|
doodah10
Regular
Posts: 31
Join Date: Jan, 2002
|
 |
« on: January 03, 2002, 11:31:26 PM » |
|
Memory tweaks
|
|
|
|
|
Logged
|
|
|
|
|
doodah10
Regular
Posts: 31
Join Date: Jan, 2002
|
|
« Reply #1 on: January 03, 2002, 11:31:26 PM » |
|
I have a new Soyo Dragon+ with an XP 1700 CPU, overclocked using 13x multiplier and 137Mhz FSB settings. This adds up to 1781Mhz, however at post it reads XP 1656Mhz. Which is correct? The machine became unstable at 139Mhz and +.5 volts so I backed off to 137 and +.25 volts and I feel comfortable here. Can I go faster with +.75 or more volts and when does more volts become unsafe?
Anyway my main question is, where can I find information on what the BIOS memory settings can do for me? I have OCZ 256Mb PC 2700 memory installed and it seems to be very fast. SiSoft Sandra reports memory speed of:
Int.Buffered 2109
Float Buffered 2000
I'm thinking this memory can go faster but I'm not sure which way to go. So far changes I've made are, CAS 2, Bank Interleave 4, DRAM Command rate 1T. System performance is set normal, because the machine would not boot under turbo, no matter what settings I had. There are many settings I haven't changed because I don't know what they do.
|
|
|
|
|
Logged
|
|
|
|
|
GeorgeBreese
Ace
Posts: 2,088
Join Date: Dec, 2001
|
|
« Reply #2 on: January 04, 2002, 12:58:06 PM » |
|
WCPUID and MBM509 will display CPU speed correctly. Is this an unlocked CPU (L1 bridges closed)? If so, could one bridge be broken?
Look out for CMD=1T timing. Some boards will run okay, but will corrupt data when CMD=1T is set. 3DMark2001 can report corrupted files, or CD burning with Nero and file-compare turned on will report verify failures. It depends on the motherboard (sometimes the revision of the board too) and the quality of the RAM.
|
|
|
|
|
Logged
|
|
|
|
|
PersianImmortal
Regular
Posts: 138
Join Date: Dec, 2001
|
|
« Reply #3 on: January 04, 2002, 08:32:21 PM » |
|
Yep, the Command Rate = 1T is incredibly difficult to run stable on a system. I tried every other memory timing I could to see if 1T would work, and I have Kingmax PC2400 DDR. It just wouldn't work. The main symptons are corrupted file errors in 3DMark2001 and random crashes in 3D Apps. It makes a big difference in performance, but it's highly unstable.
Set to 2T, and with your memory you should be able to get back some of the performance with more aggressive timings, Burst Length of 4, etc.
|
|
|
|
|
Logged
|
|
|
|
|
doodah10
Regular
Posts: 31
Join Date: Jan, 2002
|
|
« Reply #4 on: January 04, 2002, 09:13:37 PM » |
|
This is an unlocked XP 1700, L1 bridges are closed. I suppose one could be open but I doubt it. I thought this wouldn't work if any were open.
Well I set the DRAM Command back to 2T and Int. Buffered dropped to 2078, Float to 1959. Approx 1 & 2% loss. Didn't find burst length setting, could it be called something else?
I only used the 1T setting for 2 days but everything seemed quite stable using that setting.
|
|
|
|
|
Logged
|
|
|
|
|
doodah10
Regular
Posts: 31
Join Date: Jan, 2002
|
|
« Reply #5 on: January 04, 2002, 11:08:28 PM » |
|
I just ran Sandra burn-in wizard for 5 cycles of CPU, multimedia, and memory. The machine shutdown on the 3rd round.
So I returned to 133Mhz and default volts, but still at 13X. Five runs went smoothly.
Then I changed DRAM Command rate back to 1T, ran the burn-in memory test 15 times, and made it through.
Also got a copy of WCPUID. System dies immeadiately upon launch, even at the slower settings. Maybe a bad copy?
|
|
|
|
|
Logged
|
|
|
|
|
GeorgeBreese
Ace
Posts: 2,088
Join Date: Dec, 2001
|
|
« Reply #6 on: January 04, 2002, 11:32:14 PM » |
|
If one L1 bridge is open then you won't get the correct multiplier. Each L1 pin provides one binary bit of the multiplier number, so depending on which pin is open, you might be off by 0.5, 1, 2, 4, or 8.
The burst length value is "new" for the KT266A, and an old BIOS based on the KT266 might not have it. It is related to the abilities of your RAM. Micron's latest RAM can handle 2, 4, or 8 for example. I'm still learning more about it.
Some boards can really handle 1T without a problem. Just be aware that there is a possible problem with it, and be ready to set it to 2T if you discover a problem later. The chance of a problem increases with the number of DIMMs and the quality of each DIMM.
|
|
|
|
|
Logged
|
|
|
|
|
doodah10
Regular
Posts: 31
Join Date: Jan, 2002
|
|
« Reply #7 on: January 04, 2002, 11:42:36 PM » |
|
The copy of WCPUID I had was bad. I downloaded another and it worked correctly. It identified my CPU as XP 1900 and the multiplier at 12X, even though it's set at 13X. So does this sound like I have an open bridge? It's only 1X off, sounds like you might know which bridge is open, George. BTW, I opted to use rear window defogger fix kit, rather than the conductive paint Tom's used to close the bridges. Do you think that was a mistake?
|
|
|
|
|
Logged
|
|
|
|
|
GeorgeBreese
Ace
Posts: 2,088
Join Date: Dec, 2001
|
|
« Reply #8 on: January 05, 2002, 11:47:55 PM » |
|
It would take a magnifying glass and an ohmmeter to examine the situation. I am trying to remember which bridge is 1x. I want to say the 2nd one from the left but I'm not sure. Somebody documented it when the Athlon XP was new.
|
|
|
|
|
Logged
|
|
|
|
|
