Pentium D805.

Donut

As some of you may know, my wife's XP rig seems to be dying a slow agonizing death. So I decided it's time to upgrade.

I'm on a budget so I tried to reuse as much as possible. ie ram and rockin' GeForce2 vid card. Mind you I did follow Leo's thread on ocing his D820 and it's potential as a folder.

Here's what I ended up with,

Pentium D805- dual-core, 2.66ghz. To bad it's 20*133 and seems to be multi-locked

Asus p5p800 se- AGP graphics and able to use standard DDR ram
Epower 450w psu- it advertises 28a on the 12v rail, my dual proc rigs
recommend a strong single rail,we'll see if this one pans out
TT Big Typhoon cooler- I want this rig cool!

total so far is about $325 (all from Newegg)

Have, Kingmax pc3200-2-512mb sticks. (It was cheap when I bought it.)
GeForce2 MX-400 Wife doesn't do anything that would tax even onboard graphics.

Nut's outa smokes, gotta go make some!

Donut

I'm back,

I got everything put together last night, no prob. Loaded Windows let it prime for about 6 hrs. Aok.

Here's where the fun/frustration (however you look at it) begins.

Multi is locked at 20 , all I can see is playing with the voltages and fsb. Stock fsb is 133 (*4, it's Intel). I managed to get into Windows up to 180 fsb, but would not prime. Bumped the voltage, and I mean a bump .0125v per step, still no go.:-/

Back down to 175 fsb, no prime @ stock volts. bump,bump bump, up to 1.35v
still no prime, but watching temps climb into the high 50's low 60's. What seems to be discouraging is it is the same core that keeps failing

Is this normal for dual cores to be so different? Maybe I just need more volts?

Maybe some of you gurus can point me in the right direction. What is the MAX vcore for this chip, and the max temp I should be keeping it under? I'm still of the XP mindset of keep it under 50c.

I would like to push this chip some, if it turns out I only get a modest oc that's fine also. Wife doesn't need much processing power, but folding sure could.:bigggrin:

I've got to take a break for a while-believe it or not I have to try to pull a junk car off the front lawn

Thanks
Donut

lemonlime

Hey Donut, those 805s seem to be a great value to say the least.. If you are at about 170*20, thats a pretty significant OC as it is.. could be that you are already reaching the limits of the chip. Move voltage may help, but that will likely just increase temps more, worsening the situation. If you can get ~3400MHz stable, thats quite impressive.

It is pretty normal for one core to be weaker than the other. The same usually holds true for the AMD dual cores as well. Unfortunatly, you can not overvolt just one core (that would be cool wouldn't it), so its hard to help.
Lets see some benchies!

Donut

Latest update,

I tried 20*175. I had to knock the vcore up. Intel lists this proc between 1.25-1.4v, so I thought I'd see if it would do. At 1.4v (failed prime @1.375) cpuz was recording the voltages bouncing between 1.4-1.25. Asus probe temps were up to 67c! Northbridge was hot, and I think I burned my thumb on the mosfets (they're not sinked). It did manage to prime for 20 min. before I shut it down. Previously it would only go for about 5 min before failing. I think I'll go back to the stock vcore and see what I can do.

lemonlime

Donut wrote:

Latest update,

I tried 20*175. I had to knock the vcore up. Intel lists this proc between 1.25-1.4v, so I thought I'd see if it would do. At 1.4v (failed prime @1.375) cpuz was recording the voltages bouncing between 1.4-1.25. Asus probe temps were up to 67c! Northbridge was hot, and I think I burned my thumb on the mosfets (they're not sinked). It did manage to prime for 20 min. before I shut it down. Previously it would only go for about 5 min before failing. I think I'll go back to the stock vcore and see what I can do.

