Jump to content

joe111

Members
  • Content count

    9
  • Joined

  • Last visited

About joe111

  • Rank
    New Member
  1. cs02jhf: Glad you found your problem. At least you can continue running on 1 stick while Gskill replaces the other. That lifetime warranty is a good thing. I did notice your post said you are running tref at 3120 ([email protected]) is that something peculiar to the "expert" board? The refresh rate of the LE memory by SPD is [email protected] Have you tried RMMA (Right Mark Memory Analyzer)? You can change tref timings using A64 (on the fly) with RMMA running stability test and see the effect it has on memory stability.
  2. Sorry those didn't work for you. You haven't posted your voltages. If I knew a little more about your settings I might be able to help. The images in your initial post indicates a discrepancy in the tref value (one is 3.9 and one is 15.6). What HTT and multi were you running when you saved the A64 image in your first post 237x11? After you boot your machine and it is running....you change the settings using A64 and clockgen to the settings I sent, what happens? Lock-up? Blue screen? what exactly happens? If you want to run Cas 2.5 at 250 or above with "LE", you will almost certainly have to run Tras at 7 or 8. I know 5 generally will not work, 6 may but very probably will not be stable unless all the other settings are just right. 6 becomes unstable in the 265-270 range on my sticks. Give the following a try & see if you can bench at 254 or some other HTT value in the mid 250's. Select 2.7v or 2.8v Vdimm in Bios (without any +v option). You haven't specified your processor voltage in the posts but I assume it has been increased to a level you are comfortable with (personally I am running 1.48v, 1.35x110% on my opteron with an xp-90c as 24/7 setting)...see more on voltage below. Get it running. If you are at 11x, change the mulit to 10 with clockgen, then use A64 to change the settings. Increment the HTT with clockgen and run some tests...whatever you like. In A64, try the same settings as sent previously except adjust Trrd to 3, make sure async latency is 8.0 ns and read preamble is 5.0 ns. Turn off any "skew". You might also want to try changing tref to [email protected] or other values during testing (my old 3500 liked [email protected] when running 262 HTT) but use [email protected] first. If that works, go up until it doesn't. The big kicker for not booting may be the CPU "start up" voltage in your BIOS. Set to at least 1.45 before trying to get your machine to boot at the settings derived from your quick testing. Personally I don't think 1.5 is to high for "start up" on a 90nm core but that's just my opinion. It only runs that voltage for a few seconds until it reverts to the "operational" voltage. Which I am assuming you have changed to 1.35vX110% or something close that you are comfortable with. This is a shot in the dark....as I mentioned, I am not familiar with the "expert" version of this board (only the SLI-DR version) but I am still of the opinion that your memory is okay. Your "expert" board and the Gskill "LE" combination may just need a specific setting to be something that you don't expect (or that may not necessarily make sense). For example, there are some people that use a zero value in the timings when running BH-5. Additionally, you should be aware that if you are at the limit of your memory controller, the HTT and multi won't always interchange to produce the same end value i.e. 250x11 and 275x10....one may work and the other may not even when your memory is capable of both. Hope this helps.
  3. cs02jhf: I am inclined to agree with icon57, except for the 2.9 volts part. Some Tccd (depending on the batch) will throw errors with higher voltages...some like 2.9+. Being unfamiliar with the "expert" version of this board, I can't comment on the BIOS. There should be some info in the forum on which on is best with your memory. I am running the same memory with a 148 Opteron on a SLI-DR (not the expert version) and don't know all the differences between the two DFI boards. However, you might want to change your timings. They look pretty tight for "LE" at 260. My memory will run to 312 (stable) at 2.8v although timings are significantly looser. If you can run 300htt at 3-4-4-8, I don't think there is anything wrong with your memory. I don't believe that "LE" was intended to be a low latency memory even though it is advertised as 2-2-2-5 at DDR400, 2.5-4-3-7 at DDR550. Try the following a64 settings (v6.0 beta): 2.5,10,14,4,2,8,3,3,2,2,[email protected],1clk Dqs skew off 16x,128,Dis,Dis,7x,8ns,5.0ns,16 or 32clks,enable,200,enable,weak,disable memclk (all enable) These work for me, all the way to 305. You should be able to easily run cas 2.5 @ 260, 2.7v. If it boots, change the 8 to a 7. This is within the specification for the "LE". If you are after very low latencies, you will probably need another flavor. It also may be the case that you have found the limitation of the memory controller on your 3700. I had a similar problem with a 3500 that would not run over 2.65 no matter what...just a brick wall. After replacing the processor, using the same memory, same board, same everything....bingo 3Ghz.
