O/Cing the LGA775 3.2 & Abit AG8

[batboy]

Got my new system togather last night (see green rig in signature). Here are my first impressions.

1) CPU looks cool with no pins. After hearing about how fragile the LGA775 socket was, I was being very careful about installing the CPU. There are a couple of tiny notches that you line up so the CPU will only go in one way. When I tried closing the hold-down cover, it didn't seem to go all the way down. I took it all back out and looked at it again. Seemed ok. I tried again. Same deal. I pushed the lever down with a little more force. There was a audible snap when it locked into place. That scared me, but apparently it was supposed to do that.

2) The CPU cooler mounting holes on i915/i925 mobos are completely different than the socket 478 mobos. Meaning you need a heatsink designed for the LGA775 socket. The Thermaltake Jungle heatsink I bought is of a similar design to the Intel retail heatsink. Basically it mounts onto the mobo the same way. There are 4 legs with a type of push pin clip. I found out the hard way that you push the pin down until it clicks and then you turn the head of the pin clockwise to lock it into place. Apparently I didn't lock two of the pins and accidently tested out the thermal throttling of the CPU and high temp shutdown feature of the mobo. After I reseated the heatsink, all was well.

3) The PCI-E version of the Radeon X800 XT that I have requires a special power connector that apparently no power supply has yet. So, the vid card makers provide an adapter that comes in their retail box with the vid card. Trouble is, I got an OEM vid card to save a few bucks, so the adapter was not supplied. The card gives an error beep and a warning message when you turn on the computer and stops the boot. I found an article and pic over at Driver Heaven that described how to "hot wire" the card. Something else to be aware of if you you build a system with a PCI-E vid card.

4) The i915/i925 mobos recommend using a 24-pin power supply. No problem, I bought an Antec True 550w 24-pin PS. My Abit AG8 mobo will work with a standard 20-pin ATX power supply, you have to follow the instructs on which 4 pins to leave open. The extra power pins provide additional juice to the mobo power circuitry to feed the hungry Prescott. I think there is 24-pin to 20-pin adapters, so I bet there will soon be 20-pin to 24-pin adapters too.

5) CPU info: P-4 Prescott 540 LGA 3.2 with 1M of L2 cache and 800 system bus. S-spec=SL7KL, week 26 of 2004, and made in Malay. Default vcore=1.3625v.

6) Cooling and temps: I have four 80mm case fans: front intake, side cover intake with duct, rear exhaust and top blowhole exhaust. CPU cooler is a TT Jungle installed with AS3. Stock N/B cooler and vid card cooler. Temp at default speed (204 FSB for some reason) is listed below.

Idle temps: CPU=43, Sys=33, PWM1=35, PWM2=36
Load temps: CPU=60, Sys=34, PWM1=43, PWM2=44

7) Overclocking first impressions. So far I'm not a happy camper. I only got 3.4 gig out of it last night before I hit a wall. I must admit I haven't figured out the N/B voltages yet. The BIOS settings are different than what I'm used to, so I'm still learning. The Uguru thing allows overclocking in the Windows environment, so I'm testing that too. I was also using the original 11 BIOS. I just now updated the BIOS to version 15, but haven't tested it yet.

8) I will post more later when I get more time to tweak stuff. Feel free to ask questions or better yet, offer suggestions on how I can get a higher O/C.

 

[DanIdentity]

Great write up! A few comments:

1. For anyone building a system with a PCIe vid card, the Antec NeoPower PSU has a PCIe power connector.

2. Something is definitely wrong if you're only getting a 200MHz OC of it. You've probably already done this, but maybe try relaxing the RAM timings and dividers and try to find the limit of your mobo and CPU. From the reviews I've seen the chips OC better than that. Definitely update to the newest BIOS, looks like they've made a good amount of improvements.

3. I just heard back from Thermalright about their LGA775 coolers. According to them, in ~10 days they're releasing their LGA775 mounting kit for the XP-120 and XP-90. Just something to consider.

Good luck!