Wow those temps are crazy high. Is that with the big typhoon or the stock cooler? If u can keep the temps under 60'C (high fifties or lower) you'll be in much better shape.

muddocktor

Hey Donut, sounds like you've been making some progress with the 805 system, but temps looks like they are holding you back somewhat. I've found that my Big Typhoon needs a little more airflow than the stock fan gives myself. I'm using a 120 X 38mm Panaflo M1A fan on mine, which gives 87 cfm at around 35dB noise, so it still is pretty quiet. I also saw a significant drop in temps (5-8° C) after several heat cycles as the AS5 I used bedded in good too. The BT/Panaflo combo is tall as hell though and I actually had to bend the crossbar on the Antec SX600 series case I'm using it in just a little for clearance. A higher cfm fan on the BT should also move more air across the nb heatsink too and help cool that down a bit. If you want to heatsink your mosfets, give me a PM with your address and I can send you what you need for a reasonable price (my cost + shipping), since I have some extra aluminum ramsinks that can be easily modded to mosfet use.

I looked at this site here, but they don't show the 805's specs yet. They do show the rest of the Smithfield's specs and they show a Vmax of 1.55 volts, so you should be safe with 1.4-1.5v vcore as long as you can keep the temps in check. BTW, for temps on the Pentium D, I would try to keep them no higher than 60-65° C loaded, which is pushing right at Tmax.

Besides temps on the nb and mosfets, how is that P5P800 SE working out? Any problems setting it up or bugs in the bios?

Leonardo

Wow, a lot to comment on and a lot of questions:

20X175 is a good overclock, almost 900MHz above stock. I assume you are overclocking via settings in the BIOS, and not any overclocking software, right?

CPU Cooling:
67*C is way too high if you are using the Big Typhoon. I’m maxing at around 56*C with overclocks at 800MHz and higher above default CPU frequency. I’m thinking your Typhoon is not mounted quite flush with the CPU or that you’ve got too much thermal paste, or that the thermal paste has air bubbles in it. Put about half a BB-size spot in the middle of the CPU heatspreader, set the heatsink down and rotate clockwise about 5* and then counterclockwise once each direction. (Bubbles/air in the paste will interfere with heat transfer.) Yes, don’t spread the thermal paste – let the pressure of the heatsink do it (assuming Arctic Silver 5). If you remove the heatsink after this, clean the paste off and reapply. (You are very unlikely to get a stable overclock above 58 or 59*C on the Smithfield cores. My system one can be stable at 800MHz OC at 62*C, but it's an exception. I found this out from a poor Typhoon installation that I subsequently corrected.)

Also, how well ventilated is the case? If only hot/warm air is getting to the heatsink’s fan, cooling won’t be very effective.

BIOS:
Have you locked the DRAM frequency? We need to isolate FSB raises from the DRAM. I don’t know that motherboard, but the BIOS should have an entry to lock the DRAM frequency. Lock it at stock. You can play with DRAM freqs after you determine FSB limits.

Look for PCI Clock Synchronization Mode (or similar term) – lock at 33.33MHz, board default. We want to isolate the CPU FSB from the PCI clock

Voltages:
What are you measuring the vCore with? Asus probe? What you set and what you get can be two very different voltages. Rely on the Asus Probe’s readings. You will run out of thermal headroom before you risk overvolting the CPU. I’ve got my D820s all set at about 1.425, with Probe readings under load and OC at about 1.35-1.37v.

Back to the BIOS. Boost your FSB voltage (if your mobo has that setting option) to 1.4v. This should help stabilize the CPU at higher frequencies.

You think the two cores are operating somehow differently from each other? They shouldn’t be. How did you draw that conclusion?

At 1.4v (failed prime @1.375) cpuz was recording the voltages bouncing between 1.4-1.25.

When the machine first boots into Windows, the vCore should be pretty close to what you have set in the BIOS. When a heavy load is applied, such as F@H, one instance for each core, expect the vCore to drop almost immediately when the load begins. Even on my high-end Asus P5WD2 boards, the vCore measured by Probe will drop from about 1.42 to 1.35/6. But after it drops, it is very steady at the lower level. If you have big fluctuations in vCore under a steady CPU load, it’s either a PSU quality problem or a motherboard voltage regulation problem. Vcore may droop, and it will even with the best unmodified boards, but it should remain steady if loads are steady.

