HighTreason wrote:

2A? Most of my manuals state between 250ma and 500ma but it seems to vary widely, so be careful ...

775 mobos are relatively modern and usually came with decently rated headers.

But as you well say better check the manual.

In this 2003-vintage 478 mobo the CPU_FAN header is rated @ 0.74A max:

Both the two Arctic F12 (0.15A each) and the CM 212 EVO fan (0.37A) are all daisy-chained into the CPU_FAN header (0.67A total, 90% of header rating):

No nuclear melt downs. In fact with SpeedFan in charge it works like a charm, the OCed Preshott is kept well at bay with very little noise:

On the topic of decent socket 775 boards and outrageous prices.

I just scored this board with an E6850 for ~35 Euro, not a bad deal. (One of the sellers pictures)

A boxed Asus X38 DDR3 board is usually ~60 Euro

It seems nearly every motherboard manual I've looked at has never given specifications for what the fan headers can handle. The only board I have which documents it is a Tyan server board (socket 940, not 775), which allows 2A on most headers, and 3A on a couple others which lack PWM support.
It is one of those boards that has PWM on 3pin headers. The manual calls it PWM, not voltage control. I don't know for certain that this is correct, but I assume it's pulsing the voltage on the +12v pin, or pulsing the ground side, whichever. I have noticed that those headers do not regulate as steadily as the 4pin CPU headers do. The 3pin PWM headers have a more noticeable pulsing effect. I know this happens when it's left under BIOS control, but I can't remember if I had the same issue when using software control.

In the absence of documentation, I think it would be a fairly safe bet that any board from the 775 era onward should support at least 1A on each header. In case it isn't obvious to anybody reading, please be careful with older boards though - their headers get less robust the further back you go. An old Epox slot-1 board of mine was found to have transistors only good for something like 0.15A, about the same as a typical CPU fan back then. It wasn't documented, as usual.

From my research best combination is an DDR3 P45 motherboard that supports an Xeon mod, eg X5492.

Stojke wrote:

From my research best combination is an DDR3 P45 motherboard that supports an Xeon mod, eg X5492.

Now when I have spent some time with the Gigabyte EP35C-DS3R Im starting to think that everybody was wrong and that even these older Combo boards works really well with DDR3.

The current setup Im trying out.

Gigabyte EP35C-DS3R
C2D E8400@4500 MHz (FSB 500)
2*2GB DDR3 @ 1600 7-8-7-20
XFX GTX 285 Black Edition

So far I have only tested this setup with XP but for a dual core system it sure feels fast.
Most P35 boards should support all socket 771 Xeons.

I'd like to tell you of my own tale, maybe it'll help someone.

About a couple of years ago, I found a P45 board for dirt cheap. Bought it for about 10E, because the seller thought it was troublesome. Turned out to be an MSI P45D3 Platinum, an utterly insane board equipped with the P45 chipset, DDR3 memory support and glorious copper heatsinks. Anyway, I threw a C2D 6550 in there (my first Core 2 Duo), a CoolerMaster Hyper 212, some DDR3 sticks and a SilverStone ST75F and did some experimenting.
I'm not really an expert when it comes to extreme OC. I tweaked the CPU and NB voltage a little bit and basically left everything else on auto. I was able to reach 3.5GHz with my poor 2.33 Core2Duo, which I thought was amazing. System was relatively cool at ~30-35C on idle.

The motherboard was kinda flakey with DDR3 sticks, but that is to be expected somewhat. I had great issues with using more than 2 sticks, but in the end I managed to reach 8GB DDR3 1333MHz, by using identical sticks. So, I thought it would be really cool to acquire a Xeon quad core and see how high up I could overclock it, fool around a bit and then maybe sell it all together to recoup the costs or keep it for a second machine. I found the E5450 to strike a great balance between TDP/multiplier. It's basically a Q9650 with a TDP of 80W.

I finally got one for 50 Euro a few weeks ago, installed and watched my system crawl on stock settings. I tried removing RAM sticks, replacing the GPU even trying to format the system, but it would usually hang at the Windows 7 boot animation or just do a soft-reset. I could never get to installing Windows 7, the system was utterly unstable. The seller was local, so they had no problem with me returning the CPU. Way I have it figured, the BIOS for my mainboard did not support the Xeon and as such performed like it did. So, I just quickly set up the system as it was before, with the Core 2 Duo in place, and sold it for a decent amount.

