ziddey
03-30-2007, 12:17 AM
Just wondering what you guys are getting on your gkx's / firestix1000's.
I'm running mine at 2.18v with a fan blowing down. Figure since they're firestix 1000 with rated volts of 2.1v, I don't want to run much more than 2.18v (less than 5% tolerance) to keep my warranty in tact. Actually, setting 2.1v results in 2.18v, so that's where I'm getting my numbers from.
But just curious what you guys are seeing.
For instance, at this voltage, I'm seeing 550mhz 5-5-5-15. However, at 500mhz, I can't seem to keep 4-4-4-12 without running into errors really quickly. It seems that one stick can handle it if I keep a fan over them to keep them really cool. However, the other one insists on throwing errors (just two in test5, however, the longer I let it run, without a fan, the more errors I get. with a fan, it's the same two errors every time).
So just curious how these gkx breeds do. Figure it'd be a worthy post since there's a firestix800 / gmh thread going around. All in all, I'm quite disappointed. I've got a corsair ramcooler on the way to hopefully give me a more permanent fan cooling solution than just resting a 120mm on top of the modules. I do have a 120mm about 2.5 inches away mounted on my case, blowing down, but I'm figuring they might not be that effective, since my ocz 8800 gold xtc burnt up after a month of usage at spec'ed volts of 2.4v.
It does look like these firestix are stable at rated speeds of 500 5-5-5-15 2.1v, for what it's worth, so these modules aren't defective.
FWIW, for those curious of what to run subtimings at, these gkx modules have sub's at (in asus order) of 5-35-3-11-5-13, while running 500mhz off spd. Figure that'd be something for you guys to base subs off on asus boards where there's no "auto" setting.
edit: I'd rather not run more than 2.18v, since these will hopefully be my keeper sticks, and I want to maintain warranty, since my past experience on 2 different sets of d9's (albeit gmh) have both died within a few months of operation at not xtreme voltages (<2.3v, and with direct airflow).
edit: My friend got these firestix too. I tested them in my machine. Running 2.18v 500 4-4-4, it gave two errors at 99% of test5 in the second stick. Then I threw my 120mm fan on top blowing down, and so far it's 6 loops in test5 stable. It's good to see it stopped erroring with a fan, proving a heat issue, but it still makes me a little wary, knowing that if I for some reason ever didn't have the airflow I needed that the sticks would throw a fit. Grr.. Guess I really should learn to accept that there's always a reason for something being cheap. update: passed 14 loops test5 4-4-4 no prob before I canceled. guess my friend's sticks are better than mine. bleh. edit: hmm passes test5 14 loops but bsod when enter windows. oh well. screw gkx
I wish there was a program that could just run loops of latency checking. Currently, I'm finding myself tweaking timings and subtimings with memset and constant iterations of mbench, hitting ctrl+c after the latency reading. I figure once I get latency squared away as best as possible, I should be in a good bandwidth position as well. It's such a damn shame that I've got 4 firestix1000 in my possession to test with that are all way under expectations, even in single stick testing. guess these must be 7xxxx d9's.
I'm running mine at 2.18v with a fan blowing down. Figure since they're firestix 1000 with rated volts of 2.1v, I don't want to run much more than 2.18v (less than 5% tolerance) to keep my warranty in tact. Actually, setting 2.1v results in 2.18v, so that's where I'm getting my numbers from.
But just curious what you guys are seeing.
For instance, at this voltage, I'm seeing 550mhz 5-5-5-15. However, at 500mhz, I can't seem to keep 4-4-4-12 without running into errors really quickly. It seems that one stick can handle it if I keep a fan over them to keep them really cool. However, the other one insists on throwing errors (just two in test5, however, the longer I let it run, without a fan, the more errors I get. with a fan, it's the same two errors every time).
So just curious how these gkx breeds do. Figure it'd be a worthy post since there's a firestix800 / gmh thread going around. All in all, I'm quite disappointed. I've got a corsair ramcooler on the way to hopefully give me a more permanent fan cooling solution than just resting a 120mm on top of the modules. I do have a 120mm about 2.5 inches away mounted on my case, blowing down, but I'm figuring they might not be that effective, since my ocz 8800 gold xtc burnt up after a month of usage at spec'ed volts of 2.4v.
It does look like these firestix are stable at rated speeds of 500 5-5-5-15 2.1v, for what it's worth, so these modules aren't defective.
FWIW, for those curious of what to run subtimings at, these gkx modules have sub's at (in asus order) of 5-35-3-11-5-13, while running 500mhz off spd. Figure that'd be something for you guys to base subs off on asus boards where there's no "auto" setting.
edit: I'd rather not run more than 2.18v, since these will hopefully be my keeper sticks, and I want to maintain warranty, since my past experience on 2 different sets of d9's (albeit gmh) have both died within a few months of operation at not xtreme voltages (<2.3v, and with direct airflow).
edit: My friend got these firestix too. I tested them in my machine. Running 2.18v 500 4-4-4, it gave two errors at 99% of test5 in the second stick. Then I threw my 120mm fan on top blowing down, and so far it's 6 loops in test5 stable. It's good to see it stopped erroring with a fan, proving a heat issue, but it still makes me a little wary, knowing that if I for some reason ever didn't have the airflow I needed that the sticks would throw a fit. Grr.. Guess I really should learn to accept that there's always a reason for something being cheap. update: passed 14 loops test5 4-4-4 no prob before I canceled. guess my friend's sticks are better than mine. bleh. edit: hmm passes test5 14 loops but bsod when enter windows. oh well. screw gkx
I wish there was a program that could just run loops of latency checking. Currently, I'm finding myself tweaking timings and subtimings with memset and constant iterations of mbench, hitting ctrl+c after the latency reading. I figure once I get latency squared away as best as possible, I should be in a good bandwidth position as well. It's such a damn shame that I've got 4 firestix1000 in my possession to test with that are all way under expectations, even in single stick testing. guess these must be 7xxxx d9's.