AND LAST:

Northbridge was hot, and I think I burned my thumb on the mosfets (they're not sinked).

You are treading dangerous ground here! If those mosfets and the northbridge heatsink remain “hot”, you may lose your motherboard. I destroyed an Asus P5PL2 under high overclocking – mosfets overheated and burned out. The mosfets even scorched the motherboard PCP. Have you applied Arctic Silver to the northbridge chip? You could also install a fan on it. I recommend lowering your clock or installing heatsinks on the mosfets.

Perhaps consider another board with factory installed mosfet sinks. Most Asus motherboards intended for overclocking come stock with mosfet sinks. Motherboard manufacturers make a number of motherboards that will overclock, but that were not necessarily engineered to do so. My P5PL2 was a prime example.

(In a couple weeks or less, I'll have an Asus P5LD2 and MSI 945P Neo-F available. Both these boards are stock with mosfet heatsinks and both are decent overcockers. But then, you probably wouldn't be interested because both of these boards require DDR2.)

Donut

I dropped the vcore back to 1.3v and the fsb to 166, it seems to be priming stable. What I don't like is cpuz's reading on the vcore, it seems to be bouncing between 1.2-1.3v. I feel it's a board problem, just a feeling. (had an nf-7 that if you saw the drop in cpuz it would fail prime right then)

I might put a larger fan to the heatsink, (Lemonlime, those temps were with the BT) as far as sinking the mosfets, nah. If it were my rig I'd make something up, but it's going to my wife, it has to be a little more "maintainence free" than one of mine. Heck, @ stock it was a large upgrade for her 2400xp.

muddocktor

Yeah, he is going with a value board that will let him keep his old components, Leo. But he can always add the sinks on the mosfets to control the heat. The sinks I have work well on mosfets after cutting them down a bit with a hacksaw to reduce their size to fit in the space on the mobo around the mosfets. And I was able to double up a couple of mosfets with these ramsinks since there were some mosfets next to each other and were flat enough for good contact on both. I did this on my Dual Sata2 board. Since the board uses the 865PE chipset, it should have plenty of options in bios for locking the pci bus and ram settings too.

Leonardo

Muddocktor's offer to help you find some heatsinks for the mosfets is just what the doctor ordered. Seriously, I would lower your OC until you get your mosfets cooled.

Leonardo

dropped the vcore back to 1.3v and the fsb to 166, it seems to be priming stable. What I don't like is cpuz's reading on the vcore, it seems to be bouncing between 1.2-1.3v. I feel it's a board problem, just a feeling.

Just use Asus Probe for monitoring. It's fairly accurate and puts all the voltages and temperatures in one desktop bar. It's much easier.

Donut

Leo,
I agree the northbridge and the mosfets being that hot is not good! On my PC-DL all the thermal "bubblegum" Asus used was replaced, and a small fan installed to try to cool them. If the VRM temps hit 60c I start dropping wu's like crazy

I did lock the pci/ agp bus to 33/66, so I should be all set there I will be changing the TIM on the northbridge and checking spare parts bin for a mosfet sink. It also looks like I'll be reseating the BT also. Right now the board is running on a piece of plywood, not in a case. I guess I wanted to see for myself what a furnace this chip really is.

Asus probe is also showing the vcore to be bouncing around, but it is nice to know it is half-way accurate. I was using throttlewatch when I noticed it was the same core failing prime at higher settings.

Thanks for everybodys help so far,
Donut

Donut

Leonardo wrote:

Muddocktor's offer to help you find some heatsinks for the mosfets is just what the doctor ordered. Seriously, I would lower your OC until you get your mosfets cooled.

Damn, you guys type to fast! As it is priming right now everything is warm to the touch, not blistering hot, and cpu temps are 58c. Still a little hot for my taste, especially since I'm not in a case yet. In the Bios,the vcore is set for 1.31v and Asus probe is reading 1.2v. Later I'll try another PSU, I just want to verify that I'm stable @166 with this setup before I go further.