 

[DanIdentity]

Oh, and the default FSB is 204Mhz because Abit and Asus have started doing that to give them the edge when they're benchmarked at "stock" FSB. I've seen it mentioned in several reviews.

 

[eyepiz]

I thought the LGA775 would be a sperior overclocker then the 478 since there made for prescott chips

 

[jenko]

Keep the info coming batboy!
I think the 520watt OCZ psu comes with new card plug.

On agp motherboards you have to lock the agp/pci right can you do this now? could be a overclocking prob?

 

[batboy]

YIPPEE!

Update: I'm now running at 3.6 gig on default vcore and default nbv (northbridge voltage).

The BIOS update was the key. I had read in a couple reviews where they had good results by just leaving the PCI-E frequency on "auto". Well, with the original version 11 BIOS there was no setting options, only "100 MHz". The new 15 BIOS added the "auto" setting, which is what I have it set to right now. Wow, what a BIG relief.

If I try to lock it at "100" setting (default speed for the PCI-E is 100 MHz), I can't get anywhere. That was my problem last night. I've read several reviews today that claim there is something wrong with the PCI-E lock. That is quite obvious to me now too. Hopefully they can fix it with just a BIOS update. People that are thinking about jumping to the LGA775 platform might wait a little longer to see what happens in that regard.

Ok, I'm going to test for stability, then run some benchies at 3.6 gig. Might be a while before the next update.

 

[batboy]

I'm now at 3.8 gig (237 FSB) and still at default voltages (except vdimm). I was able to run several benchmarks at this speed. Seems quite stable, although I haven't Primed it yet.

The bad news is that load CPU temps are up around 68-69 degrees. Looks like this TT Jungle heatsink ain't the best for hardcore overclocks. If you plan on running one of these LGA775 systems on air, wait until Thermalright has that mounting kit for the XP-90 heatsink.

Oh well, if this system works out good, I'm getting a Swiftech MCW6000 water block and LGA775 retention plate. Hmmm, maybe I can figure out a way to adapt my MCW5000 water block to the socket, that would be sweet.

EDIT: I spoke too soon. I got a lock up on pass #9 of Sandra's CPU Burn-in Wizard.

 

[batboy]

Looks like 3.9 gig is about as good as it gets on air and still be stable. Had to bump up the vcore a couple of notches. Maybe water will help O/C a little higher. I'm actually a little surprised it did this well on air. But, I was hoping for that elusive 4 gig screenshot (boo hoo).

 

[stang8118]

wow great info batboy! Gonna wait myself till mid next year to switch over to the pci-x/LGA socket.

 

[flapperhead]

Looks like 3.9 gig is about as good as it gets on air and still be stable. Had to bump up the vcore a couple of notches. Maybe water will help O/C a little higher. I'm actually a little surprised it did this well on air. But, I was hoping for that elusive 4 gig screenshot (boo hoo).

nice, wow youve added quite a collection of rigs in the last year..

 

[batboy]

I'm almost convinced it's the lack of a working PCI-E frequency lock that's holding me back.

I can do a stable 3.9 gig at just 1.4v, but it won't even start Windows above that. I've tried using the 6:5 memory ratio and extra vcore, more N/B voltage, everything I could think of. This CPU has more in it than that, I know.

Went through my junk box and found an ancient ISA vid card and an old 4 meg AGP 2X vid card. Then I remembered the old P-200 tucked away in the corner of the spare room. Sure enough, it has a working Orchid PCI vid card.

I'll be back in a bit. Going to slap this fine piece of old low tech component into my new high tech machine.

 

[batboy]

So much for that theory. Exactly the same O/C even with an old PCI vid card instead of the new PCI-E vid card.

So far this year I've hit 3.8 gig with a 2.8E, 3.8 gig with a 3.0C, 3.9 gig with a 3.0E, 3.9 gig with a 3.2C, and now 3.9 gig with a LGA775 3.2. Sheesh, I'm cursed. Can't hit 4 gig at all.

 

[DanIdentity]

So much for that theory. Exactly the same O/C even with an old PCI vid card instead of the new PCI-E vid card.