  4. Thanks for the tip on .5 multis, but it has never been a problem in my setup. The .5's have worked just as well as whole numbers. My timings have always been 1T. The real issue here is loss of 60% of my bandwith. Here is the background info I posted at another forum. I apologize if it's a bit long winded.... ---------------------------- My LDT multiplier has apparently stopped functioning or is not functioning properly. Memory bandwidth has dropped from 7000 to barely 2000 in Sandra and Super Pi has gone from 31/32 sec to 42 sec for no apparent reason. I've isolated the problem to the LDT multiplier. After about a week of testing Refresh Rate (tref) values against Idle Cycle Limit values using prime95, I found that my memory bandwidth had dropped off from 7000+ to around 2000 and my super PI times had dropped from 32 / 33 sec to 42+sec. This was using the 310-p bios with FSB and multi settings of 264 x 10 and 274 x 9.5. Thursday evening when I powered up the system and ran some benchmarks for screenshots to send a friend, the problem began. Initial values indicated a problem...I had to reset cmos and re-enter my settings to run the benchmarks and obtain up till then "normal" results. Friday, power up revealed the same problem but this time, resetting cmos didn't work. This weekend, I have flashed bios with 414-3, 510-1, 510-3, back to 310p, then again to 414-3 using awdflash per the "correct" procedure outlined on DFI-street (except for removing the battery). None of the different versions made any difference. At this point nothing I've tried seems to correct the basic problem... With LDT set to anything other than "auto" I was unable to boot. It would boot using auto, (but the performance hit is tremendous) however, the system is stable. I tried 2x, 3x at 276x9.5...2x, 3x at 264x10 and 4x at 250x10 none will boot with any of the bioses listed above at anything other than LDT set to "auto". There is another anomaly that seems very strange and I don't understand... from time to time the when using clockgen to check settings, the PCIe frequency comes back as something other than 100 ( which is where I have it set ).....it may be shown as 83.49 or 89.23 or some other irrelevant number that doesn't make sense....Other times it returns the correct value. If I set it to 101 clockgen will report it as 101.9 ( which is probably okay ). It will take an updated setting from clockgen but that still doesn't explain it. The system will fail stability tests very quickly with the out of spec PCIe clock setting. I don't think this problem is in the clockgen program. As a last resort, this weekend I have moved the raid array from the NV controller to the 3114 controller and reinstalled XP & all the drivers. Installed 414-3 bios (and even removed the battery) It will now boot at 264x10 with LDT at 3x or auto but the memory bandwidth numbers and super Pi times are still bad....the LDT is just not working. Here is the system and settings: DFI SLI_DR R .A00, AMD 3500 Clawhammer (CAAXC), XP-90c G.Skill 2x512 "LE", Leadtek 6600GT, 2 x MaxlineIII 250GB (raid 0 on 3114) 680W Thermalltake. 264x10 1:1 and 274x9.5 1:1 (either will run stable) Mem 2.5,3,3,7 1T Vdimm 2.83v, cpu 1.62v, LDT 1.36v, NB 1.56v All voltages seem stable and max temps are okay. ------------------------------ Maybe it is not an LDT problem, maybe something else. Like I said before I'm stumped..no clue. Tried just about everything I can think of. Thanks
  5. Sharp: As you can see in my sig, I am running a fsb of 264 0r 276 by 10.0 or 9.5 respectively, so the answer is yes I specify the CPU ratio. It doesn't seem to matter what LDT is set to now, it appears the transfer rate is locked at something very low. Maybe my hypothesis is all wrong and it is something else... "Can't fix it if you can figure out what's wrong" With regard to clearing CMOS, this has been done at each bios flash and also independently (as well as even removing the battery). Memtest passes but the bandwidth is still bad. Thanks.
