[18th Anniversary] Asus Black Pearl - P5k-Premium Testing Report

[wilson]

Hi guys ive just built up this board and need some advice please. Im stable at 3.4Ghz well i think i am as it pass`s most of the stress test`s but sometimes it seem`s to reboot when loading windows ,say once out of every 8 boot ups. I know its not alot but i would like to know what volt settings i should use for my nb,fsb ect.
Cheers Paul

[jonny_ftm]

Here it is, my P5K Premium plunge.

Here's the box and motherboard. It is rev. 2.00G (the only revision I know about!). If I do believe previous topics, I have a new pack (blue box with new manual and all red caps)





For the discussion on new/old packs, look these links:
http://www.xtremesystems.org/forums/...&postcount=377
http://www.xtremesystems.org/forums/...22#post2490022

First thing is to remove the crappy stock cooler (in my opinion, it is crappy, but I didn't give it a single chance)



Looks easy once removed but it took me 45mn. The holes of the push-pins are machined less than 0.5mm to the integrated circuits on back of the PCB. So removing and placing the push-pins is a risky thing. You need much care. Also, even that I never fired it up, the ASUS thermal compound was sticking solid on the SB and NB. I had to give it a serious heating time with a hair drier. Also, take care here, not more than 20-30 secs. Better test each time, than give it another shot of hot air. The heatsinks should come without any effort. Removing the mosfets coolers is much easier as there's only a thermal pad.

Harder part now: removing the crappy solid and sticking ASUS thermal compound. You need a lot of alcohol and patience. Don't exert much pressure on the naked NB. The SB is covered with and IHS, so it is more resistant to pressure. To remove the compound quicker, I gave it some more hot air shots (but be carefull again).

The problem is that the SB, NB and first mosfet heatsinks are soldered with a heatpipe system. If you want to reuse the SB heatsink, you'll have to cut that heat pipe and loose your warranty. I agree that NB and mosfets coolers are not enough for high overclocks, but the SB would be more than enough. Having to buy a new one is a waste of money. There are very cheap ones hopefully, like the Zalman ZM NB-47J. The problem is that the SB is in the length of the PCI-E slot. A long dual slot VGA card (like my 8800 GTS 512) will conflict with any SB cooler higher than 10-11mm. There are low profile SB heatsinks at Enzotech and Petra's shop, but they are some expensive and hard to obtain. Thus, I went with the aluminium Zalman ZM NB-47J which I cut a bit so it can fit.

Now, placing my two HR-09U type 2 on the mosfets, my HR-05 IFX on the NB and a Zalman ZM NB-47J on the SB. Thermalright web site shows that you need a HR-09U and HR-09S both type 2 for the mosfets. If you're watercooled, two HR-09U type 2 will fit perfectly. But, on air cooling, a HR-09S type 2 could be needed as the U variant can conflict with the CPU air cooler. Be carefull that S type can conflict with some removable motherboards plates. Check the specs on thermalright site before buying them.

The layout of this motherboard is one of the worst I'v seen. The HR-05 IFX won't fit horizonally because of the RAM slots too close to the NB. It also won't fit vertically because it will interfere with PCI-E card (heuh, I'm no longer sure of it) and it will cover the CPU cooler fixing holes. We're left with a diagonal fit. If you're aircooled, take care as it can conflict with CPU cooler as the NB is too close to the socket.





In this last picture, you can see the needed clearance on the SB with a VGA card such as the 8800 GTS 512. The card is so long that it will cover the misplaced SATA ports (SATA 1 and 3). The only solution is to use angled SATA cables. But the angle will than render useless the ports SATA 2 and SATA 4. So, in all cases, with such VGA cards, you loose two SATA ports. ASUS really did crap here.

To continue with layout issues, look at the Analog CD Audio connector near the PCI 1 slot. The 8800 GTS 512, being dual slot, will conflict with the connector. I could plug the audio cable, but I have to force the card and the mb is bending a bit in that place

Now, the system mounted

The PA120.3 with 3 Scythe S-Flex SSF21F at 5v, speedfan regulated:



The lateral panel modded with a 120mm fan and the MCR-220 mounted on the rear with two Delta fans fixed at 5v



Now, a closeup to see the layout problems



In the circled area N°1, we see that the RAM slots are too close to the PCI-E card. It will be hard to fit a DDR cooler. In my case, the OCZ XTC Memory Cooler can't fit as it should. I had to remove the lower fixation and it is suspended only on its top. Also, it is so close to the VGA card that I recovered the lower part with an electrostatically protective film to avoid a short on the VGA card.

In the circled area 2, we see the 8800 GTS 512 blocking the SATA ports. I could use angled cables, but as you see, the ports 2 and 4 are dead.


