F2-8800CL5D-4GBPI on Gigabyte EP45-UD3P

[dclayw]

I have F2-8800CL5D-4GBPI running on Gigabyte EP45-UD3P with E8400 CPU.

Having some problems getting this memory stable at 1066MHz. No overclocking yet, just doing some basic testing with everything at stock. Application hang with Prime95 after serveral hours with the "Prime95 has encountered an error and must close......." error. This points to memory problems so I started testing with memtest86+ v2.01.

First I loaded optimized defaults in BIOS just to make sure everything was nice and clean. Then changed CPU Vcore to 1.9v (stock voltage), everything else left as is, including XMP at AUTO, so timings are set to 5 5 5 15 (stock timings). I get memtest errors in test 6, not every pass. (Test 6 seems to be the only test that gives errrors so I'm just testing with test 6). Error is always in the same place, either at around 1250 (module 1) or 3300 (module 2). Increasing MCH Core seems to help a little, delays error 'till around pass 6, but not totally consistent. Increasing FSB voltage (which I believe is called CPU Termination voltage on this board), also seems to help a little but again, not consistent.

I also have a very loose setting for tRD (static tRead) of 12, just to see if I can get things stable.

If I disable XMP, memory defaults to 5 7 7 20 at 800Mhz and it seems stable there, 10 passes of memtest with no errors. I don't think the memory is faulty but I can't seem to run the memory at 1066 without any memtest errors. I didn't think I needed to tweak all settings just to get the memory to run at stock.

Can you suggest what settings I need to tweak to get the memory running at 1066 without any memtest errors on this board. There are a lot of voltage settings on this board. Is it possible the memory is faulty, because it seems to be OK at 800MHz but not at 1066MHz

[G.SKILL]

I have F2-8800CL5D-4GBPI running on Gigabyte EP45-UD3P with E8400 CPU.

Having some problems getting this memory stable at 1066MHz. No overclocking yet, just doing some basic testing with everything at stock. Application hang with Prime95 after serveral hours with the "Prime95 has encountered an error and must close......." error. This points to memory problems so I started testing with memtest86+ v2.01.

First I loaded optimized defaults in BIOS just to make sure everything was nice and clean. Then changed CPU Vcore to 1.9v (stock voltage), everything else left as is, including XMP at AUTO, so timings are set to 5 5 5 15 (stock timings). I get memtest errors in test 6, not every pass. (Test 6 seems to be the only test that gives errrors so I'm just testing with test 6). Error is always in the same place, either at around 1250 (module 1) or 3300 (module 2). Increasing MCH Core seems to help a little, delays error 'till around pass 6, but not totally consistent. Increasing FSB voltage (which I believe is called CPU Termination voltage on this board), also seems to help a little but again, not consistent.

I also have a very loose setting for tRD (static tRead) of 12, just to see if I can get things stable.

If I disable XMP, memory defaults to 5 7 7 20 at 800Mhz and it seems stable there, 10 passes of memtest with no errors. I don't think the memory is faulty but I can't seem to run the memory at 1066 without any memtest errors. I didn't think I needed to tweak all settings just to get the memory to run at stock.

Can you suggest what settings I need to tweak to get the memory running at 1066 without any memtest errors on this board. There are a lot of voltage settings on this board. Is it possible the memory is faulty, because it seems to be OK at 800MHz but not at 1066MHz

could you tell which cpu you are using? and is it overclocked?
we'd recommend you to install them to slot 2 and 4 first
and please increase the NB voltage to 1.4v and set the timing to 5-5-5-15 2T@1.8-1.9v
TRD should not be the problem, if it still can't work with the setting above, try to increase tRFC
thanks


G.S

[dclayw]

Thanks. CPU is E8400 (see original post). Not overclocked, I'm just concentrating on getting memory stable first. I do have the memory in slots 1&3 at the moment. I will try 2&4, can I ask why you suggest 2&4?

With the last memtest I ran I had MCH (NB) at 1.24v and CPU Termination (FSB) at 1.24v. Errors started in pass 8 in memtest, so getting closer. But I will give 1.4v a go and report back.

