Asus P5K Deluxe - Intel P35 Bearlake + Intel E6600 B1 ES / Q6600 Kentsfield / X6800

Discussion in 'Intel x86 CPUs and chipsets' started by eva2000, May 13, 2007.

  1. eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    After setting up my P5K Deluxe the other night, I decided to start off with my old trusty E6600 B1 ES cpu so as to be able to compare temps and results with my other motherboard adventures with the same cpu.

    The P5K Deluxe shipped with 0123 bios but was updated to latest 0304 bios and flashed using USB drive and Ez Flash 2 in bios. You can see the bios screen shots here.

    Asus P5K Deluxe Photos​

    [​IMG] [​IMG] [​IMG]

    [​IMG] [​IMG][​IMG]

    [​IMG] [​IMG][​IMG]

    [​IMG] [​IMG] [​IMG]

    Notes: - Will be continually updated as I progress.
    • Update to 0304 bios to fix the incorrect cpuz reported cpu/memory clock and timings.
    • Asus P5K Deluxe supports cpus from 800FSB and upward. But doesn't support 533FSB cpus. Read here.
    • Added: June 13th, 2007: Confirmation from bingo13/Gary from Anandtech here, black dimm slots are designed to be optimised for higher memory clocks/low latency clocks compared to yellow dimm slots. I found this out went revisiting Corsair 10000C5D Dominators in black dimm slots here.
    • Heads up for folks with PS/2 mouse, these new Asus P5K/P5K3 Deluxe boards removed the mouse PS/2 port in favour of USB mouse devices. So if you have a PS/2 mouse you'd either need to grab a USB mouse or PS/2 to USB adaptor http://www.auspcmarket.com.au/show_product_info.php?input[product_code]=CB-SKUSBPS. Unfortunately, for me that particular adaptor didn't work with my KVM switch mouse PS/2 connector cable so a USB mouse was needed.
    • Recovery from failed overclocks seems to be much better on Asus P5K Deluxe. Just turning off from power switch and powering psu off, wait a bit and power psu back on and power up would initiate the recovery mode - I see the bios detecting cpu at default 9x266FSB for my E6600 so i can F1 back into bios to set proper settings. No need for clear CMOS so far :)
    • Memory dividers available on Asus P5K Deluxe depend on the FSB Frequency set. Not all FSB Frequencies have all dividers available for them (well for 266FSB at least haven't tested the rest yet).
    • At 266FSB you have access to DDR2-667 (4:5), DDR2-800 (2:3), DDR2-889 (3:5), and DDR2-1067 (1:2). Photo
    • At 333FSB there's DDR2-667 (1:1), DDR2-800 (5:6), DDR2-833 (4:5), DDR2-1000 (2:3), DDR2-1067 (5:8), DDR2-1111 (3:5), and DDR2-1333 (1:2). Photo
    • At 400FSB there's DDR2-800 (1:1), DDR2-960 (5:6), DDR2-1000 (4:5), DDR2-1200 (2:3), DDR2-1280 (5:8), DDR2-1333 (3:5) and DDR2-1600 (1:2). Photo
    • At 500FSB there's DDR2-1000 (1:1), DDR2-1200 (5:6), DDR2-1250 (4:5), DDR2-1500 (2:3), DDR2-1600 (5:8), DDR2-1667 (3:5) and DDR2-2000 (1:2). Photo
    • Seems for at least 1:1, 4:5 and 2:3 dividers AUTO subtimings are set pretty tight, performing in Super Pi times just behind manually set subtimings of 3-30-8-3-6.
    • CPU Voltage Damper option in bios reduces vcore droop by as much as 75%. From approximately 0.032v droop to 0.008v droop! Finally, Asus answers our prayers in regards to vcore droop :)
    • Transaction Booster bios option when enabled, seems to boost memtest86+ v1.70 reported memory bandwidth. By default Transaction Booster is set to AUTO, but when Enabled or Disabled a sub menu appears in the form of levels. When enabled, Transaction Booster gives you an option to select 2 Boost levels of 0 or 1. When disabled, Transaction Booster gives you an option to select 3 different Relaxed levels between 0-3 - level 0 at least looks the same as AUTO set Transaction Booster in memtest86+ v1.70 bandwidth.
    • Interesting to note that for tests at 9x444FSB 1:1 4-4-4-4 2-20-2-2-2 at 2.1v vdimm that Transaction Booster AUTO and TRFC = 20, memtest reports 5383MB/s memory bandwidth compared with TRFC = 25 or 30 at 5201MB/s. But at TRFC = 20 if you enable Transaction Booster with Boost level 1, memtest still reports 5383MB/s. Could it be Transaction Booster Enabled just manipulates a tighter TRFC or other subtimings values ? Without memset for windows we won't be able to see what the subtimings are for real for now though.
    • The Stilt has figured out the strap changes for P5K Deluxe/P35 chipsets http://www.xtremesystems.org/forums/showpost.php?p=2243939&postcount=17





    .
     