So, a word of warning, check very thoroughly if your motherboard supports the Xeons or Xeon model you have in mind before buying or at least make sure the seller accepts returns. Maybe someone else has a better idea as to what may have happened in my case?

F2bnp wrote:

I'd like to tell you of my own tale, maybe it'll help someone. […]
Show full quote

I'd like to tell you of my own tale, maybe it'll help someone.

About a couple of years ago, I found a P45 board for dirt cheap. Bought it for about 10E, because the seller thought it was troublesome. Turned out to be an MSI P45D3 Platinum, an utterly insane board equipped with the P45 chipset, DDR3 memory support and glorious copper heatsinks. Anyway, I threw a C2D 6550 in there (my first Core 2 Duo), a CoolerMaster Hyper 212, some DDR3 sticks and a SilverStone ST75F and did some experimenting.
I'm not really an expert when it comes to extreme OC. I tweaked the CPU and NB voltage a little bit and basically left everything else on auto. I was able to reach 3.5GHz with my poor 2.33 Core2Duo, which I thought was amazing. System was relatively cool at ~30-35C on idle.

The motherboard was kinda flakey with DDR3 sticks, but that is to be expected somewhat. I had great issues with using more than 2 sticks, but in the end I managed to reach 8GB DDR3 1333MHz, by using identical sticks. So, I thought it would be really cool to acquire a Xeon quad core and see how high up I could overclock it, fool around a bit and then maybe sell it all together to recoup the costs or keep it for a second machine. I found the E5450 to strike a great balance between TDP/multiplier. It's basically a Q9650 with a TDP of 80W.

I finally got one for 50 Euro a few weeks ago, installed and watched my system crawl on stock settings. I tried removing RAM sticks, replacing the GPU even trying to format the system, but it would usually hang at the Windows 7 boot animation or just do a soft-reset. I could never get to installing Windows 7, the system was utterly unstable. The seller was local, so they had no problem with me returning the CPU. Way I have it figured, the BIOS for my mainboard did not support the Xeon and as such performed like it did. So, I just quickly set up the system as it was before, with the Core 2 Duo in place, and sold it for a decent amount.

So, a word of warning, check very thoroughly if your motherboard supports the Xeons or Xeon model you have in mind before buying or at least make sure the seller accepts returns. Maybe someone else has a better idea as to what may have happened in my case?

Most problems seem to have to do with bad stickers or bad contact from either bad sticker placements or not enough trimmed down sockets.
X38/X48/Q35/Q45 dosnt support any E/L/X54xx Xeons at all though so those with any of those chipsets dont need to worry about the Xeons 😀

The X5450 sells for 25 Euro in Sweden, they rarely get sold. I payed 30 Euro for my X5460 so im very glad our Ebay clone (Tradera) isnt connected to the international Ebay.

shamino wrote:

... In the absence of documentation, I think it would be a fairly safe bet that any board from the 775 era onward should support at least 1A on each header. In case it isn't obvious to anybody reading, please be careful with older boards though - their headers get less robust the further back you go. An old Epox slot-1 board of mine was found to have transistors only good for something like 0.15A, about the same as a typical CPU fan back then. It wasn't documented, as usual.

^ Wise words.

When the CPU_FAN header is PWM, and all the fans in the case are also PWM, there is a trick to control any number of them from the CPU_FAN header, regardless of amperage.

· ONLY the CPU fan RPM signal wire is connected to CPU_FAN.

· The PWM duty cycle wire from ALL fans in case is bundled and connected to CPU_FAN.

· The +12V and ground wires from all fans in case are connected directly to PSU.

This way the mobo only 'sees' rpm fron one fan, but all fans are controlled and pulsed in unison, and there is zero load on the mobo headers.

There are plug & play gadgets for this purpose.

I do remember reading that 4 modules would be challenging. So I only ever used two.

philscomputerlab wrote:

I do remember reading that 4 modules would be challenging. So I only ever used two.

