Hi to everybody!
I encountered several problems with this mainboard and I hope someone here on the forum is able to help me. It will take some time to read all that I wrote but I think if you will be patient, you will agree that is definitely worth it!
I highlighted every part that is important, and my questions are preceded by "Q."
First of all my configuration:
- 2 x Intel Xeon 2687W
- 4 x 8GB Samsung M393B1K70DH0-CK0 modules, ECC / Registered
- 1 x Seasonic Platinum 1000W PSU
- 1 x Samsung 830 SSD
- 1 x NVidia 7800 GTX
- Windows 7 Ultimate 64 bit OS
Please note that I kept the installed hardware @ the bare minimum , just to test and solve the issues I'm experiencing.
I used another graphics card from ATI too, cheap but way more recent than the old 7800 GTX.
Also you must keep in mind that I made a "raw" Windows installation , without any specified or updated driver for the hardware installed and I did not installed any of the software/drivers on the CDs that comes with the mainboard.
I will try to explain the problems encountered by dividing them into phases, making a chronicle of my experience with this mainboard, both indicating solutions to commonly encountered problems (when I managed to find them!) and asking you tips and explanations for the problems that still are there.
1) Initially I could not get a POST no matter what, the mainboard showed on Q-CODE display virtually the entire range of possible error codes, like a causal extraction of numbers.
With both 2 and just 1 CPU installed, and with all 4 or just 1 RAM modules.
I soon realized that the only possible explanation for such behavior was some kind of problem with the memory.
After dozens of attempts, resets, clearing CMOS, etc.. the board still refused to boot.
Then something very strange happened: I tried with some cheap RAM - same result - BUT after that, when I put back the Samsung modules, the mainboard has finally begun to "collaborate" and I had no problems getting a POST and boot.
As if, once "tasted" the worst, she understood that was best for her to be "satisfied" with the ECC / Registered PC12800 (1.600Mhz) modules from Samsung, @ 100 Euro x piece ..........
2) After some boots I noticed that the BIOS indicated that portion of the installed memory was not used, or rather seemed "invisible" to the system: at first only a "half" module (the BIOS gave about 28GB available instead of 32GB, so half a module of 8 GB was missed), then after other attempts and resets, even disappeared the entire RAM bank, showing only 24GB (!).
An option on the menu "Advanced / Chipset / Memory" showed sometimes the message (bank) "Present: 0 MB", sometimes the equally ominous "Present: Disabled".
Here, too, after dozens of attempts and after reversing the two CPU from socket 1 to 2 and vice versa, I managed to "convince" the mainboard to accept the harsh reality of the existence of 32GB of RAM on the DIMM .......
PLEASE NOTE that I have examined very carefully all the pins of both sockets and all of them are absolutely PERFECTS.
I read on variuos forum that other users have had this same issue with this mainboard and no one was able to find an explanation and it seems that they are not even able to overcome the problem: in this regard I have some luck, at least.....
Q. Someone can enlighten me on this issue?
3) In the process of course I tried to update the BIOS from the original 3206 to the newest 3302.....not only did not help to solve the problem of RAM "disappearance", but it has added another: after each system restart, the mainboard remained "alive" only for a few seconds and then suddenly turn off and restart, repeating the whole loop endlessly.
The only way to stop this behavior was - @ every single system restart"!!! - the CLEAR CMOS option.
Alternatively, I found out that setting jumper ME Recovery on the "enable" - AND KEEPING it as such - allowed the mainboard to systematically avoid the "deadly loop" but it is clear that this could not be a solution ( I do not think that the ME Recovery jumper must be always on) and the entire boot seemed slower than normal.
Q. By the way, what is exactly the ME Recovery? Enabling it what exactly does?
In the end, I found the definitive solution for pure casualness: after restoring the original BIOS 3206 and above all after disassembling the entire system and reversing the CPU from the socket 1 to 2, everything turned back to "normal" in the sense that the mainborad stopped the strange loop on every reboot.
Note that this issue,too, was already reported by some users, as I found out in other forum.
4) The BIOS shows that the processors speed is 1200 Mhz (instead of 3100) and also the voltage is much lower than normal (obvious consequence).
If I do a CLEAR CMOS, the mainboard sets speed again to 3100 Mhz but this ONLY LAST until the next reboot of the system, then lowers again the frequency to 1200 MHz.
All this does not disturb the performance because obviously Windows autonomously manages the CPUs clock frequency and voltages depending on the workload.
It seems to be a bizarre setting of this mainboard that seems to try to save energy at bootup.
Or is not? It can be a sign that something's wrong? I personally do not think that's the case because this "issue" has already been reported by many users (again!) on other forums and it seems to be a peculiarity of the board or - most likely - of the newer BIOSes.
Q. Can someone tell me more about it?
5) Now, the problems that remained unresolved.
Various monitoring software in Windows environment (such as CPU-Z, HWINFO64, AIDA64, etc.) show that both CPUs have a frequency of 1200 Mhz (which should be normal in the absence of workloads - Q. right?) BUT all the various cores - almost at random - experience a sudden increase in frequency with "Turbo"activation, but not caused by any software running or any element that can stress the CPUs , along with an equally sudden increase in voltage.
This phenomenon is repeated continuously without stopping.
Sudden changes of frequency and voltage (about 1.1 V), then return to 1200 MHz and minimum voltage (about 0.8 V).
