If the DRAM is set to 667, the CPU FSB to DRAM frequency ratio is 3:5. If the DRAM is set to 533, the FSB-DRAM ratio is 3:4.
Do you also have a 1:2 divider? Put the ram at as low speed as possible with a divider and i would think that the 1:2 divider would be the lowest setting. Or can't you just change the divider without setting the Dram frequency?
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
There are only two CPU/DRAM ratios available - 3:4 and 3:5. Both ratios are automatic when the RAM is set to 533 and 667 MHz respectively. This board is definitely not an Asus or Abit.
Download that and see if it works on your board. You might be lucky and get some more ratios there.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
Mack, I could not find the ICH controller number in my MSI motherboard specs that matches the ICH numbers in the Clock Generator. Maybe I don't know how to use this thing. I'll pop over to CPUID and read.
I checked the CPUID site, ClockGen page. the I945P/G chipset is not supported yet under the MSI category or the chipsets category. Thanks anyway. I'll watch and see if it is supported later.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
I know you mentioned using FAH and general PC use for stability testing.. any reason you have not used Prime95 or OCCT for stability testing runs? I only mention it because my systems can run FAH for a long time (sometimes weeks) without issue, when they fail prime95 in 10-15 minutes. You may start to mysteriously loose some WUs in the long-run.. Just a thought..
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
You make a very good point. Actually, in 2004 I "mysteriously" lost a lot of production on my System No. 1, which previously contained the P4C 2.8 Northwood that is now in System No. 3. I learned my lesson with that ongoing episode and now monitor all my systems very, very closely. And yes, you are correct - Prime95 will stress the system in ways that Folding@Home does not.
if all you'll be doing is stress testing I would recommend sp2004 instead of prime ...it has a gui and it basically runs the prime code. The best part is that if you want to run 2 instances you just click the icon again to open another instance and set the cpu to 0 and in the other set it to 1. I ran mine last night for 6+ hours both cores @ 2600 and everything was fine ...prior to that I had run it for 5 mins both cores @ 2610 and cpu1 failed. I'll probably give 2610 another shot this weekend.
Btw ...you can test the cpu or the ram or blend both in a test.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
All a moot point at this juncture. Overclocking the D820 has come to an abrupt halt. Operation of the whole computer has come to a complete stop. My wife pulled out my M1911 and put a .45 fat lead one right through the power supply.
Well, no, she didn't. I killed the motherboard. I was installing a large passive heatsink on the northbridge to replace the stock one. I was over confident, having done that procedure too many times, and cracked the NB's core. CMOS resets, CMOS battery removal...nothing will bring back video. Tested with a spare, good video card too. Uh, yeah, also smelled that sickeningly pungent smell of burning electronics when the computer would not boot. I did confirm the core chip when I disassembled the computer in detail. Looks like I'm shopping for a new motherboard. Probably have to order one. CompUSA, the only affordable local shop, doesn't have squat. They've got the Asus P5ND2, but it's not compatible with D820s, only 830 and up. CompUSA's site doesn't reveal that, but web research points out that most NF4 Intel SLI boards weren't engineered/aren't compatible for the D820.
Shipping still might be an issue for Leo, since he lives in the Great White North, but eBay is definitely an alternative. It will most probably have to be an Intel chipset to work with his 820.
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
Mudd is right. There are compatibility problems with many of the Intel-Nforce 4 boards with early series D820 CPUs. Mine is definitely 'early', being an engineering sample. Can't get earlier than that. The Intel 945P/G/~ and 955 chipsets work with all Pentium Ds. I'll probably end up ordering, as CompUSA's offerings are pretty slim.
What about this board ...I don't know much about Intel chipsets but this board looks good to me ...probably out of the price range ballpark but what's your opinion on this board compat and performance wise?
Those are supposed to be some damn fine boards, from what I've read. Supposed to very good overclockers too.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
Well, gentleman, the dual Presshot is back in business warming my home office! Cheery, isn't it!