I have heard that aswell. The Gigabyte DDR2/DDR3 board only has two DDR3 slots so I cant test 4 modules but when the X38 board gets here I will try 4 modules. My problem is though that most of my DDR3 memory is in sets of two or three modules, not 4.

I moved from my old Q6600 build to my 4770K build about a year and a half ago. Some of my games were indeed heavily CPU-bottlenecked to the point where I noticed a good performance boost even using the same GTX 480 on both.

That said, the Q6600 system was built in late 2007 and lasted me a good six years of viable use. That's more than I can say for any computer made before the mid-2000s with the rapid hardware advances of the time. Just get it clocked to 3.2 or even 3.6 GHz if you've got the cooling to swing it, and it'll hold its own. To think people were saying back in the day that a slower-clocked quad-core wouldn't be worth it, and now we've got games that don't even run correctly on dual-cores!

It would still be quite viable for many of today's games and everyday computing, really. I might hand it down to someone else in the family once I get something to replace the dead GTX 480, because an 8800 GT doesn't cut it at all for today's games.

I finally found the time to test the Gigabyte EP35C-DS3R DDR3 system with Windows 7 -64

With half the amount of cores and only one GTX 580 this system is perhaps a bit slow for some of the newer games, others will probably run OK at 1080P.

After some testing Im starting to doubt the viability of the Asus P5E3 X38 DDR3 board as the platform for a high end quad core socket 775 gaming rig targeting modern games.

If I had a QX9650 or QX9770 things would be different. I could use the 11x multiplier with 400 FSB or even the 10x multiplier with 450 MHz FSB. As it is I only have a really bad 1.3 VID Q9650 with a 9x multiplier. Getting 450 MHz 100% stable with the Q9650 quad and 8GB memory was not hard but it takes ~1.6V northbridge voltage, anything higher or really tights timings needs 1.65+ northbridge voltage. I know my Gigabyte X48-DS4 is more capable but its a DDR2 board so I will not be able to upgrade to 16GB memory or run 8GB with only two sticks and I have never tried the boards limits with 4x 2GB DDR2.

I would really like to hit at least 9x475 = 4275 MHz with at least 8GB of memory running at a decent speed. Perhaps the Gigabyte P35 Combo board could handle the speed but it only has two DDR3 memory slots so I would be limited to 8GB which is fine. The lack of good vdroop and vdrop control is worse as I would probably need at least 1.5V idle voltage to keep the load voltage high enough for total stability at 4275 MHz. With the P35 board I would also lose PCI-E 2.0. While PCI-E 1.1 perhaps isnt much of a bottleneck for a GTX 580 I would like to be able to use my GTX 770 to see how much it gets bottlenecked by the CPU and memory...

Another choice would be a Nforce 780i board, the one modded for s771 Xeons has two bad caps, not a board I want use with really high FSB for extended periods of time. I do have two other more or less new boards but I somehow doubt their stability running at 475+ MHz FSB with 4x 2GB DDR2. Sadly I do not own a Nforce 790i board which would be a better candidate. I think I will have to pick up a few X48 DDR3 boards to get what I want as even those seem to be hit and miss when it comes to FSB over 450 MHz with a quad core and lots of memory. I do not want to pay the prices these boards normally cost so I will play the waiting game.

For now here is at least proof that you do not need an "Internal Memory Controller" inside the CPU to reach decent latency and somewhat decent memory bandwidth. This is with a dual core E8600, the memory bandwidth would be even better with a Quad at the same FSB and memory speed, lowering the CPU multiplier to 9X dosnt change these numbers much.

Latency

Memory read

philscomputerlab wrote:

I do remember reading that 4 modules would be challenging. So I only ever used two.

FWIW:
On DX48BT2 it got a little squirrelly about a mis-matched four DIMM configuration with a pair of mismatched kits; one of them was rated to DDR3-1333 and the other up to DDR3-1600. If the kit rated for 1600 wasn't in the primary DIMM slots, it would refuse to boot. Switching that around and it was stable, and ran 1333 with no problems (which is correct).

My Dell with an X38 (its in a Precision T3400) didn't seem to care about mismatched DDR2 kits - I have whatever the original Dell memory was, and some discount no-name stuff I found online, and it just booted right up and worked.