Thanks all,
Donut

Leonardo

You seem to know what you're doing. Hope you haven't thought that any of our advice and questions were condescending. Was that CPU 67*C with the motherboard out of the case? Double ouch! You defiitely need to reseat the heatsink. Don't fret. I had to install the Typhoon three times in my System 1 before I got it seated right. It's so big and top heavy that seating correctly is not as easy as it would appear. Also, I really dislike the tiny mounting nuts that secures it to the four posts.

Asus probe is also showing the vcore to be bouncing around

Bouncing around by how much?

Donut

Leo

Everybodys advice is welcome! When I built my PC-DL there was an extensive database of what had to be done to get it to OC. I keep refering to that build because it is an Asus, there are quite a few similarities between the boards.
(undervolting of the vcore,heat with the northbridge, heat with the mosfet/vrm)

There doesn't seem to be a whole lot of info on the 805 and this board, most people are either running a single core proc. or they get a newer board for the 805. This seems like a crossover part, to get dual-core cheap. Most of the OCing options are there, but I don't feel the board can really take a heavy OC without some serious work. Someplace I read there is a vdroop mod for this board, Just a little to intense for me.

yes I was up to 67c! I went to kill Prime95 when I noticed it had hit 65c.
2 instances, max heat, 1.4 on the vcore. I think stock is 1.25 vcore. The thing is, prime didn't fail! (It had run for about 20 min. total)

I'm seriously beginning to wonder how accurate the temps are. I've reseated the heatsink twice, and added a 2000rpm fan instead of the 1300rpm fan that was on it, and still it's reading 57c. (1.3v, 2-prime max heat) Hey, the northbridge is at least a little cooler now.

Asus probe is reading 1.2-1.25 on the vcore, seems to have settled down a little. (it was 1.2-1.3v) Tomorrow I'll try a different PSU. I've got a couple Thermaltake 425s in other rigs. They're only rated 18a on the 12v line, but it's a solid 12v. (One is running a dual-opteron rig now)

Time for bed,
Thanks
Tom

Running prime overnight, looks like it's holding a 660mhz oc for now.
BTW Leo, I've read your thread on the 820 more than once today.:bigggrin:

Leonardo

The Smithfield cores are very hot, but they will also run at temperatures that would shut down other cores. The big disadvantages of Smithfield-based CPUs, and I should know as I've got three of them, is their power consumption and the user having to purchase top of the line heatsinks. But for the 805s and 820s, they are a super price-performance proposition. I am really impressed with the 805 - that's just a whole lot of performance for a small price.

muddocktor

Plus, I think you will see your cooling performance get better after your AS5 settles in after a few days and heat cycles. Also, if you pull your BT off again, check and see how flat the base is. I noticed on mine that it was a little bit concave (not bad though) and the surface finish was kind of rough. I lapped mine and after lapping, saw a drop in temps immediately of 1-2° C.

Let me know if you want the ramsinks I have. I can let you have around 6 or so if you need that many (I still have some boards to do myself). They were pretty cheap and I've definitely seen that they transfer heat away from the mosfets on the 2 boards I have done with them. You can use either the thermal tape that's on them or bond them permanently with Arctic Adhesive. I used the tape on them and it seems to work well.

Donut

Good morning,

Prime is still running. Later today I'll try to swap the psu,as Asus probe is showing my 12v rail down around 11.7v. The temps are running 55c-57c. I'm really beginning to wonder though, the room temp is around 65f, down from almost 80f yesterday. (see what happens when you don't pay attention to the wood stove.) but not a big change in chip temp. My other rigs are down 5c under load.