CompUSA had a board that was not listed on the web site and that the useless information telephone information center did not list. I purchased an Asus P5PL2: DDR2, PCIe, PCI, SATA II, overclocking BIOS, Windows overclocking software.
It's running stable now with all updates. It took me a couple hours to get the settings right in the BIOS. It's an American Megatrends BIOS, of which I was not familiar previous to this. Overclocking will commence later this evening.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
After many fits, starts, stops, and mild frustration, I finally figured out the BIOS. It really is much differnt than the Award BIOS setups that I was accustomed to. Actually, this board has a load of overclocking settings available, to include FSB voltage, south bridge voltage, and lots of DRAM frequency and timings options. I uninstalled the Windows-based overclocking utility and am now focusing on the BIOS. 846MHz overclock per core. I think there's more FSB headroom yet!
Very nice!! Now just to keep the wifey away from the pistol...
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
Still stable. I just bumped the vCore setting down to 1.45v from where I had it at 1.5. Default is 1.34. Tomorrow I'll see what headroom there and experiment with lower voltages on memory and CPU. Memory voltage is bumped from the default of 1.8 to 1.9. Overclock on the memory is is 14MHz only. It's 667 DDR2 (333MHz X 2) running now at 347. I can set it in the BIOS as 533, which will allow max FSB overclocking while keeping the RAM closer to spec. Good ram though, considering it's a value brand, Gigaram.
53c load isn't bad at all on quiet air with those clocks. Not far from a Amd X2 at 2750. I get around 52c load after 24 hours of dual Prime with that heatsink. I removed the heatspreader on the cpu and gained 8c but no overclocking.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
Thanks for coming along for the ride, guys. That's what makes the forums fun. Some more experimenting: upped voltages, reduced voltages, changed FSB/DRAM dividers..... Booting into windows at idle is a piece of cake at 3800MHz, but it's not stable when I start piling on the load. The system freezes, but does not shut down. This requires a reboot. The Asus motherboard has a feature that under excess overclocking the BIOS will reset itself to default settings. I haven't had that happen, but am pleased to report that it is indeed a tough, robust BIOS. I have not once had to reset the BIOS with the CMOS jumper. Having backed the memory down to default settings, SPD 333MHz at 4-4-4-11, and being able to lock the PCI/e bus at 33MHz, I think I've explored the FSB limits quite well. It's stable now with about whatever I can throw at it:
I had to break one of my own rules and install a fan on the Asus' passive northbridge heatsink. Both the MSI's and the Asus NB cooler were simply inadequate for more than moderate overclocking. They were both hot to the touch, not warm. My parts bin yielded a little 40mm fan that I cannot hear. (Amazing how many worthless Abit NB fans I found in the parts bin! LOL)
This is a fairly simple motherboard - no RAID, no AGP, only one IDE, passive cooling. It would appear though that the quality is good. Not too shabby on the OC - over 1690MHz (2 cores) for free. I'm pleased.
Looks like you are really having some fun with that 820, Leo. And it's turned out to be a good overclocking chip after all, too. One thing I was reading about over at the Intel cpu forum at overclockers.com is that you are probably better off with that old single 12v rail Robanton psu, amp-wise, than many of the newer split 12v rail high dollar psu's are with overclocking a DC Intel beast. Many people are finding that these Smithfield (and it's also looking like the early Pressler results too) draw so much current on the 12v rail that it's maxing out 1 of the 12v rails on these newer psu's (such as the AntecTP2 series among others) and causing instability when really pushing that DC Intel. With your single 12v rail, you have a higher max allowable 12v draw on it, so the proc isn't starving for power. And since this is a basic folding box, you aren't pulling a lot of other 12v load too, so you are staying within the psu's operating parameters.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
Thanks for the advice, Mudd. I've heard a little about that before, but you've given the first details I've seen. I just purchased a Tp II EPS12v Antec from Ebay, but the auction winning price was so low I don't mind just putting the PSU in my emergency parts queue in the parts bin. As I mentioned in a previous post, mulitmeter readings showed this Robanton to be pretty healthy. I have not checked the 12v line with a multimeter since the motherboard change, but Asus Probe is showing 11.88 - 11.93 and SpeedFan is showing approximately the same. (I can't get Motherboard Monitor to work correctly with this board.)
