View Full Version : DFI 875P-T with bugged 5:4 Ratio?
Hi Guys,
I recently got myself a new Board and RAM (DFI Lanparty 875P-T and OCZ 3200EL Rev 2... upgraded from an AS8 with EB) and I'm pretty pleased with it so far.
The CPU is still the same and ran at 4.6 (300x15) with the old setup.
The 5:4 Ratio is giving me trouble though - i can run up to 278 Mhz 1:1 and rockstable, however, if I set the FSB to 300 Mhz and the Ratio to "DDR320" (5:4), im getting tons and tons of errors in Memtest - the System isn't stable at all.
Still, the CPU will do up to 310 Mhz FSB rockstable with a 3:2 ratio. :confused:
To sum it up:
278 Mhz FSB / 278 MEM -> perfectly stable
300 Mhz FSB / 240 MEM -> extremely unstable
300 Mhz FSB / 200 MEM -> perfectly stable
I´ve also fiddled with timings and PAT, it doesnt change anything at all. :stick:
Any ideas?
megahurtz-oc
11-17-2004, 05:41 PM
i have the same prob guess the dividers are screwed. maybe a bios update will help.
Szymek
11-18-2004, 07:14 AM
Alter PCI freq in BIOS if you can
helsyeah
11-18-2004, 11:48 AM
Hi Guys,
I recently got myself a new Board and RAM (DFI Lanparty 875P-T and OCZ 3200EL Rev 2... upgraded from an AS8 with EB) and I'm pretty pleased with it so far.
The CPU is still the same and ran at 4.6 (300x15) with the old setup.
The 5:4 Ratio is giving me trouble though - i can run up to 278 Mhz 1:1 and rockstable, however, if I set the FSB to 300 Mhz and the Ratio to "DDR320" (5:4), im getting tons and tons of errors in Memtest - the System isn't stable at all.
Still, the CPU will do up to 310 Mhz FSB rockstable with a 3:2 ratio. :confused:
To sum it up:
278 Mhz FSB / 278 MEM -> perfectly stable
300 Mhz FSB / 240 MEM -> extremely unstable
300 Mhz FSB / 200 MEM -> perfectly stable
I´ve also fiddled with timings and PAT, it doesnt change anything at all. :stick:
Any ideas?
It seems as though the 875 and 865 chipsets have semi hosed 5:4 dividers. I own two Is7s and an ic7 max3, all fully voltmodded. Only one of the is7's can do over 290fsb with the 5:4 divider and be stable. The other is7 wont post beyond 286fsb with the 5:4 divider, and the max3 wont post over 282fsb with the 5:4 divider. this is with memory that runs completely stable over 244fsb 1:1.
From my experience and what i have gathered in the last 6 months or so, the memory is stressed more when using the 5:4 divider. Example. With the memory @ 233 mhz with the 5:4 divider i get more bandwidth than with the memory running 233 mhz 1:1.
Also the 5:4 divider seems to stress the NB quite a bit as well. I found that upping the vdd (vnorth) can increase the upper fsb using the 5:4 divider a bit, but not by a large margin.
I have tried increasing the agp/pci frequency as well and have yet to obtain an increase in stability in the 280 to 300fsb range using the 5:4 divider. Increasing the agp/pci frequency does help when going over 300fsb however.
Overall, at this point i believe that with a few exceptions, the majority 875/865 boards are very difficult to get stable when using the 5:4 divider over 280fsb to 290fsb. However, if someone finally cracks this 5:4 problem they would be my hero. My 2.4c would sure like to run @ 3.6 with 300fsb and my memory cookin along with the 5:4 divider @ 240mhz.
OPPAINTER
11-18-2004, 12:03 PM
I had 5:4 problems on this mobo also. It seems the 1:1 and 3:2 work fine, the 5:4 is all FUBAR though.
OPP
Clevor
11-18-2004, 05:23 PM
The DFI board is finicky with ram anyway, and yeah, 5:4 don't work right. My BH-5 only works at 3:2 and 1:1.
You can do high 5:4 if timings are slack. Hard to do high up >290-293 with 2-2-2-5. I believe you need to juice the NB and AGP. I've gone over 300 5:4 OK with ADATA PC4000 stuff at 2.5-4-3-6, various boards like Epox and Abit IS7/AI7/IC7.
The Ballistix PC3200/4000 stuff will also do 290-300, 5:4, 2.5-2-2-5 on the IS7-E (only needs 2.7-2.8 volts too). The PQI TCCD did 297, 5:4, 2.5-3-3-6 on the 4PDA2+ @ 2.8 volts.
All above I'm talking 3D stable settings. I seen Memtest clean higher but you will pop out of the 3DMarks pronto.
holst1202
11-19-2004, 04:27 AM
Hi! Meop
Had you tried to disable the CPC Addr/control ?
Hi,
Yeah, I tried just about every possible combination... PAT on/off, CPC and Selective CPC on/off, "Fast Chip Select" (whatever that is) on/off... nothing seems to work. All the other "Turbo" stuff is off as well.
I've also maxed out the Vagp since that helped my AS8 a lot with PAT, I can't see any difference with the DFI though...
Well, ATM i'm running 280 1:1 @ CL2.5-3-3-6 with 2.6V and the "low-voltage" OCZ 3200EL Rev 2.
I tried 300 5:4 with all sorts of Settings and RAM Sticks (the 3200EL, 3700EB and a random Corsair BH-6 Stick), none of them works flawless at Memtest... not even with ridiculous timings like CL3-4-4-8.
I also tried the PCI/AGP Trick, I've heard of that before... it doesn't help either... also, the next higher setting from 66/33 is 73/37, whch seems a bit high for my finicky PCI Cards (Gigabit, SATA Raid)
Bleh, I already sent a mail off to DFI, no response so far.
