G.Skill memory for EP45-UD3P

[dclayw]

Re: G.Skill memory for EP45-UD3P

There seems to be quite a craze around buying 1066 and above RAM in combination with CPU's that have a respectably high multiplier. You just won't see as much performance from "high speed" RAM as you will from a fast CPU.

95% of the performance increase will be in benchmarks like Sandra, realworld CPU beats RAM speed hands down. Running RAM on any kind of divider also introduces extra latency as RAM can only send data down a bus when it's ready to transmit that data.

I'm not saying give up on getting max RAM speed, rather if it's holding you back, go for your processor first, then maybe try again later. Who knows, the tweaking you do to get stable at 1:1 may just be what's needed to also reach 1100MHz-DDR.

I think you might be right. I've put 2 builds together now and had nothing but problems with 1066 memory on both of them. I was attracted to the g.skill 8800 PIs because it was fast memory and ran at low voltage. My previous build was with crucial 8500 tracers rated at 2.2v and they were too hot to touch. Didn't realise it was going to be such an effort to get stable settings for this memory on this board.

I'll persist a little longer, then probably drop back to 800Mhz and come back to it later.

[dclayw]

Re: G.Skill memory for EP45-UD3P

Are there 2 performance profiles on that memory?..even if you just read the advanced settings and enter them manually..Sandra shows some of them as does everest.

CPU-Z only shows one EPP. I'll get the advanced settings as shown by memtest and enter those manually and see how that goes.

On multipliers I found B better most of the time but have a D multiplier running now at 490 for 9gb/s synthetic bandwidth it showed 9.2Gb/s at 500 on the D multi but I cant get it stable.

This is what I'd like to know. Is it OK to use B multiplier for FSB at 400 and above. The BIOS indicates that B is for FSB 333. I don't really understand why there is different types of multipliers (B, D etc) if it turns out they can be used interchangeably.

[dclayw]

Re: G.Skill memory for EP45-UD3P

Do you have Extreme Memory Profile(X.M.P.) set to Auto or to Disabled?

On my UD3R, if I leave it Auto, the BIOS picks up the main 4 speed settings from the EPP settings, but all other timings come from the SPD settings resulting in a mixed setup which can't be ideal. I have now set XMP to Disabled and manually set the RAM up according to the EPP values displayed by Memset and Everest.

I started off with XMP at AUTO but now set the timings manually. I haven't set all the advanced timings manually but will give that a go too.

[grishkafool]

Re: G.Skill memory for EP45-UD3P

I don't know about the Pi memory, but I have 8GB of Gskill PC2-8500. I have been able to run stable (24hr Prime) at 3.oGHz, 400x7.5 & 3.4GHz, 400x8.5) with no errors, but I have been careful not to push my memory above 1068MHz.... but to make a comment on theme with your original question, I like my RAM... I especially like the shade of blue they came in, matches my Gigaboard nicely...

Is there any problem using the B strap for FSB in the 400's

I think I remember Lsdmeasap saying something about Bstrap wanting more volts than D...?
It may or may not have been in this thread, I don't remember.

GL with your overclocking though, sounds like you know what you're doing to me.

[MarktheC]

Re: G.Skill memory for EP45-UD3P

I don't really understand why there is different types of multipliers (B, D etc) if it turns out they can be used interchangeably.

On my UD3R, I can boot and memtest86+ and Prime95 with 2.66C, but can't even POST with 2.66D...

[dclayw]

Re: G.Skill memory for EP45-UD3P

On my UD3R, I can boot and memtest86+ and Prime95 with 2.66C, but can't even POST with 2.66D...

What setting are you using for FSB? I never did test the A and the C straps. I tried both just now with FSB at 400+ and memtest just spits out hundreds of errors. The "B" strap seems to work for me but the machine sometimes won't start after exiting the BIOS, just shuts down totally. And sometimes when getting into the BIOS it gives that warning message that some of the settings shown may not be what is actually being used. And sometimes when exiting memtest the machine will reboot a couple of times, and sometimes just turn itself off. Also the machine will reboot a couple of times when all I've done is increase MCH voltage by .02v. So I'm a bit suspicious of using the B strap above 400 FSB and I don't really trust it. But it is the only one that has given me no errors in memtest after 15 passes with memory at 1066. I've only run one test with it so far, I've been concentraing on using "D", the supposedly correct one for 400+, and the one that does not cause all those spurious reboots and shutdowns. (I know it's normal for the machine to reboot a couple of times after changing settings, but what's happening when I use "B" is not normal).

