unbelievable/impossible. It seems to me if going beyond a 1:1 ratio gains very little, then RAM fsb and CPU FSB would scale together but they obviously don't. So my question is why are we replacing DDR2-800 with DDR-2000 if there is little to no benefit?
 
Running the RAM at above a 1:1 ratio gains very little (in comparison) to upping the fsb and running the RAM at 1:1. 1066 at 1:1 at 533 fsb is going to be a lot better than 1066 at 445 fsb.

Why do you think the voltages to reach 500 or 533 would be "unbelieveable/impossible"? Maybe for a Quad on a p35 or below(and I am not totally convinced of this as my RAM is the limiting factor on my fsb).. but not for a Dual as long as the NB and SB are cooled adequately.

And IF the RAM can actually run at what it is sold to run at, a fsb of 533 should be no problem.
 
I'll have to try to drop the multi and run at these speeds tonight, but I just don't think it's going to be possible. I'm currently running 4.0 with 1.275v right now and loading around 52'C, but it's hot as balls in my apartment. My suspicion is that the highest linked I'm going to be able to achieve is not going to be equivalent to benchmarks run with 4GHz proc, 1068 MHz ram unlinked. I'm thinking that running that high of a FSB is going to make the NB hot a hell too and my stock passive heatsink it isn't gonna be happy.
 
So grab an extra 80mm case fan and point it in the general direction of the chipset heatsink.
 
18 is # 1 said:
look here: http://www.hardforum.com/showthread.php?t=1418981
Click to expand...

Yeah, I already saw them. I really wish there were more reviews of these so I could tell if the 1200Mhz was just a fluke or something. I've seen several "Engineering samples" reviewed, and a friend even has one of the first e6600s which was an engineering sample, and they seem to work a bit better compared to the real product. Usually samples and review kits are pre-tested and high binned chips, but maybe that's just my bias based on the couple examples I've seen.

cyclone3d said:
So grab an extra 80mm case fan and point it in the general direction of the chipset heatsink.
Click to expand...

Last night I lapped both my TRUE and my cpu; nice mirror finishes now. I'm out of MX-2 and using AS5, but here's what I've done as far as testing and benching.
Processor Setup~~~~~~~RAM Setup~~~~~CrystalMark~~Memory Score~~Load Temp
444x9=3.995 GHz~~~444*2.4=1065 MHz~~~~179434~~~~~~~21984~~~~~~~47'C
445x9=4.005 GHz~~~445*2.4=1068 MHz~~~~179840~~~~~~~21738~~~~~~~47'C
500x8=4.000 GHz~~~500*2.0=1000 MHz~~~~178493~~~~~~~20660~~~~~~~51'C

Miraculously I only had to bump FSB voltage +.1 for the 500MHz FSB; I suppose I should admit my wrong in terms of judging the ability for my board/proc to overclock. My last proc didn't overclock nearly as well. For the test all other settings (volts, accelerators, OS, etc) kept the same between tests with only CrystalMark running at time of test. Each benchmark was run 3 times and then averaged. I did the above and below in the multi of 9 because I didn't want you saying that the reason it beat the other was due to 1 MHz FSB. I find it interesting the 500 FSB raised the temp 4'C when the only increase I made was the FSB +.1v for the Northbridge.

However, I've yet to attempt for the 1:1 scaling at 533 MHz, mainly because I won't be able to match it to the 1:1.2 scaling with the ram at that speed. I however think the result would be the same. This is a MASSIVE overclock over the stock 333 MHz overclock and the 1:1.2 is still beating the 1:1 ratio pretty handily, not just a minuscule benefit.

Questions? Comments? Concerns?

I'm still buying a kit today, probably in 2-3 hours from now so I can get Newegg to rush it today, and I would still love input as to my purchase. So please, anyone who has knowledge of DDR2 vs DDR3 please weigh in. Thanks
 
If it's simply DDR2 vs DDR3, on socket 775 I'd stay with DDR2 unless you get something like the ASUS Rampage Extreme or a fast P45 board and go for some really high OCing.
 
What are you cpu overclocking goals? I honestly don't think you going to get much more than 4.0GHZ with an E8400 on air. Most poeple get them to ~3.8Ghz-4.1Ghz on air. I would just get a 4GB DDR 1066 kit. You can overclock the ram a bit too if you need the head room.
 
You must log in or register to reply here.
Back
Top