[Tradesman]

What OS are you running? You mentioned going to optimized defaults, what FSB bus are you running...1333? Also what 's the CPU VCore, you mention it in your POST, but am thinking....hoping, you were referring to the RAM voltage at 1.9, not the VCore. You probably want the VCore about 1.1 to 1.15. @1333 FSB. DRAM Performance Enhance to Normal (believe turbo is the default). With this board and RAM combo you might want to try relaxing the timings to 6-6-6-18, 2T command rate, 1.8 (or 1.9).
If those don't work and you go to the tRFC (found under the Advanced Timing Control, I'd suggest trying 54, maybe 55-56. With your sticks, I think up around 80 would degrade performance...
The UD boards are pretty good with RAM but a little picky with the off standard speeds (900, 1100, etc)

[dclayw]

Running 32bit XP Pro. FSB is 1604MHz (CPU Host Frequency in BIOS set to 401).

Yep, that's a typo in my original post, CPU Vcore is 1.25v, the RAM voltage is 1.9v. I'm not overclocking the CPU at this stage, just trying to get memory stable. Performance Enhance is set to Standard.

The spec timings for this memory is 5 5 5 15 for 1100Mhz at 1.9v so I'm hoping to at least achieve that. I could relax the timings but I could have bought cheaper memory if I wanted to do that. But I agree, it's worth trying.

Either the memory is faulty or the board is being exceedingly picky with it. I have been concentatring on MCH and FSB voltages and I have tried pretty much every combination (within safe voltage limits) and I can not get error free memtest at 1066. That's with FSB of 401 and SPD of 2.66D (=1066Mhz). I have run many memtests on this memory and I'm getting inconsistent results. I think I'm getting close to stable settings when memtest doesn't fail 'till pass 10 or so, but then I run the test again with exact same settings and it fails at pass 1.

I bought fast memory and I don't want to push FSB beyond 1600. (I'm only going to look for a mild overclock of 3.6Ghz on my e8400 (401*9)). I did try FSB of 444 with SPD set to 2.40B (=1066Mhz) and this ran about 15 passes of memtest without error. I don't know why (444*2.4) would seem to work when (401*2.66) doesn't, with every other BIOS setting exactly the same in both tests. I'm going to have to run that test again. I'm not sure of the consequences of using the "B" strap for FSB in the 400s, the "D" strap is supposed to be used for that.

I'm starting to think this memory is faulty. I've run several tests with the memory running at 800Mhz and it seems OK at that speed. I'm running another 800Mhz test right now and will leave it overnight. If that fails I'll have to look at RMAing.

I'm also now testing in slots 2&4 with the same type of results I was getting in slots 1&3, i.e. memtest errors. I've also come across a couple of other users who have the same board, same poceessor and same memory and I've used their settings with no success. They were using MCH voltage of 1.26 and 1.3. So far I've only tried up to 1.3.

[Tradesman]

What is your tRFC, about 46-48? Might try 52-54, or if there inch it up slowly.

[GSKILL TECH]

yes for F2-8800CL5D-4GBPI, trfc require 48+, 54 is the best

[GSKILL TECH]

memory maybe bad. however, to get both bad, chance of that is very low.
i will suggest you install only 1 stick a time. if both not working, it might be other things. i got to tell you, EP45-UD3 is one of the picky memory. this memory is not happen a lot of problem on Asus board. i am not saying GA is no good but it just this board. lots of issue from 1066 and up

Running 32bit XP Pro. FSB is 1604MHz (CPU Host Frequency in BIOS set to 401).

Yep, that's a typo in my original post, CPU Vcore is 1.25v, the RAM voltage is 1.9v. I'm not overclocking the CPU at this stage, just trying to get memory stable. Performance Enhance is set to Standard.

The spec timings for this memory is 5 5 5 15 for 1100Mhz at 1.9v so I'm hoping to at least achieve that. I could relax the timings but I could have bought cheaper memory if I wanted to do that. But I agree, it's worth trying.

Either the memory is faulty or the board is being exceedingly picky with it. I have been concentatring on MCH and FSB voltages and I have tried pretty much every combination (within safe voltage limits) and I can not get error free memtest at 1066. That's with FSB of 401 and SPD of 2.66D (=1066Mhz). I have run many memtests on this memory and I'm getting inconsistent results. I think I'm getting close to stable settings when memtest doesn't fail 'till pass 10 or so, but then I run the test again with exact same settings and it fails at pass 1.