@muddocktor, I checked the flatness of the sink before I installed it, looked good. The finish was a little rough, but nothing that would catch a fingernail. (Not as good as my SLK-900a, but smoother than my Coolermaster hs's). I'll get back to you later today on the ramsinks, as I know I've got a couple old heatsinks in my junk room that might have to be cut up.:D (I have to wait for the sun to come up a little more)

I was wondering, does AS go "bad" over time? I know my tube is over a year old.

muddocktor

AFAIK, AS5 doesn't go bad over time. But I have noticed that it takes 3 days or so to really "break in" and start delivering better temps. Take my Dual Sata 2 rig for instance. I put my BT on it and my temps were running about 47-49 C while running 2 F@H clients when I first got it back running. After around 2 days I noticed that the temps had dropped to the 44-45 C range and after 3 days my temps had dropped to the 42-43 C range, right before I came back to work offshore. I did do a few heat cycles during that period also, shutting it down for around 30-45 minutes, then restarting. Now I haven't seen this kind of behavior with AS5 on a naked core proc like a mobile A64 or an AXP processor and I assume this hapens due to the wide load bearing surface that an IHS gives compared to a naked core.

Donut

I reset the heatsink, still running 57c under load. ( in place FFTs, max heat+power) at idle it's 36c I guess I'll wait a couple of days and see what happens. I'm almost scared to put this thing in a case! ,but I will! I did put sinks on the mosfet, if nothing else for piece of mind.

I also pulled a psu out of a folding rig and tried that to see if voltages would improve. (Thermaltake 420w). NOT! The 12v rail dropped to 11.4v under heavy load, read by Asusprobe. Granted it only has 18a on the rail, but still it put a hurt on this psu. (this is the psu that has been powering my dual Opteron rig!) Talk about a power-hungry pig!;D

Overall I'm happy so far, we'll see if I ahve to go case shopping soon. Let's see how it folds. Wife should be happy.

Donut

I forgot to add, The bios seems incomplete, the ram can be set to 133,166 or auto. Maybe I'm just not used to using an AMI bios.

lemonlime

Wow, I just can't believe how much power these smithfields draw! I thought my opteron was bad. You still at 3400MHz Donut?

Donut

Nope, I dropped it back to 3320 (20*166) I could probibly squeak a little more out of it, but loving wife is starting to growl at her other machine. I think she wants her new rig soon!

Leonardo

I also pulled a psu out of a folding rig and tried that to see if voltages would improve. (Thermaltake 420w). NOT! The 12v rail dropped to 11.4v under heavy load, read by Asusprobe.

Something is just not right. I use 430W Antecs - one dual 12v rail, and the other single, and an Akasa 480W. None of them dips below 11.90 no matter the load or overclock.

Donut

Remember the mosfets being screaming hot? The non SE board did not support dual cores, If I remember right(I think I read it somewhere) the reason was the power draw. It seems they fixed it on this board, barely. I have found a vdroop mod around, though I don't plan on doing this anytime soon.

Also under load my vcore would drop from 1.3v to 1.2v. I guessed this would be a problem that's why I got a psu with 28a on a single rail. On my dual Xeon it was recommended min.30a on the 12v rail. (with the OC I've got on it)
I figured dual proc./dual core the power requirements wouldn't be all that different.

muddocktor

The reason you aren't seeing a 400 ram divider is because of the fsb you are presently running, Donut. It's a limitation of the 865 and 875 series chipsets, where you can't set the ram to a faster speed than the fsb (I'm pretty sure).

I just got back in from the rig a little while ago (job finally ended up) and I found the DC system with the BT off; must have had a power failure and it's the only one I forgot to set the bios to last state on power fail. But I'm seeing the same thing happening with my main machine too, which is a DC Opteron and the AS5 breaking in and giving much better performance. It has an XP90C with a 92mm Panaflo M1BX and when I first put it on, my temps were 48-49° C with 2 clients folding. I right now though, with about a week on the AS5, my temps are running 40-41° C, so I'm definitely seeing a break-in with the AS5 on these 2 AMD dual core machines.

BTW, I'll be going back out to another rig in the morning to finish my 2 week hitch off.

Donut

muddocktor, I do believe your right.

I had completely forgotten I had to insulate a pin on my Xeons for the 200fsb to be available

I have the board installed in a case now, what a pain! I've only got about a 1/4in. clearance between the BT and the psu. Loading winders as we speak.:D

On a side note, Does anybody know how well the airflow is on a Coolermaster Centurion case? (This might be the replacement case if I run into cooling issues)

Donut

Latest update:

Rig is dead. I powered down (getting ready to put it in a new case), got sidetracked. I went to power it back up a couple of hours later so my wife could check her E-mail, Nada-No post. Removed everything, changed everything, still nothing. Can't even force an error beep by removing all ram.