  6. Been working with my SLI-DR about a month now and thought things were smoothed out until runnng into a problem that I can't explain..... Memory Bandwith has dropped from 3000+ (memtest) , aprox 7000 (sandra) to approx 1000 (memtest) and 1700 (sandra) for no apparent reason. My Super PI times have gone from 32 / 33s to over 42s. Since this problem has developed, I have flashed the bios several times (tried 414-3, 510-1, 510-3, back to 310p, currently using 414-3). I have moved my raid 0 array from NV to the 3114 controller & reloaded xp / drivers. Nothing has made any difference. I don't know what to do next. The system is stable and runs fine at the settings I've been using for over a week (Passes memtest, runs all the bench test etc.) - it's just much slower. Does anyone have any ideas or experienced a similar problem? I started a thread on xtremesystems (LDT multiplier not working) hoping someone there might have some experience and / or solution. The is some additional background info there for anyone willing to throw out an idea. Please....any ideas at all. I'm stumped
  7. Wanted to update my post on the Maxline III drives.... The problems that I reported earlier have gone away. I am not quite sure how or why but my disks have worked very well for the last couple of weeks. After a fresh install of XP at the time of my last post, and a week of nothing but testing (much of which was I/O intensive) the problem has gone away. I backed my fsb off to 275 and upped the voltages .... no more problems. Maybe the controller just needed a good burn in. I have no other explanation At this point I am actually considering the purchase of two more to add to the array.
  8. Contacted Maxtor by phone from work but they couldn't pull up the information I had submitted on line in their website and I didn't have the serial numbers so they weren't willing to discuss much. The good news is the drives are behaving themselves now and I haven't had any corruption problems. It seems that there may be nothing wrong with the Maxline III drives, just a problem with the operator (me) and the multitude of settings on this DFI board. It appears that the drive problems were related to some of the settings I was attempting to use during testing....in some cases they were extreme, and as I have now learned not well balanced. The drives corrupting during a hard crash should have been an indication that the problem lied elsewhere.
  9. For those who may be interested, I'm running (2) Maxtor 250GB Maxline III drives on NV Raid 0, and am experiencing disk corruption problems as well. The 300's and the 250's are essentially the same except for size. Rgone referred me to this thread when I asked him about my drive corruption problems with the Maxtors. It sounds like a very similar scenario. Initial setup of my SLI-DR proved to be problematic and XP would not install without numerous BSOD's...It was finally accomplished with one stick of DRAM and loose timings. Subsequent testing of this configuration (with the original shipping bios) resulted in a number of corruption events requiring a reformat and reinstall. I flashed the bios...tried both 414-3 and 310p. Got both Dimms installed and running initially with 414-3 but experienced the corruption problems mentioned above. Flashed to 310p and things seem to be working better but I am afraid that really pushing the system will result in additional corrupttion issues. Currently I am able to run at 285x9.5 without the corruption problem (so far) but the board has more there. The limiting factor is definately either the drives or the controller....Benchmark tests will execute and complete when running from CD but lock up or cause a hard crash if attempted from the drive array or if the test requires disk I/O. I have contacted Maxtor by email and received the same response as "Alien"...due to the nature of the problem, call. I don't know if the firmware issues apply to the 7B series of disks (maxline III) but the constant reformatting and reinstalling of XP is getting old. I have used Maxtor for a number of years and own four other Maxtor drives - they are fast and dependable. Any contact with their technical staff I have had in the past has been swift and competent. I'm sure that they are becoming very aware of this problem as it doesn't seem to be occuring in disks of the other manufacturers. (It's not showing up in the forums or if it is, I just haven't found them) It may turn out that these drives just aren't up to the task at 300x9 or 290x9.5....I hope that's not the case. Performance should not be limited by disk drives. I will post again if the call to Maxtor reveals any definate fix or firmware update for these drives. Additionally, if anyone following this thread knows if or how it would be possible for the NV controller to unlock (or it's frequency to vary) I would be interested to know. I believe it is (or should be) locked at 100MHz and cannot be changed. Is this correct?
×