Both were/are significantly nicer about 4 DIMM than some older motherboards I've used/owned (e.g. my S939 boards), which is a plus, but the DX48 isn't as painless as some server/workstation boards I've used (e.g. PC-DL Deluxe) or my modern Z97x. Still, with the memory in the proper slots, it ran rock solid for years like that. I'm guessing if it had all 1333 (or even more ideally, four of the same DIMM) it would've been a complete non-issue.

obobskivich wrote:

FWIW: On DX48BT2 it got a little squirrelly about a mis-matched four DIMM configuration with a pair of mismatched kits; one of t […]
Show full quote

philscomputerlab wrote:

I do remember reading that 4 modules would be challenging. So I only ever used two.

FWIW:
On DX48BT2 it got a little squirrelly about a mis-matched four DIMM configuration with a pair of mismatched kits; one of them was rated to DDR3-1333 and the other up to DDR3-1600. If the kit rated for 1600 wasn't in the primary DIMM slots, it would refuse to boot. Switching that around and it was stable, and ran 1333 with no problems (which is correct).

My Dell with an X38 (its in a Precision T3400) didn't seem to care about mismatched DDR2 kits - I have whatever the original Dell memory was, and some discount no-name stuff I found online, and it just booted right up and worked.

Both were/are significantly nicer about 4 DIMM than some older motherboards I've used/owned (e.g. my S939 boards), which is a plus, but the DX48 isn't as painless as some server/workstation boards I've used (e.g. PC-DL Deluxe) or my modern Z97x. Still, with the memory in the proper slots, it ran rock solid for years like that. I'm guessing if it had all 1333 (or even more ideally, four of the same DIMM) it would've been a complete non-issue.

This is my experience with socket 775 and DRR3 so far.

The Gigabyte EP35C-DS3R Combo board would run 2x2 GB and 2X4GB fine in its two slots at any speed up to 1600 MHz as long as you entered all timings manually including all availible subtimings. The board would not even post with auto setting other than in failed OC recovery mode.

The Asus P5E3 Vanilla will not run DDR3 at 1600 MHz what ever I do, not even with two 2GB sticks. Speeds up to DDR3 1550 MHz CL 7-7-7-16 1T works though but only with 200 and 333 straps and really tight subtimings, looser timings wont even post most of the time. The board shows the same behaviour as the Gigabyte board with two Corsair 2GB sticks, auto settings wont even post other than in failed OC recovery mode. With two Corsair 4Gb sticks the board do post at default settings but fails to load Windows. For some reason the board likes a tRFC setting of 60 when pushing memory and FSB, no other tRFC setting will even post with the memory at 1500+ MHz and the FSB at 450+ MHz.

Both these boards seem to have issues with my sticks default sub timings, I have tried two BIOS versions with the Asus board, the release BIOS and the latest BIOS with pretty much the same result. As long as I set all timings manually the Gigabyte board has little issues and the Asus board works fine even with 8GB but is rather picky when it comes to straps and timings, especially at really high FSB.

Edit

I think I just found memory that works a bit better with the Asus board, Corsair Dominator GT 2133 MHz 4GB sticks. The board isnt stable with everything on auto and wont even load Windows just as with the other 4Gb sticks but as long as I set strap and memory speed to known working settings the memory seems to work fine with auto timings. With luck this memory will help me get a bit higher FSB at a lower northbridge voltage than with the really tight timings I had to set with the other sticks. Its a 16GB kit meant for a low power socket 1155 system Im planning to build for the hot summer months though.

Im advancing in the wrong direction...

I managed to get the Asus P5E3 board stable (well 3 hours of Prime Blend...) with 2x4 GB at 4050 Mhz with the memory running at ~1350 MHz CL 6-6-6-14-60-1T at "only" 1.35V Northbridge voltage and 1.35V core and 1.34V VTT. But as I really wanted some more speed I switched to the Gigabyte X48-DS4 DDR2 board. Now Im starting to think that this CPU is really really bad. I could get the Gigabyte board to load Windows at 475 FSB with 2x2 GB memory running at 1140 MHz CL 5-5-5-15 2T but stability was nowhere to be found. Lowering timings and or TRD did nothing, testing other memory speeds did nothing, increasing voltages did nothing unless I used 1.6V+ northbridge voltage and 1.5V+ VTT.

