Sign in with
Sign up | Sign in
Your question

Is it harder to OC with 4 DIMMS vs. 2 DIMMS?

Last response: in Overclocking
Share
a b K Overclocking
March 20, 2010 2:27:02 AM

I renovated an older computer that we had at work. One of our CAD guys was using a custom built computer with Q6600, Intel motherboard, 4GB of ram and Windows XP 32bit.

With some of our current Revit projects they felt that they needed 8GB, and I thought I could OC the processor, so I replaced the motherboard, put on a good aftermarket cooler, bought 8GB of ram and Win7 Pro 64bit and loaded it up.

Here are the actual parts I used:
mobo: Gigabyte GA-EP45-UD3R
cpu: Q6600 (about 18 months old)
ram: G. Skill F2-6400CL5Q-8GBPQ
cooler: Xigmatek Dark Knight-S1283V
PSU: Antec Earthwatts 650W
GPU: ATI FireGL V3600 (about 18 months old)

I loaded the machine up and it ran great. I tested it a bit at stock speed of 2.4GHz, then tested it from 3.0 up to 3.4. At 3.4 it failed Prime95 testing, but ran fine at 3.3 with max. temp of 58.

In the process of installing apps we had some flakiness that I'm not totally convinced is a problem of the hardware. The CAD operator said he did have it fail in prime95 at one time so he took out 2 of the DIMMS and then it ran prime95 overnight.

I thought I had read that it is harder to OC with more DIMMS, is this true?

Are there other tests I can do to pin this down better? Should I try other things such as increasing ram voltage or loosening up the timings?

More about : harder dimms dimms

a b } Memory
March 20, 2010 9:56:36 AM

Yes, it's much harder with 4 DIMMs.
You could increase the northbridge voltage to get a better stability.
Test the stability running the "Prime 95" and you can use "Core temp" to control your cpu temperature.
m
0
l
a b } Memory
a c 197 K Overclocking
March 20, 2010 10:24:01 AM

Generally, yes, especially with Core2 CPU's.

The Q6600, depending on setup and the particular CPU chip, is usually good for 3.3 - 3.6 GHz. with a good air cooler.

You have one more possibility. Check to see that you are not inadvertently overclocking your RAM. If you are at 3.3 GHz, you are probably running 367 MHz X 9, yes? Look in your BIOS. Your memory clock should be running at 733 MHz.

If it is not, change your System Memory Multiplier from Auto to 2.00B. Reboot. If your memclock is now 733 MHz, try pushing your FSB higher.

Intel's max recommended CPU voltage is 1.3625 volts. Keep your temps under 70 C.
m
0
l
Related resources
a b } Memory
a b K Overclocking
March 20, 2010 2:48:50 PM

Yes, especially if they are not identical. If you only have two sticks, then one or the other may limit you. If you are using 4 then the slowest stick will limit you, your memory controller will probably have to use higher latency, and of course it's more stress on the memory controller. Changing your memory multiplier as jsc suggested will help with stability.
m
0
l
a b K Overclocking
March 20, 2010 10:17:26 PM

I bought the ram from newegg, and bought a set of 4 DIMMS so that they would match.

Yes, my ram is running at 733. Its rated at 800. I tried to loosen the timings but I didn't change the ram voltage or the northbridge.

I used one of my favorite coolers, the Xigmatek Dark Knight, and temps at 3.4 were 60 max.

My home computer is running a Q9400 which is hampered by a multiplier of 8. I bought better ram for it so I could run it at higher speed. It is only running 4GB but it has run at 3.8GHz under short term testing. Normally I run it at 3.4GHz, puts its ram at 850.
m
0
l
a c 104 } Memory
a c 121 K Overclocking
March 20, 2010 10:32:34 PM

Short answer: yes, it is harder.

When you use 4 sticks, you may have to up the ram voltage a notch. That is because The same voltage controller has to deal with a bit more load.
As long as you limit the increase to what is printed on the sticks, you are OK.

Run memtest86+ for a few full passes, and you should get NO errors.

Similarly, run prime95 fft with rounding error checking. Run it long enough for the CPU temperature to stabilize at it's maximum. Verify that there is no throttling. You should see NO errors if it is stable.