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
Cool ...it's like this thread has gained it's second wind!
Third wind has commenced. 3739MHz, steady under load.
Today I replaced the two low speed Antec 120mm fans (not three-speed TriCool) with Panaflo L1As I had in the parts bin. The Big Typhoon heatsink so big that I couldn't fit the Panaflo in the back interior of the case, which is a Chieftec (Antec 1030B). I had to modify the back of the case, cutting the exhaust fan hole wider with a jigsaw. It works quite well. Load CPU temperature is now reduced 4*C, which has bought me a little more front side bus. Here's where we are now:
I've again booted into Windows at higher freq's. Temperatures seemed OK, but after a while the screen would freeze. Who knows. I'll boost voltages and try again tomorrow.
Tomorrow I'll boost the vCore a little more and also pump up the vFSB and northbridge voltage. We'll see.
Right now I've got the FSB/DRAM ratio at 5:4 - just playing it safe for now. I think the DRAM will handle 1:1. I may try that tomorrow as well.
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2006
Looks like I'm going to have to stay at 3646MHz. Even after lowering case and CPU core temperatures, and regardless of memory timings and frequency, and boosts to vMem, vNorthbridge, and vCore, after a few hours of running the system will freeze. As has been advised me here and at another forum, the freeze is probably the limit of the CPU itself (FSB wall) or due to vdroop with the motherboard or PSU. I'm happy - it's still an 846MHz overclock and the box is a tremendous Folding machine. :smiles: Mission accomplished.
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited February 2006
Gentlemen, be prepared for:
D820 OVERCLOCKING, PART III
Within a week, I will be overclocking my 820 system with an Asus P5WD2, supposedly the big daddy of Socket 775 overclockers. I hadn't intended on this, but my P5PL2 killed itself a couple nights ago. Seems the board's components were not meant for the overachieving BIOS. As far as I can tell, the mosfets way overheated and either damaged traces in the PCB or the mosfets themselves were rendered non-functional. Presently, I'm attempting to contact Asus online customer service for RMA.
The P5WD2 is a refurbished unit purchased for steep discount from Newegg. I'm crossing my fingers that what caused it to be returned in the first place was user error by the original owner, and not a problem with the board. We'll see, I should have the board next week Monday or Tuesday.
Stay posted to this thread. I'm hoping for a 1000MHz overclock per core.
Comments
Do you also have a 1:2 divider? Put the ram at as low speed as possible with a divider and i would think that the 1:2 divider would be the lowest setting. Or can't you just change the divider without setting the Dram frequency?
Download that and see if it works on your board. You might be lucky and get some more ratios there.
I checked the CPUID site, ClockGen page. the I945P/G chipset is not supported yet under the MSI category or the chipsets category. Thanks anyway. I'll watch and see if it is supported later.
3640MHz, 14X260FSB/1040, DDR2 346.6MHz/channel
Settings:
DDR2: 533MHz mode, timings 5-5-5-13 (SPD 4-4-4-11).
vCore: 1.52 set/1.40 indicated (default 1.3)
vMem: 1.95 (default 1.8)
vPCI/e: 1.65
CPU Core Temp: 52*C
System Temp: 30*C
Couldn't get 260FSB stable last night. Looks like the PCI/e boost to 1.65 (over 1.55 default) may have done the trick. Still locking up above FSB 260.
I know you mentioned using FAH and general PC use for stability testing.. any reason you have not used Prime95 or OCCT for stability testing runs? I only mention it because my systems can run FAH for a long time (sometimes weeks) without issue, when they fail prime95 in 10-15 minutes. You may start to mysteriously loose some WUs in the long-run.. Just a thought..
Btw ...you can test the cpu or the ram or blend both in a test.
Well, no, she didn't. I killed the motherboard.
what are the qualifying motherboards? I mean I know nothing about this chip so how can I help search? What am I looking for????