Everything is now mounted. And there's a last bad layout (yes again). The PCI-E slot is too close to the bottom of the card compared to my previous P5B Deluxe. The 8800 GTS 512 conflicts with my pump, it was hard to fix it despite I have a rather big tower (Thermaltake Shark)



Some discreet lights:



Now, time to see if all this work is worth it. Well, here are my results, rock solid, with my Q6600 L740B (check my signature)

Near 12h Prime95 Small FFT:



And, the best program to test the NB/RAM stability with these agressive settings on a quad core: OCCT in RAM mode





Some SuperPi runs:






And 3DMark06 run






This is nothing close to a world record, but, the voltage settings are impressive, here are the bios settings to achieve it: (edit: lowered vNB to the minimum now)





My old P5B Deluxe couldn't stabilize my chip under OCCT above 390 FSB with same components. It also was impossible to over clock my DDR2 above 900MHz. So I'm quiet impressed.

About the voltages, I didn't try lower as I think they are already quiet too low. The vPLL did give me neither less vcore nor better overclocking. It is fine at 1.5v. The NB and CPU Ref voltages would give me random lockups on startup and within Adobe Encore CS3 if set lower than the maximum.
I see my volts are one of the lowest for such a high FSB on a quadcore at 3.8GHz. Maybe I got a good board, but for sure the NB and RAM cooling are helping. At my OC, with an excellent incase airflow, the system temperature is at 48°C (25°C ambient) after some hours of OCCT in RAM mode. I do believe that an active cooling on the NB is the only way to grant you a high FSB on a quadcore.

I also see a lot of people putting vPLL and clock over charging on Auto. A good advice is never put sensitive voltages (vDDR, vcore, vFSB, vNB, vPLL) on Auto. vPLL is a CPU killer, and often useless except some handpicked CPUs. And on ASUS boards at least, the voltage will increase by its self when you overclock on Auto and you have no way to monitor the killer vPLL. My piece of advice...

The Static read control feature gives the memory a dramatic boost in performance. I couldn’t enable Transaction Booster but relaxing it to 0 instead of 3 worked and gave the memory benchmarks another impressive boost. SuperPI 32M improved by no less than 5 sec with also a remarkable 3DMark06 improvement.

Finally, here’s a SuperPI 1M at 4GHz. I didn’t even check for stability as I didn’t like to bump many volts in my components since I’m limited by the heat from the CPU for extra overclocking (should get a new CPU WB instead of my bowed Apogee GT maybe not well contacting the surface of my lapped IHS)





I made many tests using multiple combinations (8x475 MHz and 9x423MHz with different memory speeds and timings). I also gave the strap 266 a look. It is a very interesting setting giving some more memory performance and low latencies. The strap 266 settings need high voltage on RAM (2.30-2.35v to stabilize it) and also higher volts on the NB. My RAM is rated at 2.35v 1200MHz and CAS5, so it could handle it, but for the relative performance gain…
In yellow is the setting I chose for 7/24 use. I also added now results with new bios 0612




It looks like the 8x475MHz setting at strap 266 and CAS 5-5-5-18 is the best performer though if you don’t mind putting 2.3-2.35v in your RAM
You can also see the important boost by relaxing tansaction booster to 0 instead of 3

Bios 0612 adds also some gain in memory performance. It is reported to give better memory overclocks

Here’s also a Crysis benchmark result at my settings:

16.02.2008 06:15:44 - Vista

Run #1- DX10 1280x1024 AA=2x, 32 bit test, Quality: High ~~ Last Average FPS: 24.10
Run #2- DX10 1280x1024 AA=2x, 32 bit test, Quality: High ~~ Last Average FPS: 24.41

Run #3- DX9 1280x1024 AA=2x, 32 bit test, Quality: High ~~ Overall Average FPS: 24.39
Run #4- DX9 1280x1024 AA=2x, 32 bit test, Quality: High ~~ Overall Average FPS: 24.53

Run #5- DX10 1280x1024 AA=2x, 32 bit test, Quality: VeryHigh ~~ Last Average FPS: 14.16
Run #6- DX10 1280x1024 AA=2x, 32 bit test, Quality: VeryHigh ~~ Last Average FPS: 16.65

Run #7- DX9 1280x1024 AA=2x, 32 bit test, Quality: VeryHigh ~~ Last Average FPS: 16.66
Run #8- DX9 1280x1024 AA=2x, 32 bit test, Quality: VeryHigh ~~ Last Average FPS: 16.76

Run #9- DX10 1280x1024 AA=2x, 32 bit test, Quality: Custom ~~ Last Average FPS: 17.19
Run #10- DX10 1280x1024 AA=2x, 32 bit test, Quality: Custom ~~ Last Average FPS: 17.50

Run #11- DX9 1280x1024 AA=2x, 32 bit test, Quality: Custom ~~ Last Average FPS: 17.70
Run #12- DX9 1280x1024 AA=2x, 32 bit test, Quality: Custom ~~ Last Average FPS: 16.86

I must say I'm impressed how easy it can be overclocked. It is really not a frail board like the X38 motherboards as I read in their topics. You put your volts and settings and you quickely know if it handles it. In my case, I'm only limited by my other components. Also, two features that definately should be enabled are: static read control and transcation booster. Sadely, when you overclock, you can't enable them both. Transaction booster won't work on important memory overclocks. Despite this, disabling the transaction booster and trying not to relax it much (0 instead of 3) would still give an important boost in benchmarks that you also see in SuperPI and 3DMark06.