Do not bother with the timings, it will make very little difference in actual performance(vs. synthetic benchmarks) perhaps 1-3%.

m
0
l
a b K Overclocking
March 21, 2010 4:34:44 AM

This ram is rated 5-5-5-15, and the mobo reads this and uses it. The reason I mentioned timings before was because I wondered if it would help if I loosen up the timings a bit.
m
0
l
a b K Overclocking
March 23, 2010 12:06:52 AM

Update-
The 2 of us have been working with this machine a bit. Today I tried changing ram voltage and northbridge voltage, with no improvement. We put 2 DIMMS in it and stress tested it quite a bit with no problems, but put the other 2 DIMMS in it and it won't boot. I did some more experimenting and discovered that one of the DIMMS seemed to be the problem and it wouldn't run at any speed with that DIMM in it. I suppose it's RMA time.

I had some old DIMMS laying around so I'm actually testing it right now with 2 new ones and 2 old ones in it, 6GB total. If that works I've got some older 2GB sticks in another computer to try it with.

Funny thing is when I first upgraded the machine and loaded the new OS a week ago it ran fine with all 4 new DIMMS in it. It didn't start getting flaky until we started loading apps on it.
m
0
l
a b K Overclocking
March 23, 2010 4:13:09 PM

Another update-
The machine ran though an hour of stress test with 2 G.Skill DIMMS and 2 other old DIMMs in it that I had laying around, for a total of 8GB.

Convinced I had a bad DIMM, I contacted G.Skill for their recommendations. They said I should install the latest motherboard BIOS, and increase the northbridge voltage to 1.4. I was skeptical since I just bought this motherboard 6 weeks ago but I did what they asked and discovered that there was one newer version of the BIOS, and Gigabyte said that it dealt with memory issues. So I installed the new BIOS, put all of the G.Skill DIMMS back in and the machine loaded Windows. I've been stress-testing it at 3.2GHz and all is OK for 2 hours so far.
m
0
l
a b K Overclocking
March 24, 2010 8:11:07 PM

Yet another update-
Stress-testing the machine per G.Skill's recommendations (latest mobo BIOS, 1.4v northbridge) it finally failed, and curiously after it fails it won't reload Win7 when you try to reboot.

Reluctantly I fired it up with mixed memory, and it stress-tested in Prime95 for 12 hours.

I downloaded the memtest86+ iso, burned a CD and tested that. It ran fine for a couple of hours with either memory configuration.

At this point I've decided I can't trust the 4 sticks of new RAM so I'm going to have to RMA it. It was clear to me while experimenting with different combinations of modules in the machine that it always failed with one particular module in there, and never if that module wasn't in there.
m
0
l
a c 104 } Memory
a c 121 K Overclocking
March 24, 2010 8:51:44 PM

I think you have properly identified a bad set of ram, and it is appropriate to return it.

You might have better luck with a different set; Corsair and Patriot would be my first picks.

Go to the ram vendor's web site, and access their configurator.
Corsair, Kingston, Patriot, OCZ and others have them.
Their compatibility list is more current than the motherboard vendor's QVL lists which rarely get updated.
Enter your mobo or PC, and get a list of compatible ram sticks.

Here are a few links:

http://www.crucial.com/index.aspx

http://www.corsair.com/configurator/default.aspx

http://kingston.com/

http://conf.ocztechnology.com/index.php?c=1

http://www.patriotmemory.com/configurator/index.jsp

Cpu performance is not very sensitive to ram speeds.
If you look at real application and game benchmarks(vs. synthetic tests),
you will see negligible difference in performance between the slowest DDR2 and the fastest DDR3 ram.
Perhaps 1-2%. Not worth it to me.
Don't pay extra for faster ram or better timings unless you are a maximum overclocker.

---good luck---
m
0
l
a b K Overclocking
March 25, 2010 12:05:14 AM

Quote:
Don't pay extra for faster ram or better timings unless you are a maximum overclocker.


My strategy for purchasing this ram may have been incorrect. I wanted 8GB which pretty much means 4 DIMMs of 2GB each. I thought they had to be matched to some extent so I just went to newegg and searched for sets of 4x2GB. I sorted by price then started at the lowest price and worked my way up until I found a brand that I liked. I would have had a lot more to choose from if I had selected 2 kits of 2x2GB each.

I've had good luck with G.Skill memory in the past. When I built my home machine 14 months ago I did some research in Tom's memory test articles and selected G.Skill as one of the "good" brands. Since that time I've built 3 i5 machines using 8GB of G.Skill each with no problems. So the score stands 1 questionable DIMM out of 18 DIMMs I've bought from G.Skill in the past 14 months.

I'll do some more research to see if there are any problems with compatibility of the G.Skill memory and the Gigabyte mobo.
m
0
l
!