Can anyone else comment on the use of "B" for FSB 400+. Is it something that shouldn't be done. I've read about it somewhere but during my countless hours of research I can't remember where.

Update: Here is some info on it: http://forums.tweaktown.com/f58/2-00...arision-31656/

I've also been posting over at the g.skill forums too. First bit of advice I got there was to increase MCH voltage to 1.4v. I'm happy to take the advice from the g.skill techs but doesn't that seem a bit high, I only have 2x2GB. I don't want to damage the board.

[dclayw]

Re: G.Skill memory for EP45-UD3P

I think I remember Lsdmeasap saying something about Bstrap wanting more volts than D...?
It may or may not have been in this thread, I don't remember.

I remember reading about it somewhere too but can't find any info on it, I don't think it was in that thread.

[kick]

Re: G.Skill memory for EP45-UD3P

Yeh i meant to link u lsds strap comparison.
I found the best results on 2:00B(333mhz) early on i mean in terms of cpu overclocking D just didnt seem to work so well from 400-500 cpu/fsb ref speed.
As i said tho im now on a D for increased bandwidth with less voltage (lower tRD).
You already know the board and the cpu..If id just bought some 1100 mhz memory id be tryin to get it to work at 1100 or very close to it because its supposed to :0.
So assume you cant get it on a 2 multiplier (550 fsb)tho maybe that is possible with a dual and hi voltage.
leaves you with the 2.40 B (333) or 2.66 D (400).
Perhaps cpu with a multiplier that has it running under 3.5 ghz just to see first?
2.40 B would get you a higher base fsb ,higher fsb= better bandwidth,if the memory timings dont change.

Some memory doesnt run happily below its rated speed (so ive read) but you should have some advantage with 1100 mhz tested memory.
Perhaps you can run it on 4-4-4-12 timings or 4-*-*-* on a 2 multiplier to above 460 fsb?

[dclayw]

Re: G.Skill memory for EP45-UD3P

I did try running the memory at 1100, no change in results. I guess I'm sticking with 1066 as that works out nice and even with FSB 400 and 2.66D, no other reason. I'm perfectly happy with 400 FSB and only want the e8400 at 3.6GHz (400*9). I can always push it harder later, just want a reasonably fast stable machine for now.

Actually I'm testing memory right now so I have the e8400 running with only 6 multi atm. I'm currently just on 2 hours with Prime95 blend tests with 2.40B strap (FSB 444). If that proves stable then I'll be happy with that and move onto the CPU. I need to check but I think tRD defaulted to 13 (I had it on auto) with 5 5 5 15 and 2.40B strap. I'll see if I can get it lower then that.

I haven't tried 550 FSB as I prefer to keep things cool, but when I get time I'll give it a go. From my research it seems that higher CPU multi is preferable to higher FSB, lots of debates around on that though.

This all came about because with the board pretty much at auto for everything Prime95 fell over with what indicated memory to me. So I started testing memory and found I just cannot get D strap to run stable. I don't know if it's the board or the memory, I could try other memory, but I'll be happy if I'm stable with B strap. The settings I have now are MCH voltage of 1.24v and FSB voltage (CPU Termination) of 1.22v. They work with B, but not D.

[MarktheC]

Re: G.Skill memory for EP45-UD3P

What setting are you using for FSB?

E8400 @ 333 MHz x9.

I must be rare, because I'm NOT trying to OC my PC, I just want to learn as much as I can so I can manually configure my G.Skill PC7200 to run @ spec. If I had known it would be so hard, I would have bought RAM that was certified by Gigabyte for my mobo. I have the RAM @ 444 Mhz now Prime95 10 hours, Everest read ~7400 MB/s so I'll stop at that point.

[kick]

Re: G.Skill memory for EP45-UD3P

Oh yeh higher cpu speed the best,within reason,tho, highest fsb with highest cpu speed as long as you dont have to gimp the memory timings too much is the general plan .
Me< a lower multiplier isnt really an option. 6 or 7 are the only choices and 6 doesnt cut it.
for a 10mhz(20mhz mem) fsb speed diff my memory is working better at a lower voltage,sure the cpu loses 70 mhz but I feel the total performance is about the same.kinda like the "Prof Jimi Hendrix experience"

http://forums.tweaktown.com/members/profjim/

[Lsdmeasap]

Re: G.Skill memory for EP45-UD3P

1.4 MCH Voltage is not suggested at all, and yes that is very high for sure!

P45 is a tricky beast, sometimes when you would expect more MCH to be needed less is the answer, the amount needed also directly depends on the amount of Vtt applied, MCH Ref and tRD in use

You can use B @ 400+ FSB, it is just harder to do depending on the ram you have, and of course some ram will not do it at all

I did do a simple strap comparison, but of course that does not show all that is possible with higher speeds or higher multi's on the straps. >>

http://forums.tweaktown.com/f58/2-00b-vs-2-00d-399-500-fsb-simple-strap-comparision-31656/

May or may not be the thread you was thinking of?