I think it's heat/power related. While removing the backing plate from my heatsink, I noticed the back of the PCB discolored where the different mosfets were (3 places). I e-mailed Asus tech support to see which way I should go next.

It was nice while it lasted,

Leonardo

Yup - dead board. I had the same thing happen to an Asus P5PL2. Asus RMA went through with no problem. Some Asus boards will overclock beyond the safe limit for their construction, that's for sure.

The MSI board I'm shipping you Monday has beefy mosfets heatsinks.

Donut

Leonardo wrote:

The MSI board I'm shipping you Monday has beefy mosfets heatsinks.

Yep, checked them out.:bigggrin:

Of course, with a board fresh from RMA,(Asus) I might have to look for a new processor to go with it.:rolleyes2

Leonardo

Get another Smithfield core. Your electric bill will go up but you can turn the heat off in half of the house!

johansson

Hi,
I´ve been reading this post with interest, since I`ve got simular setup and simular issues.

My system;
ASUS P5P800SE bios 0503
Intel D805 cpu 2,66Ghz@3,6Ghz ( 1.42v)
Zalman 9500LED cpu cooler
4x256 A-Data DDR500
Nvidia 6600GT AGP
Maxtor 160GB Pata
Lite-On DVD
Sony Floppy
HEC 350w ( 18amp / 12v line)

At start I was quite happy since I did run the system OC at 3.6Ghz dual-prime stable with 1.425 vcore.
CPU temp wasn´t good but acceptable ( 34c idle / 60c load).
Tried 3.8Ghz as well but it wasn´t quite prime stable even with 1.5 vcore.
Core0 was running OK, but Core1 went yellow in a couple of minutes.
Looking into details I got a bit surprised to see that vcore fluctuated between 1.28 - 1.48v
That´s 0.2v !!, which would make anyhigher OC very difficult.

At this time I also added some extra Mosfet cooling and replaced Northbridge thermal paste with some AS5,
and replaced cpu cooler and removed some AS5 to a very thin layer.
Then I came up with this genious idea, to set vcore at 1.6v to add up for the MoBo undervolting and fluctation
and somehow achieve a lowest vcore at around 1.4v to keep a stable OC at 3.8Ghz.
When booting at 1.6v I started to smell something was overheating and turned off the system.

Was a bit worried, but no problems booting again and set bios back to 3.6Ghz.
Don´t know what the smell came from, it may have something to do with the new mosfet / NB cooling /and AS5
or it may have something to do with too high vcore.
However, now I wasn´t able to run at 3.6Ghz ( FSB 180) dual-prime stable like before.
I was only able to run primestable at 3.3Ghz ( 166FGSB) at 1.425vcore.
Only thing that got better after this incident was the cpu temp that got from 34c idle to 24c idle and from 60c load to some 36c load !.
Don´t know if something happened to the cpu itself or the fact that I did replace the cpu cooler / cleaned of some AS5 ?.
All these temp-values are measured by speedfan.

Although my PSU isn´t the best, it still delivers stable 12v power +/- 2% with it´s 18 amp on the 12v single line.
I was thinking about getting a better one, but I´m not sure anymore that it would solve any problems.
I´m starting to believe that the P5P800SE powercomponents is´nt made for OC, which in a way would be logical with a cheap combo-mobo.

So I´m now in a position that makes me wonder if I´ve partly damaged MoBo or cpu after my mooded cooling efforts and OC adventures.
Except from that, I´m still dissapointed about the extreme vcore fluctation the P5P800 gives.

Question I ask myself is if I will encounter further instability, would it be the cpu or mobo that would be the subject for replacement.

What MSI Mobo model are you guys talking about ??
Is it a socket 775 with DDR / AGP supported chipset / mobo ?

Thanks.......