Offler

Hi. I just get one of these nice boards so after more that 8 years i am going to build up a new system.

Config:
Xeon x3360
Ram Nanya 1600 mhz at 1,5 volts.

The board is bit older, bios is the newest, custom cooling waits to be builded.

The board was owned by one OC guy here and the board was able to run on 500 FSB for some hours, and also on fsb 520 with dualcore E8500. As long i know Quad cores are not so overclockable as dual cores. I have seen FSB 500 with my own eyes.

My goal was to run the system on FSB 533 while memory will be underclocked to 1066 with better latency timings, but only the best OC maniacs were close to such high FSB and the chipset was not x48.

I have read (and i am reading) articles by EVA2000 about this board, but most CPUs are dualcore, so there will be some difference for me.

I will post here some additional information, such as GTL tables released by DFI some months ago, and other information. I just miss some experience with modern boards and i really have no experience with quadcore cpus...

Also the board was described as one which is not so easy to overclock, but i believe that good cooling system can do its work, also i am quite amazed at its bios.

Current situation:
FSB 400 - No bios changes needed, everything running juuust fine, except memory which needs to be underclocked.
FSB 433 - Voltage has been added everywhere and system is still not fully stable. At this point i am still 100fsb from my goal.
temperatures are around 40 on CPU with stock cooling, 55-65 on NB with stock cooling.

Expectations:
533 is really too high for normal day to day running. The goal was set as highest possible frequency on which i plan to work, but never higher. FSB 450 should be fine and everything beyond that limit should be excellent.

Current problems:
- I really dont have idea which cooler can replace stock cooling on CPU power source - on volterra chips.
- dont know if i am limited by NB, SB, Clockgen, memory, CPU or temperature. Maybe setting lower multiplier for CPU will solve some questions.

If there is anybody who tried this board with quadcore i will appreciate his experience. Thanks.

eva2000

hi Offler and welcome

First off quad core 533FSB on DFI board is a bit unrealistic, I've had 2x E8500 and 5x E8600 dual core and best fsb on DFI UT X48-T3RS i could do on water cooled cpu was between 505-515fsb only for benchmarks, 24/7 was around 500fsb for dual core. I only have QX9650 tried on this board and max fsb was around 445-450fsb for benching around 435-440fsb for 24/7. So 533FSB for quad core is very unlikely although you might have a luck that I don't have heh.

PWM temps I just had fans over the area to keep cool. But fsb limitation is most likely the board chipset used by DFI and the clockgen used for it. The DFI DK X48-T2RSB Plus uses same clockgen as Asus Rampage Extreme and allowed all my E8600 to hit consistently higher max fsb around 550-570fsb same cooling with max validation around 590-600fsb for dual core. Didn't try quad core on DFI DK X48-T2RSB Plus though, but I'm sure it will have potentially higher quad core fsb than DFI UT X48-T3RS.

Offler

Ok. This seems interesting for me...

1, I need more description about both clockgen used. net will to this for me, so dont care about it. Voltage, manufacturing process, and all possible differences between them.
2, Mostly it is known that quads have lower FSB but the reason was always referred to the CPU and dual-die construction, saying that those interconnects just "buzz". If we have problem with clockgen, why there is a difference between dual and quad core?
3. I am amazed that i was even able to hit same limit FSB as you with stock cooling (i just changed thermal paste).
4. Former owner was selling this board because FSB 500 with his E8500 was unstable - BSODs. I reccomended him installing an additional fan nearby chipset and the BSODs were gone.

When i saw that cooling i noticed that large portion of board was air cooled with a 12cm fan, including memory and Clockgen.

If there is a trick with cooling the clockgen it might help. Arctic Cooling Silver 5 thermalpaste has one very special feature. It can soak into chips covered by ceramic and improve heat dissipation. (tested with my old chipset). Its because ceramics are like sponge.

Clockgen list:
http://www13.plala.or.jp/setfsb/

My board has ICS 9LPRS918JKL-F. have googled a bit and i found some Asus Rampage boards with it.
At all the chip is quite hot even on default FSB.

Offler

ok, i just return a few moments to my board...

From my experience the best temperature for each chip in computer is 40 degrees of celsius. Normally the NB and CPU have 50-60 degrees, while average temperature of environment is 22-27 degrees. No component in my machine is too hot, for normal operations, but i am pretty sure they are too hot for normal OC (lets say more than 20 percent OC).

When the temperature difference between air and component is more than 15 degrees the heat dissipates througt other componets, or worse the heat is accumulating somewhere.

This is also problem of clockgen on this board. The chip was cool on touch, but i put on it copper heatsink and temperature of it was more than 45 degrees. I fear there are some places on the board which can accumulate heat.

At this point i still dont know weak places of this design, with exception of cooling system.

so what can i do...
1. Memory underclocking.
This was planned, because i want to run memory on synchro frequency in dualchannel. Right now i reached FSB 450, paid with blood because voltages on many components are high.

2. CPU - FSB Overclocking/multiplier underclocking
Just for test purposes i will lower the multiplier. This will show how the board handle its own parts without influence of CPU.

Mostly those boards were critcised because DDR3 which does not run correctly on them. Thats why i bought Nanya Elixir memory with 1,5v and 1600Mhz. Low voltage proves that the memory is not overclocked by default. Problem is that I was unable to run the computer on memory speed 1600Mhz, but when the memory was underclocked on 1333 the system worked.

From my experience board runs on 400mhz fsb, CPU too without any additional settings from optimized defaults, but the memory dont... (according to Evas OC guide). I didnt read the guide completely, but i remember something about memory profiles for certain memory producers and chips, but nanya was not mentioned.

Memory cannot reach its stock settings, but i was not planning to reach them, the memory will be very underclocked...

eva2000

Usually a 120mm fan over select parts of the any motherboard help

Could be memory compatibility issues with certain modules ?

Offler

I think that stock board cooling is going down. Even with better airflow it still accumulates heat in some places of the tubes. Again small copper and aluminium heatsinks designed for memory modules will do better job, i just dont know how to connect them on some places. I dont like thermal doublesided stickers, but its better than nothing.

Memory... i have just one single piece of those nanyas. Their parameters are excellent, on datasheets of course, but i wanted better latency rather than high frequency. I just cant wait to test dualchannel but im out of money. Surely not problem with module compatibility on dualchannel, just i fear that i have to do a lot of fine tuning in Bios. With such amount of settings it may took months.

Are there any exactly and technically described problems with x38 and x48 memory controller hubs? Compatibility issues... anything... Globally were issues with those boards related to ddr3 memories, while ddr2 variants were running more smoothly on higher frequencies, but i dont really know any technical reason of such behaviour.

eva2000

No idea for memory compatibility as all memory i used worked but mainly used well know brands such as crucial, corsair, team, ocz, mushkin

Offler

I have also checked Bjorn3d review... They had almost same OC limits i had. 424 mhz was first block, and 447 is almost impossible to get stable. They used q9550 - practically same quadcore i have.

Two same cores have same limits on same boards. so... this is too similar. There must be some kind of problem. When i underclock the CPU and memory, while FSB will remain same, and stability too i will be pretty sure that problem is with boards chipset, but not the CPU power source.