So far this year I've hit 3.8 gig with a 2.8E, 3.8 gig with a 3.0C, 3.9 gig with a 3.0E, 3.9 gig with a 3.2C, and now 3.9 gig with a LGA775 3.2. Sheesh, I'm cursed. Can't hit 4 gig at all.

Man I feel your pain, maybe it's time to go watercooling on that LGA.

What's interesting is that based on your findings, it looks like Abit has managed to completely lock the PCIe bus frequency.

 

[batboy]

No, not completely locked, but managed. I saw that the PCI-E bus at 112 MHz at either 3.8 or 3.9 gig.

Yep, this CPU deserves to be watercooled next. Since I need a new water block, I might just completely revamp my H2O system. I need better hoses. The ones I have are too thin. I'm half tempted to eliminate the video water block, perhaps that will help the CPU run cooler? Might be a few weeks before I get around to doing that stuff though.

I might explore ways to improve N/B cooling. I noticed system temps were getting a little high. I doubt there is any aftermarket coolers yet. Looks like they went back to the standard 2 hole mount for the N/B sink.

I also need to check out whether the mosfets are getting hot. There sure are a lot of them. Ah yes, lots to do and play with. And I haven't even started overclocking or benching the vid card yet. Cooling mods for the X800 XT might be the next thing on the agenda.

 

[situman]

Heartd the Maze 4 fits thge x800

 

[DanIdentity]

Question batboy, are you running IDE or SATA hard drives with that overclock? I've heard SATA drives limit overclocking on the new 915/925 chipsets too.

 

[batboy]

Hey Dan, I'm glad you thought of the harddrive issue. I too have read about some people having problems with SATA drives and then I sort of spaced it off. Yes, I'm using a SATA HDD.

I just so happen to have an extra 80 gig Maxtor IDE harddrive that is not being used at the moment. Perhaps it might be worth while to give it a try just to see what happens, hmmmm.

 

[Zebbo]

Hey Dan, I'm glad you thought of the harddrive issue. I too have read about some people having problems with SATA drives and then I sort of spaced it off. Yes, I'm using a SATA HDD.

I just so happen to have an extra 80 gig Maxtor IDE harddrive that is not being used at the moment. Perhaps it might be worth while to give it a try just to see what happens, hmmmm.

With SATA-drives you'll probably hit the wall somewhere ~260FSB, just don't know though does IDE-drive fix this?

 

[batboy]

Good question Zebbo. I don't know. Looks like I need to do some more research today.

One cool thing about running this rig at 244 FSB (3.9 gig), I can use my 2X512 Mushkin PC3500 level 2 RAM at the 1:1 ratio using 2-2-2-5 timings with 3.2v. This AG8 mobo allows up to 3.2v vdimm with good VTT tracking. If you go above 3v, you have to manually turn off the high voltage "beep" (alarm).

 

[DanIdentity]

Trying out your O/C with an IDE drive instead of the serial ATA probably won't matter, but I brought it up because I know that it is a factor in the Intel overclocking "lock."

This is an explanation of the lock from Anandtech:

1. PCI Express floats in the Intel 925X/915 chipset. PCIe frequency exceeds the capabilities of PCIe cards at about the 10% overclock level. Neither Asus or Abit or any other manufacturer that we have talked with has been able to effectively lock the PCIe frequency in the new 925X/915 design. This is the major roadblock to overclocking on the 925X/915, as any attempt to lock the PCIe frequency limits overclocking.

2. The Northbridge and Southbridge link frequency also floats with the CPU frequency, and since link frequency is monitored at startup, values higher than 10% cause system shutdown. It is true that increasing the voltage to the chipset increases tolerance in this area, but you only gain about 10 MHz to 15 MHz by applying voltage (CPU frequency can increase from 220 to 230 to 235), since the PCIe and SATA issues are not corrected.

3. SATA must be fixed at 100 to function, but the SATA frequency is also influenced by the link frequency. SATA drives simply disappear when the link frequency exceeds the 10% overclock. This can be extended with a bit of voltage, but voltage is not a fix for this issue either.

Although it does look like SATA drives don't really become an issue until you're above ~260 FSB.