Initially it did not seem to happen, I noticed this thing after updating the BIOS to version 3302: I've now restored the 3206 but the phenomenon persists.
Q. Now I ask: is this normal behavior or is a symptom that something's wrong?
6) Again RAM issue, and once again a strange one.
Although now the BIOS always recognizes the 32GB installed, in Windows there is something strange: some software (HWINFO64, AIDA64) shows 32 GB accurately (32768 Mb) while only CPU-Z show values slightly lower - but not correct (32718 or 32720 Mb, depending on how he feels ...).
Of course I thoroughly tested each memory module with various software and I did also stress test for over 24h with MemTest86 +,with absolutely NO ERRORS detected.
What's more, one of the CPUs recognizes and correctly handles all modules (whether I entrust her all 4 or just 1), and behaves perfectly on both socket 1 and 2.
While the problem seems to affect only the other CPU, that show it in both sockets: does not hold all 4 modules (i.e. if I install this CPU alone with all 32 GB entrusted to her, at least 1 module of 8 GB "vanishes"...) and also whether I install both CPUs the problem I described above (several megabytes "disappeared") happens, but ONLY if I installed all 32 GB (obviously divided between the 2 sockets).
It seems this CPU has some problem on the memory controller (at least this is my inference) ..... Q. this is the only explanation or there may be something else?
Curiously enough, if the "faulty" CPU is, for example, on the socket 2 and I insert the RAM modules on the DIMM "G1" & "H1" rather than the "E1" & "F1" (as described by the manual) CPU-Z shows the correct values (32768 Mb) but, as was to be expected, there is a decline in memory performance.
Another curious thing is that I tried to install both CPUs and put all 4 RAM modules only on socket 1 (where is the "healthy" CPU), leaving socket 2 - where sit the "faulty" CPU - free of RAM : the system recognizes all 32 GB correctly BUT while HWINFO64 & Aida 64 reports them running in Quad-Channel mode (!!!), once again CPU-Z only disagrees and can not indicate the "channel mode" in which they operate, furthermore showing that the RAM run at 500 Mhz instead of the normal 800 (as if they were DDR 1060 instead of DDR 1600).
Q. By the way, it is correct that the RAM setted on pair on both sockets as the manual reccomend (16 GB on socket 1, DIMM A1 & B1 and 16 GB on socket 2, DIMM E1 & F1) run in Dual-Channel mode (as happens in my mobo)?
Q. There is a way to run them (4 modules x 8 GB) in Quad-Channel on this mobo or I need 8 modules and fill all the RAM DIMM on both sockets?
I did some tests and benchmarks and there not seems to be big difference between the "correct" RAM configuration (from the manual: 16 GB on socket 1, DIMM A1 & B1 and 16 GB on socket 2, DIMM E1 & F1) and the wild one tried by me ... although there are some inconsistencies: benchmarks from AIDA64 not seems to feel any differences, while GeekBench do not like at all the 4 modules on a single socket and heavily penalizes the score.
Finally, PassMark tests are the most ambiguous because PCMark (evaluation of CPUs) approves and returns a very high score (which I believe is not reliable) while MemMark (of the same package, which assesses the memories) returns very poor values for the RAM performance.
Q. Therefore, I believe that is not a good solution to put all 4 modules on sockets 1 and probably it may cause a performance decline: what do you think?
Last question: when I install one CPU alone, even with just 1 RAM module, AIDA64 benchmarks for memory test show decent values - for ex., let's say 15.000 on "Memory Read test" - but when I install both CPUs, even with all 32GB of RAM, I get a sensible drop in performance - like 11.800 in the very same test.
Also "Memory Write & Copy test" values drops considerably.
The latency rises from, let's say, 69ms to 80.
And this happens no matter what (switching CPU from socket 1 to 2 and viceversa, switching memory modules, less modules, more modules, BIOS setting,etc.).
Q. Is it normal? I mean, with 2 CPU is somewhat "physiological" a decline in memory performance? Or ,on the contrary, it is a sign of something wrong?
7)Last "issue". As soon as I see the Windows logo on my monitor, the Q-CODE display of the board show value 22, and stays so forever.
I have not tested the system 24h/7, only some benchmark & simple software, so I can't be 100% sure that it is absolutely stable.
Altough I feel optimistic about that.
I did some search on internet and found out that no one can say clearly what this code 22 is, which is the cause and especially if it is a real problem or not: I've read about people who had this same report but whose system seemed stable at 100%.
Q. Somenone here has an exhaustive answer? And, by the way, what should be the "correct" code (ie that everything is ok in the system) that the Q-CODE display should show after boot?
Last notes: trying to solve all the issues described above I did several tests, for example using a video card, then the other one, moving them in each slot to see if there were benefits.
I've tried changing various parameters in the BIOS: first all on "Auto", then I tried to manually setting the values of the RAM timings, speed and voltage.
Disabled PCIE slots not being used, disabled the onboard video. And vice versa.
Q. Is there anything else I could try?
By the way, I also noticed that some of the CPU settings such as QPI Link 0s and 0c are set by the BIOS to "disabled", while only QPI Link 1 is "enabled": Q. is this normal? It's useful to enable the other two settings? What are they? it is better to disable all 3?
I made a few attempts but do not seem to have any effect neither positive nor negative on the system ....
If you have read so far, thanks!!! I hope someone can help me. Thanks in advance for every reply.
Edited by TheBlackHole - 4/30/13 at 6:01am