EDIT: The Egg has another refurb MSI 945P Neo-F right here, Leo: http://www.newegg.com/Product/Product.asp?Item=N82E16813130516R
Yeah newegg refurb sounds like a great idea!
What about this board ...I don't know much about Intel chipsets but this board looks good to me ...probably out of the price range ballpark but what's your opinion on this board compat and performance wise?
http://www.newegg.com/Product/Product.asp?Item=N82E16813131547R
CompUSA had a board that was not listed on the web site and that the useless information telephone information center did not list. I purchased an Asus P5PL2: DDR2, PCIe, PCI, SATA II, overclocking BIOS, Windows overclocking software.
It's running stable now with all updates. It took me a couple hours to get the settings right in the BIOS. It's an American Megatrends BIOS, of which I was not familiar previous to this. Overclocking will commence later this evening.
Very nice overclock, that is one serious folding monster. That ES chip really turned out to be a real bargain :thumbup
53c load isn't bad at all on quiet air with those clocks. Not far from a Amd X2 at 2750. I get around 52c load after 24 hours of dual Prime with that heatsink. I removed the heatspreader on the cpu and gained 8c but no overclocking.
CPU: 3646MHz (FSB 260X14)
FSB 1040: (quad 260)
DRAM: (DDR2) 347MHz, 5-4-4-15
vCore set/indicated: 1.475/1.36
vMem: 1.9
I had to break one of my own rules and install a fan on the Asus' passive northbridge heatsink. Both the MSI's and the Asus NB cooler were simply inadequate for more than moderate overclocking. They were both hot to the touch, not warm. My parts bin yielded a little 40mm fan that I cannot hear. (Amazing how many worthless Abit NB fans I found in the parts bin! LOL)
This is a fairly simple motherboard - no RAID, no AGP, only one IDE, passive cooling. It would appear though that the quality is good. Not too shabby on the OC - over 1690MHz (2 cores) for free. I'm pleased.
Today I replaced the two low speed Antec 120mm fans (not three-speed TriCool) with Panaflo L1As I had in the parts bin. The Big Typhoon heatsink so big that I couldn't fit the Panaflo in the back interior of the case, which is a Chieftec (Antec 1030B). I had to modify the back of the case, cutting the exhaust fan hole wider with a jigsaw. It works quite well. Load CPU temperature is now reduced 4*C, which has bought me a little more front side bus. Here's where we are now:
CPU: 3739MHz (FSB 267X14)
FSB 1068: (quad 267)
DRAM: (DDR2) 334MHz, 5-4-4-15
vCore set/indicated: 1.512/1.38-1.40
vMem: 1.9
vFSB (external): default
vNorthbridge: default
vSouthbridge: default
I've again booted into Windows at higher freq's. Temperatures seemed OK, but after a while the screen would freeze. Who knows. I'll boost voltages and try again tomorrow.
Tomorrow I'll boost the vCore a little more and also pump up the vFSB and northbridge voltage. We'll see.
Right now I've got the FSB/DRAM ratio at 5:4 - just playing it safe for now. I think the DRAM will handle 1:1. I may try that tomorrow as well.
Hmm wonder what they fold like at that speed.
Within a week, I will be overclocking my 820 system with an Asus P5WD2, supposedly the big daddy of Socket 775 overclockers. I hadn't intended on this, but my P5PL2 killed itself a couple nights ago. Seems the board's components were not meant for the overachieving BIOS. As far as I can tell, the mosfets way overheated and either damaged traces in the PCB or the mosfets themselves were rendered non-functional. Presently, I'm attempting to contact Asus online customer service for RMA.
Here's a thread on the demise of the P5PL2.
The P5WD2 is a refurbished unit purchased for steep discount from Newegg. I'm crossing my fingers that what caused it to be returned in the first place was user error by the original owner, and not a problem with the board. We'll see, I should have the board next week Monday or Tuesday.
Stay posted to this thread. I'm hoping for a 1000MHz overclock per core.