A note about Samsung/WD SATA 2 drives. My Samsung 120Gb drive works without any issue. I also made the jump to a Western Digital Caviar SE16 WD6400 AAKS without any issue. On ASUS forums, the issues concern early motherboard samples. You can know your build date by serial number on pack. Mine is 78xxx meaning 2007 August. MB from 77xxx and before are at high risk to be affected by the bug. You can RMA your board to get a newer one.

Well, hope this complete view can help someone mastering this nice board. Feel free to comment on my settings / results / interpretations.

edit 27 April 2008

Added underlined text:
- Updated bios settings screens
- Added Bios 0612 feedback
- Added GTL Ref Voltages issues
- Added WD/Samsung SATA2 HD issues workaround

edit 12 May 2008

Bios 0612 gave me memory performance improvement (cf above table) but I begun getting some random hang-ups on windows startup. During daily use or benching it was still stable Prime95 small FFT and OCCT in RAM mode. I couldn't sort it out increasing vNB and vFSB. Finally, I realized it was memory issue, now needing looser timings (it became unstable) or more voltage (2.20 instead of 2.10v with bios 0504). Since I opted for 950Mhz CAS4 to have 2.10v on my RAM, and since I now had to give it 2.20v, I opted for a larger bandwidth setup (if it has to burn, at least it gets for respected performances ). Here are my expierience facts
- 8x475 at strap 266 with 1186 DDR2 CAS5 and 2.25v were perfectly stable on this new bios. It also gives a good performance increase compared to strap 333. I didn't need to boost my NB or FSB voltages

- 8x480 at strap 266 with 1200 DDR2 CAS5 and 2.25v needed 1.30v FSB instead of 1.20v to stabilize the CPU at 3.84GHu instead of 3.80GHz without adding vcore. So, now, this is my new setup, still rock stable and no more hang ups on boot.

- My Q6600 is fully stable P95 small FFT at 3.90GHz (9x434MHz) at PLL 1.5v and vFSB 1.30v but a vcore of 1.47v. I couldn't stabilize the 3.9GHz at 8x488 or even 8x485 despite any increase in vFSB, vPLL, vNB... It could be the CPU, the memory or the motherboard limiting me. I'd have to try strap 333 to exclude memory from equation and to loosen NB stress to partially exclude motherboard. Maybe I'll test that later, but no real benefit as I won't gain much CPU speed and the memory will loose performance

Here are screens of this new setup. Note that I opted for shorter periods of benching, especially with OCCT in RAM mode. OCCT will burn the RAM, and I've never seen errors after 2h of run in RAM mode. Also, at the end, you'll find all bios screens for this config

I could tweak the trfc and other subtimings (it seems stable with a trfc of 25 and settings of 5-5-4-10-4-25-3-3-3), but it only gave a poor boost in memory copy/read performance, so not worth going the hassle of OCCT RAM stability sessions to tweak it

[wilson]

Nice m8!!
Just thought i would post some pics of my rig .




Its just a shame that the nb and vrm`s dont give out temp readings

[jonny_ftm]

Nice and very clean setup. I see you chose the HR-05 SLI for the SB, quiet a big investment, but very clean setup I have to agree. I think though that your Reserator is limiting you in temperatures as your circuit is very restrictive too. But I can think you have a very silent setup...

Otherwise, I just could set Transaction Booster on Disabled but relaxed to 0 instead of 3, it gave me a huge improvement in everest. Will update my post once I finish a new 10h OCCT RAM stability test in that setting.

[turtletrax]

My HTPC is finally complete and running as planned (minus the SuperPI issue)

Specs are in my siggy.

[Supertim0r]

good old AMD64 fan cooling the chipset/ram

[turtletrax]

good old AMD64 fan cooling the chipset/ram

God I love those fans. I even cut a 70mm hole in my Xbox 360 and mounted it internally right before the CPU heatsink. Fit perfectly and has helped bigtime. It is the only (easy) source for 70mm. Thank god I had about 12 AMD K8 CPU's, many of which I clipped HS/Fan from them

[sesdave]

Here it is, my P5K Premium plunge.


First thing is to remove the crappy stock cooler (in my opinion, it is crappy, but I didn't give it a single chance)

Nice work - in case you hadnt noticed on previous post to help removal of stock heatsinks - use the small end of the barrel of a click type pen to compress the plastic compression pins . It just fits nicely and out they pop with no effort or damage to the MB and pins unlike trying to do it with pliers.

[Shocky]

Just installed this motherboard and all is well but there seems to be a cold boot issue, when first starting up it will reset itself and then powerup, why is it doing this? as it seems completely stable in windows/games and during prime.

I'm using the latest bios, 0508 I think.