Hey everybody,
I’m stress testing my system w/ Memtest86 and Prime95 and have wound up with some results that I need help interpreting.
Here’s my system:
Athlon XP-Mobile 2600 IQYHA 0401MPMW
Biostar M7NCD Ultra, nVidia nForce2 Ultra 400
2 x 512MB GeIL PC3200 value RAM, running dual-channel,
manufacturer rated at 6-3-3-2.5, 2.6 vDIMM
Enlight 420W PSU, +3.3v@40a, +5v@30a, +12v@18a - dual fans
OC: 12.5 x 200 @ 1.7vCore = 2.5GHz
Air-cooled w/ Thermaltake Volcano 10 & Arctic Silver 5
35-40C idle & 55-60C Prime95 load, depending on weather
A word about my case and case cooling: It sucks. Cheap case with one 80mm fan in the back on full jet-engine blast. One low-rpm intake fan in the front. All the tests mentioned below have been run with the CASE OPEN and cool New York fall weather. I understand that if I want my system to be rock-solid during normal operating, I need to get a better case and cooling. But in terms of the following results, they were obtained on very cool nights with case open, so I don’t think cooling is an issue below.
Now, on to Memtest86:
I’ve found that while running at stock timings of 6-3-3-2.5, FSB of 200MHz, & 2.8vDIMM (just for good measure), a loop on Test 5 always yields at least a couple errors within 20 to 40 passes irregardless of processor speed. In other words, lowering the clock multiplier has no effect on the errors. I’ve gathered from reading the forums that I should take these errors seriously, even though Prime95 has run several times for 16-24 hrs. with no errors. So I started experimenting with loosening the timings, and consequently lowering the memory bandwidth. First I tried with really loose timings just to see if I could get rid of the errors. Here’s the results of the Test 5 loops I ran in the order I ran them:
(all tests were run in dual-channel mode @ 2.8vDIMM)
1. 11-6-6-3 @ 200MHz FSB = 1179 MB/s : no errors
2. 11-4-4-2.5 @ 200MHz FSB = 1304 MB/s : errors
3. 11-4-4-3 @ 200MHz FSB = 1322 MB/s : errors
4. 11-5-5-2.5 @ 200MHz FSB = 1271 MB/s : errors
5. 10-5-5-2.5 @ 200MHz FSB = 1239 MB/s : errors
6. 8-5-5-2.5 @ 200MHz FSB = 1223 MB/s : errors
7. 11-3-3-2.5 @ 200MHz FSB = 1398 MB/s : errors
8. 11-3-3-3 @ 200MHz FSB = 1398 MB/s : errors
9. 8-4-4-2.5 @ 200MHz FSB = 1359 MB/s : errors
10. 6-3-3-2.5 @ 166MHz FSB = 1182 MB/s : no errors
So what’s going on here? Why does Test 5 return errors at bandwidths in excess of 1200 MB/s? It doesn’t seem to matter what the timings are, what the FSB is, or what the CPU speed is, so long as the memory bandwidth is below 1200 MB/s.
Here’s a link to a bunch of favorable reviews of this RAM on newegg.com: http://secure.newegg.com/app/CustratingReview.asp?DEPA=1&item=20-144-305
Another possible piece of the puzzle:
With memory at stock timings, I can pass Prime95 for 24 hours with the case open, but can only get through about 5 minutes with the case closed. Although all the Memtest tests noted above were performed with the case open, obviously there’s a component that’s getting too hot when the case is closed. Maybe this is the same component that’s causing my RAM errors? The northbridge getting too hot? I don’t know enough about what the northbridge does to know if this is plausible, or if that could be related to the bandwidth issue. What could be creating a bandwidth threshold? Is it just bad/inferior RAM? It certainly isn’t giving me the performance reported in those newegg reviews. People are talking about successful oc’s of this RAM; nothing extravagant, 210-220 FSB with stock or loosened timings. I can’t even run at stock timings and FSB w/o errors.
Any suggestions/insights will be much appreciated.