I bought fast memory and I don't want to push FSB beyond 1600. (I'm only going to look for a mild overclock of 3.6Ghz on my e8400 (401*9)). I did try FSB of 444 with SPD set to 2.40B (=1066Mhz) and this ran about 15 passes of memtest without error. I don't know why (444*2.4) would seem to work when (401*2.66) doesn't, with every other BIOS setting exactly the same in both tests. I'm going to have to run that test again. I'm not sure of the consequences of using the "B" strap for FSB in the 400s, the "D" strap is supposed to be used for that.

I'm starting to think this memory is faulty. I've run several tests with the memory running at 800Mhz and it seems OK at that speed. I'm running another 800Mhz test right now and will leave it overnight. If that fails I'll have to look at RMAing.

I'm also now testing in slots 2&4 with the same type of results I was getting in slots 1&3, i.e. memtest errors. I've also come across a couple of other users who have the same board, same poceessor and same memory and I've used their settings with no success. They were using MCH voltage of 1.26 and 1.3. So far I've only tried up to 1.3.

[dclayw]

OK, I let memtest86+ run overnight with FSB 444 and 2.40B strap (=1066), test 6 only at this stage but 45 passes or so, no errors. Just finished 5 hours success Prime95 blend tests. So looking good now. I don't know why I can't use the "D" strap, that is the strap that is supposed to be used for FSB at 400+ on this board. Anyway, that's looks like a board issue, not memory.

So the memory subsystem is looking stable now. For these tests I manually set the main timings 5 5 5 15, all the sub timings are on AUTO. tRD defaults to 13 and tRFC is 68. These are actual values read via memset utility. I'm expecting I should be able to lower both of these for better performance. I'll be aiming for 8 and 54 as you guys suggested.

Thanks for all your suggestions, it is appreciated. I might be back depending on how further tetsing goes.

[Tradesman]

Have fun, will be watching for your post

Have a gud 1 !

Tman

[GSKILL TECH]

OK, I let memtest86+ run overnight with FSB 444 and 2.40B strap (=1066), test 6 only at this stage but 45 passes or so, no errors. Just finished 5 hours success Prime95 blend tests. So looking good now. I don't know why I can't use the "D" strap, that is the strap that is supposed to be used for FSB at 400+ on this board. Anyway, that's looks like a board issue, not memory.

So the memory subsystem is looking stable now. For these tests I manually set the main timings 5 5 5 15, all the sub timings are on AUTO. tRD defaults to 13 and tRFC is 68. These are actual values read via memset utility. I'm expecting I should be able to lower both of these for better performance. I'll be aiming for 8 and 54 as you guys suggested.

Thanks for all your suggestions, it is appreciated. I might be back depending on how further tetsing goes.

well... motherboard strap is not always work. bios might have bug. MB RD is not able to test each strap with different FSB..... memory controller doesn't like that strap.. all possible

[dclayw]

OK, I let memtest86+ run overnight with FSB 444 and 2.40B strap (=1066), test 6 only at this stage but 45 passes or so, no errors.

I've still been testing this past week. I've run this test again with exact same settings and it failed. I'm guessing this points to faulty memory.

I don't know why I can't use the "D" strap, that is the strap that is supposed to be used for FSB at 400+ on this board. Anyway, that's looks like a board issue, not memory.

No matter what I do I cannot come up with any settings on the D strap that will not give memtest86+ errors. There are plenty of other people using this board with same memory and processor and using the D strap. Maybe my board is faulty. Maybe the memory is faulty.

I have been mostly testing with the B strap. There are very few combinations of settings (voltages/timings etc) that will not give memtest errors. I have spent countless hours testing this. This again points to faulty memory. I gave up on FSB444 because the memtest results were not consistent. The best I have now is FSB400 with 2.40B strap which gives me 960Mhz for the 8800s at 1.8v, so I am running them under spec. While these settings appear stable (memtest results appear consistent acrosse several runs), I have not completed Prime95 stress testing, although preliminary results look promising. FSB400 is also limiting my ability to overclock my CPU.

I'm concerned the memory may be faulty and I have just managed to find a sweet spot where it may currently work, baut fail later. I guess I won't find out 'till I've been running the machine live for a week or two. It's my view that it should not have been so difficult to get this memory working with this board and processor. Memtest errors at almost all settings combinations I tried does not give me much confidence in the memory. It's my experience that there should be many settings combinations that will run stable, or at least not give memtest errors.