I hope that's nothing that a BIOS-Update can't fix.
FUGGER
11-19-2004, 11:06 AM
New bios posted
http://www.dfi.com.tw/Upload/BIOS/LP875PT_DB17.zip
nice - has anyone checked that BIOS yet?
I'll slap it on tomorrow when I'm back at home... I doubt that they fixed the 5:4 issue though... if they are even aware that such a thing exists. :stick:
the "changelog"...
1. To add the "Disabled" option at "CPU THRM-Throttling" setting in BIOS.
2. To expand the high limit of CPU shutdown temperature in BIOS.
althes
11-19-2004, 04:23 PM
I will check it tonite
Do you only have trouble running 5:4 at very high fsb near 300?
Will there be problems running 270 5:4 2-2-2-5?
Right now I'm trying 1:1. 253 2-2-2-5 is prime stable without booster (booster fried the exactly same way fuggers did on the board)
Well I managed 272 or so 5:4 but nothing more.
I will mail DFI about this as it is holding the board back badly.
Well I managed 272 or so 5:4 but nothing more.
I will mail DFI about this as it is holding the board back badly.
Well after switching off fine tune and cpc address control i pushed upto 285 PI stable although on the ragged edge using stock cooling.
It looks to me as if its ram related more than the board.
hmm, what happens if you loosen up the timings a bit and try to run memtest at 300 5:4?
I can't see why the 5:4 ratio should stress the RAM that much... weird.
Sooo, errh,... what is a 5:4 ratio good for if you can run 1:1 at higher speeds with the same timings, and stability? :stick:
on my board 270 3:2 2-2-2-5 didn't work either....
did vagp or vdimm help you get higher?
Yeah - I cranked the Vagp to 1.8 and the Vdimm is at 2.6V ("old" TCCD - this is the sweetspot).
I can run 300 Mhz 3:2 2-2-2-5 all day long with my OCZ Plat Rev 2...
could you tell me your complete settings.
do have sata hdd's?
a pity I can't send the board back anymore...someone should have posted this some days earlier ;)
could it be a problem that I have a P4 560 and use the 14x and you have a "natural" 15x?
errh, sorry for the late reply, i missed the topic.
yeah, the cpu is a "native" 3.0e, thus running at 15x... the 14x doesn't even work for me.
i do have sata drives indeed (2x Raptor 73GB / RAID 0), both are hanging on the onboard controller.
ar7786
11-27-2004, 06:45 AM
To achieve higher than 300fsb, active cooling on NB is necessary, here is what I got with IS7 with swiftech water block on NB, Gskill PC3200LE 2x256MB
http://home.earthlink.net/~xhu7786/sitebuildercontent/sitebuilderpictures/p4m5.jpg
you may guess we had this idee before ;)
it's seems to be a special bug with 875P-T + 5:4.
nvertheless i will test wc nb cooling next week. but i think it's time for tccd ... pqi should be a good choice
erh, yeah. the NB temp hovers at 25°C, it does have a fan strapped on indeed.
looks like the 5:4 ratio (or the 875P-T in general) just really stresses the hell out of the ram.
BTW, I fiddled with the 5:4 some more and managed to run a SuperPI test at 292 Mhz / 5:4 / CL2-2-2-5 with OCZ Rev 2 TCCD (at 3.2V Vdimm though).
Also, I noticed the i865/i875 Tweaker cpuid.org shows different latency readings for each ratio... I can't remember exactly, but afaik it was 01 for 1:1, 11 for 5:4 and 00 for 3:2.
The old Abit AS8 showed the same settings for each ratio... I found that pretty interesting.
I've been told that the 11 is the "tightest" settings of all, altering it didn't yield any results (neither FSB nor Bandwidth-wise) though.
Oh well, guess DFI really hosed up those DRAM Timings on the 5:4 ratio... nothing we can do about it.
erh, yeah. the NB temp hovers at 25°C, it does have a fan strapped on indeed.
looks like the 5:4 ratio (or the 875P-T in general) just really stresses the hell out of the ram.
BTW, I fiddled with the 5:4 some more and managed to run a SuperPI test at 292 Mhz / 5:4 / CL2-2-2-5 with OCZ Rev 2 TCCD (at 3.2V Vdimm though).
Also, I noticed the i865/i875 Tweaker cpuid.org shows different latency readings for each ratio... I can't remember exactly, but afaik it was 01 for 1:1, 11 for 5:4 and 00 for 3:2.
The old Abit AS8 showed the same settings for each ratio... I found that pretty interesting.
I've been told that the 11 is the "tightest" settings of all, altering it didn't yield any results (neither FSB nor Bandwidth-wise) though.
Oh well, guess DFI really hosed up those DRAM Timings on the 5:4 ratio... nothing we can do about it.
I have mentioned this to DFI directly, with luck they will have a look at the bios for us.
thanks - that would be awesome. I doub't that it's gonna happen, though.
As far as the 875P-T and OCZ ram, did you try different combos yet?
I'm wondering if the new voltage-hungry ram would do any better than my TCCD (3200 plat rev 2, the "old 2.6V sweetspot" version)... humm.
yeah that would be nice ;)
freecableguy
02-12-2005, 04:07 PM
Is there a Vdimm mod for this board?
freecableguy
02-12-2005, 09:25 PM
Also, how the fark do you enable PAT on this board?
uclajd
02-12-2005, 10:08 PM
Also, how the fark do you enable PAT on this board?Isn't it already enabled, being an 875? 865s have the PAT to get around their disability, but 875s should already have it.
FWIW, I can't get the 3:2 to work worth a damn. Why would I want to? 14 multi on a couple of benches that won't go past 4.75...
5:4 seems lame too.
Fortunately I can do 1:1 past 284 or so...