[dclayw]

Re: G.Skill memory for EP45-UD3P

Well, I'm getting pretty frustrated. I only wanted 3.6Ghz but it was pretty easy to push it to 3.8Ghz so that's what I'm running now. I need to post all my settings to get some feedback. Will do that when I get a bit more time. Thanks for being patient.

This is what I have done with current settings. I ran 13.5 hours Prime95 with small ffts with no errors. I also ran 15 passes of InteBurnTest (several times) and that passed too. This indcates the cpu is stable. With all my memory testing I was just testing with test 6 in memtest86+ as that was the only test giving errors, and it takes so long to run the tests. Maybe running just test 6 was not such a good idea. I let it run overnight (about 50 passes on test 6), no errors. I had previously run 5 hours of Prime95 blend tests and that was stable too. As a final check I ran memtest86+ (all tests) and it gets errors in the 1st or 2nd pass, it's not a one off, I've ran it a few times. I'm currently 9 hours into Prime95 blend with no problems. But I'll post all my settings tomorrow and would appreciate any feedback on what you guys think might be wrong.

So that's where I'm at. There is obviously some small issue somehere but I can't find it. Maybe the memory is a bit dodgy, maybe I need to tweak some timings, maybe memtest86+ is telling me lies, (I don't think memtest86+ ever lies does it). All my timings are pretty lose by the way. I'm running 5 5 5 15 at 1.9v (spec) and tRD set at AUTO gets 13 and tRFC at AUTO gets 70.

A couple of quick questions. What is the actual value for CPU vcore. Is the value shown in CPU-Z accurate? I curerntly have it set to 1.29375v in the BIOS. After reboot PC Health shows 1.284v and CPU-Z shows 1.232-1.248v. I think these values sound reasonable for a E0 E8400 running at 3.78Ghz (445*8.5). Do you think so?

This is only my second build and the approach I've been taking is to get the memory subsystem (RAM and NB etc.) stable and once that is OK then move onto overclocking the CPU. When getting memory stable you generally have to increase MCH and FSB voltages a bit to find the sweet spot (for fast memory speeds anyway). Once that is all done and you move onto the CPU is it then only necessary to tweak CPU vcore. cpu clock and maybe the reference voltages, do you ever need to go back an tweak MCH and FSB voltages again?

[kick]

Re: G.Skill memory for EP45-UD3P

I believe the cpu core voltage shown by cpu-z is accurate it also agrees with speedfan and cpuid hw monitor etc.

For me the vdiff (vdroop or whatever) is approx .025-.03v with LLC enabled while under load.
For memory the voltage drop is approximately the same .ie i have 2.08v set in bios ,when running under load this drops to 2.05.
If the board is set to run auto ie standard ddr2 memory v it inputs 1.82+ in bios.
My other Gigabyte board which has no memory voltage adjustment gives the memory 1.84v in bios.
I assume this is some kind of mbrd designer "lets give it a little bit over to compensate for voltage drop/variations" thing.

I havent noticed setting different ref settings or vTT making much difference,(those that work anyway).
Some of this may be due to different psu's but most people get it as far as I know.

At v high(above 1.4v)cpu settings the voltage while running gets closer to what is set in bios.Im not sure what causes that.

OCCT running with a monitoring suite such as speedfan or everest gives nice graph readouts of all the voltages over the course of its test run . If the test fails you can see at exactly what temp or voltage the different components were, at the point of failure.

on your last point,I had a 500fsb set up withmost of the ref settings done manually on, a B multiplier(mem voltage was v high tho).
Now im on 490fsb on D and I still havent everything set.It is stable but with a lot of auto settings,ill get there .

[dclayw]

Re: G.Skill memory for EP45-UD3P

At v high(above 1.4v)cpu settings the voltage while running gets closer to what is set in bios.Im not sure what causes that.

I found that out. When I was moving to around 3.8Ghz I thought I would set vcore on auto to see what it needed. CPU-Z showed 1.424v!! Silly thing to do in hindsight, I quicky shut down and changed it, BIOS health reading was giving 1.428v.