I have not tried MCH voltage of 1.4v as suggested as that seems very high to me. Most people with this board/processor/memory are running 1.2-1.26v. Gigabyte tech advised I should not run MCH at 1.4v. I do not want to damage my board.

I don't want to be seen to be blaming the memory but I think I've given it as good a run as I can. Based on what I've written above should I look at RMAing this memory? I cannot say for sure that the memory is bad, it's only the memtest results that are pointing me in that direction. I have not yet tested each stick individually but will do that before I RMA, if that's the path I take.

[genetix]

well, you ain't gonna get to the D strap on anything below 1.4v(In bios this is 1.35-1.39v range) that's for sure. I don't own the board, but know that much. I run my stuff on 1.480v(In bios this is 1.39v-1.42v range) regularly to get stability on OCs and you are telling me 1.20-1.26v which would never work with decent OC even while the board is stable. Consider that 1.20v also isn't exact actual value. If you set something on BIOS != same as what it's actually (pretty much ever). Anyway, I'd consider having Northbridge over 1.5v to be 'limit' of an sort and even that when no effient cooling solution is on box. (You can easily see how effective cooling solution is (and actual voltages) with tools like CPUID 'HWMonitor' Lets say when NorthBridge'/MCH goes near/over 50c that's too much for 24/7 run).

Also about the memtest. that test is very very picky on tWR value specially when you are trying to run this memory on 1066Mhz. So I'd check tWR values a bit higher and try that way (start from 10 and above on 1066Mhz).


About RMA and well generally 'is the memory bad' seen errors on memory own couple bad blocks here. Best way to see if the memory is bad is to see does the errors on same set occure in same place several times, if errors are in random memory locations that usually mean some latency is simply too tight.

anyway there's my 2 cents... (just opinions really. Someone correct me, if I'm wrong.. ;P)

[dclayw]

well, you ain't gonna get to the D strap on anything below 1.4v(In bios this is 1.35-1.39v range) that's for sure. I don't own the board, but know that much. I run my stuff on 1.480v(In bios this is 1.39v-1.42v range) regularly to get stability on OCs and you are telling me 1.20-1.26v which would never work with decent OC even while the board is stable.

I assume you're talking about MCH Core voltage here. While it appears a lot of people with this board are running with the B strap for FSB 400+, there are still quite a lot running with the D strap with MCH Core around 1.20-1.26v, with 1066 (or greater memory). Just look at this thread for examples: http://forums.tweaktown.com/f69/stab...oltages-30561/ The very first example in that link is using 2.66D at 400FSB (1066 memory) and MCH Core of 1.20v. With all my research on this the only people I have seen running with MCH around 1.4v on this board are those who want around 500+FSB. I am only running with 400FSB.

I also have a Gigabyte X38 board and it has been running stable oc for 12 months on the the D strap with 1066 memory (400FSB) and MCH Core at 1.20v.

Also about the memtest. that test is very very picky on tWR value specially when you are trying to run this memory on 1066Mhz. So I'd check tWR values a bit higher and try that way (start from 10 and above on 1066Mhz).

Thanks for this tip, no-one had ever mentioned that before. I ran some tests with tWR set very loose and I must admit the results were better, although still memtest errors.

[genetix]

Yeah, noticed I am wrong on the MCH voltage on this case (Although, I still do not understand how same chipset different boards could act so much differently. Could be ofcourse be the power base). The boards start voltages are different than what I own here and am testing. Seems also chip acts a bit differently beside the multiplier.


tWR still shouldn't be below 10 and, if I were owner of that board on the tweaking forum you included. I'd actually check what the read to write etc values actually print on OS with MemSet (even if they wouldn't be wrong). However, that's just an side comment. Also I would sure check what the MCH/DRAM References & DRAM Terminations prints out (I am sorry I dislike anything saying 'Auto') or well does that rise correctly with DDR Voltage.



(Anyway, that's just paritial check on the OC someone('Outer Marker') made. I seriously don't understand couple other things on the board setup just makes no sense. Like who closes CPU virtualizations or steps or why there's an return 0 XD while he has closed Vitualizations, but that's another matter...)