Im back to 4050 MHz at low voltages but now with (slower) DDR2 memory, at least the board diddnt protest when I added two 1GB sticks for a total of 6GB DDR2 running at 1080 MHz CL 5-5-5-15-52 2T. Im to lazy to switch back to the Asus board for better memory performance, now I just want to do a fresh install of Windows 7 and install the game Shadow of Mordor I just bought to see how well it runs on an old socket 775 system (below minimum system requirements).

Here are some porn.

The Gigabyte EX38/X48-DS4 DDR2 board, The EX38/X48-DS5 and EX38/X48-DQ6 boards are identical to the DS4 just other cooling and 2 more SATA ports.

Some manufacturer started adding a 4th power phase, then someone added two more, then...

Here are the settings I wll run for now. This Q9650 is pretty much the worst Q9650 I ever heaed of, the first core is really really weak.

Don't feel bad about the Q9650, I got one in a trash find, and it runs HOT at load. The thing will idle at 45 or so, and then shoot up to even 60+ under load. Wouldn't dream of trying to push it further. And Q9650 is supposed to be all E0.

If you spent the money already on a Q9650, why not get a Q9550S instead (65W TDP)? Lower multiplier, but it might have been stable...

You also might wanna try an E5450. They cost only half of what an equivalent Core2Quad would cost, and have a lower TDP (80W), but you would need to mod both the chip and the socket.

I can tell you from personal experience that in both my experiences with C0 E5430 machines I built, they ran a heck of a lot cooler. In my main rig, I have it underclocked and undervolted, because I have it in a really old case with very poor airflow.

mockingbird wrote:

Don't feel bad about the Q9650, I got one in a trash find, and it runs HOT at load. The thing will idle at 45 or so, and then s […]
Show full quote

Don't feel bad about the Q9650, I got one in a trash find, and it runs HOT at load. The thing will idle at 45 or so, and then shoot up to even 60+ under load. Wouldn't dream of trying to push it further. And Q9650 is supposed to be all E0.

If you spent the money already on a Q9650, why not get a Q9550S instead (65W TDP)? Lower multiplier, but it might have been stable...

You also might wanna try an E5450. They cost only half of what an equivalent Core2Quad would cost, and have a lower TDP (80W), but you would need to mod both the chip and the socket.

I can tell you from personal experience that in both my experiences with C0 E5430 machines I built, they ran a heck of a lot cooler. In my main rig, I have it underclocked and undervolted, because I have it in a really old case with very poor airflow.

The CPU is a dumpster find (I found it in a Dell Optiplex 960 tower) so no worries! 😀 I do own a Xeon X5460 that I run in a nForce 780i board, that CPU does 4.4+ GHz. No X48 board will take any Xeon E/L/X54xx CPU though.

This CPU dosnt run that hot, I can cool it enough with my TR Ultra 120 to keep it under 80C even with 1.5V+ core running Prime at 4.2+ but 100% stability over 4.1 seems hard to find at least in Windows 7.

mockingbird wrote:

Don't feel bad about the Q9650, I got one in a trash find, and it runs HOT at load. The thing will idle at 45 or so, and then shoot up to even 60+ under load. Wouldn't dream of trying to push it further. And Q9650 is supposed to be all E0.

My QX9650 (C0) idles at less than that, but at full load it can get to 60+. When overclocked 20% (and overvolted somewhat) I sometimes would get it to 80C under very heavy load. There is no problem whatsoever with such temperatures occasionally.

Advance! No Retreat, No Surrender!

Well perhaps a little bit of shortening the frontline... I remembered that I have had stability problems with Micron 2Gb sticks and the Gigabyte X48 board many years ago. I switched the 2x2 GB Micron memory for another set of 2x1 Gb sticks for a total of 4x1 GB. The game Shadow of Mordor seems perfectly happy with only 4 GB memory anyhow and runs perfectly smooth at 89 fps average will all high settings at 1280*1024.

I managed to find stability at 4.2 GHz with reasonable voltages, 50PS CPU skew was the key.

Some benchmarks showing the systems performance.

FryBench and specs

3dmark 11

Memory Read

Queen

Hash