Peace
I’m stress testing my system w/ Memtest86 and Prime95 and have wound up with some results that I need help interpreting.
Here’s my system:
Athlon XP-Mobile 2600 IQYHA 0401MPMW
Biostar M7NCD Ultra, nVidia nForce2 Ultra 400
2 x 512MB GeIL PC3200 value RAM, running dual-channel,
manufacturer rated at 6-3-3-2.5, 2.6 vDIMM
Enlight 420W PSU, +3.3v@40a, +5v@30a, +12v@18a - dual fans
OC: 12.5 x 200 @ 1.7vCore = 2.5GHz
Air-cooled w/ Thermaltake Volcano 10 & Arctic Silver 5
35-40C idle & 55-60C Prime95 load, depending on weather
A word about my case and case cooling: It sucks. Cheap case with one 80mm fan in the back on full jet-engine blast. One low-rpm intake fan in the front. All the tests mentioned below have been run with the CASE OPEN and cool New York fall weather. I understand that if I want my system to be rock-solid during normal operating, I need to get a better case and cooling. But in terms of the following results, they were obtained on very cool nights with case open, so I don’t think cooling is an issue below.
Now, on to Memtest86:
I’ve found that while running at stock timings of 6-3-3-2.5, FSB of 200MHz, & 2.8vDIMM (just for good measure), a loop on Test 5 always yields at least a couple errors within 20 to 40 passes irregardless of processor speed. In other words, lowering the clock multiplier has no effect on the errors. I’ve gathered from reading the forums that I should take these errors seriously, even though Prime95 has run several times for 16-24 hrs. with no errors. So I started experimenting with loosening the timings, and consequently lowering the memory bandwidth. First I tried with really loose timings just to see if I could get rid of the errors. Here’s the results of the Test 5 loops I ran in the order I ran them:
(all tests were run in dual-channel mode @ 2.8vDIMM)
1. 11-6-6-3 @ 200MHz FSB = 1179 MB/s : no errors
2. 11-4-4-2.5 @ 200MHz FSB = 1304 MB/s : errors
3. 11-4-4-3 @ 200MHz FSB = 1322 MB/s : errors
4. 11-5-5-2.5 @ 200MHz FSB = 1271 MB/s : errors
5. 10-5-5-2.5 @ 200MHz FSB = 1239 MB/s : errors
6. 8-5-5-2.5 @ 200MHz FSB = 1223 MB/s : errors
7. 11-3-3-2.5 @ 200MHz FSB = 1398 MB/s : errors
8. 11-3-3-3 @ 200MHz FSB = 1398 MB/s : errors
9. 8-4-4-2.5 @ 200MHz FSB = 1359 MB/s : errors
10. 6-3-3-2.5 @ 166MHz FSB = 1182 MB/s : no errors
So what’s going on here? Why does Test 5 return errors at bandwidths in excess of 1200 MB/s? It doesn’t seem to matter what the timings are, what the FSB is, or what the CPU speed is, so long as the memory bandwidth is below 1200 MB/s.
Here’s a link to a bunch of favorable reviews of this RAM on newegg.com: http://secure.newegg.com/app/CustratingReview.asp?DEPA=1&item=20-144-305
Another possible piece of the puzzle:
With memory at stock timings, I can pass Prime95 for 24 hours with the case open, but can only get through about 5 minutes with the case closed. Although all the Memtest tests noted above were performed with the case open, obviously there’s a component that’s getting too hot when the case is closed. Maybe this is the same component that’s causing my RAM errors? The northbridge getting too hot? I don’t know enough about what the northbridge does to know if this is plausible, or if that could be related to the bandwidth issue. What could be creating a bandwidth threshold? Is it just bad/inferior RAM? It certainly isn’t giving me the performance reported in those newegg reviews. People are talking about successful oc’s of this RAM; nothing extravagant, 210-220 FSB with stock or loosened timings. I can’t even run at stock timings and FSB w/o errors.
Any suggestions/insights will be much appreciated.
Peace