    Last edited: Jun 13, 2007
  2. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    Stock

    E6600 B1 ES @2400Mhz - 9x266FSB stock AUTO vcore​


    System
    • E6600 QPGE B1 Step 5
    • Corsair Hydrocool 200EX 1/4" OD tubing
    • Asus P5K Deluxe 0304 bios
    • 128MB Gainward FX5200 PCI
    • 2x1GB Super Talent T1000UB1G5 Singles Micron D9GKX
    • 74GB WD Raptor 8MB on SATA 1 port
    • Pioneer DVR-08 Burner
    • Sony FDD
    • 1KW PCP&C SLI psu
    • WinXP Pro SP2

    Cpuz Validation:
    [​IMG]

    Idle:
    [​IMG]

    [​IMG]

    Load:
    [​IMG]

    Post Load Idle:
    [​IMG]
     
  3. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    CPU Voltage Damper Option​


    The Asus CPU Voltage Damper option in bios when enabled reduces vcore droop under load by alot it seems - ~75% reduction in vcore droop under load when enabled! 1.425v bios set in windows AI Suite reports 1.392-1.400v idle and 1.368v drooped under load by 0.032v. But with CPU Voltage Damper enabled, AI Suite reports 1.4080v idle vcore and 1.400v load vcore - only 0.008v vcore droop! Other bios set vcore options have similar reductions to only 0.008v vcore droop under load when CPU Voltage Damper is enabled :)

    CPU Voltage Damper disabled @1.425v bios set vcore = 2.2857% vcore droop under orthos small FFTs load
    CPU Voltage Damper enabled @1.425v bios set vcore = 0.5682% vcore droop under orthos small FFTs load

    [​IMG]
     
  4. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    E6600 B1 ES @3600Mhz - 9x400FSB at 1.416v vcore​


    Okay to my everyday stable known overclock for my E6600 B1 ES cpu @3600Mhz. The P5K Deluxe handled my E6600 B1 ES much like my P5B Deluxe did needing slightly more vcore (1.416v loaded) to run Orthos stable compared to E6600 B1 ES on Asus P5W DH Deluxe motherboard (~1.38-1.39v vcore loaded).

    System
    • E6600 QPGE B1 Step 5
    • Corsair Hydrocool 200EX 1/4" OD tubing
    • Asus P5K Deluxe 0304 bios
    • 128MB Gainward FX5200 PCI
    • 2x1GB Super Talent T1000UB1G5 Singles Micron D9GKX
    • 74GB WD Raptor 8MB on SATA 1 port
    • Pioneer DVR-08 Burner
    • Sony FDD
    • 1KW PCP&C SLI psu
    • WinXP Pro SP2

    Coretemp temperatures are pretty much same as Asus P5W DH, P5B Deluxe, and Intel D975XBX for 3600Mhz at 1.38-1.42v vcore Orthos loaded around 60-65C in 28.8C to 29.3C room ambient temps.

    Load temps:
    [​IMG]

    Post Load idle temps:
    [​IMG]
     
    Last edited: May 13, 2007
  5. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia

    Asus P5K Deluxe Timings/Subtimings
    1:1 Memory Divider


    I went back to look at my usual P5B Deluxe advance timings and decided to see how they relate to what P5K Deluxe offers so as to see what kind of advance timings I can start off testing with. After comparing them, I figured I'd start with same advance timing defaults as P5B Deluxe.

    For P5K Deluxe:
    * The comments in brackets next to values are how P5B Deluxe behaved with Micron D9xxx memory.

    I'll use Super Pi v1.50 to showcase how much difference adjusting memory timings and subtimings has on bandwidth intensive applications performance.

    Important to note, when looking at the Super Pi times, is that using E6600 B1 stepping ES cpu the Super Pi times are approximately 10-22 seconds slower clock for clock than retail Core 2 Duo B2 stepping cpus. I'll swap to a retail B2 stepping cpu after I used my E6600 B1 ES cpu to figure out the memory clock/timings behavioural characteristics of the Asus P5K Deluxe.

    I'll start off with all sub timings being set to AUTO:

    Subtimings all set to AUTO:

    [​IMG]

    Then we'll tighten just tRAS to 4 but leave everything else the same (largesystemcache disabled in Windows XP Pro SP2):

    [​IMG]

    Then tighten subtimings to P5B Deluxe default values (largesystemcache disabled in Windows XP Pro SP2):

    As you can see from the slower 32M time, the AUTO set advance timings must be alot tighter than 6-42-10-10-10-10. Maybe Asus P5K Deluxe AUTO subtimings are picking up my Super Talent T1000UB1G5 memories advance subtimings set in SPD - see Everest Ultimate memory SPD settings here.

    [​IMG]

    Then tighten subtimings even further (largesystemcache disabled in Windows XP Pro SP2):

    Faster than subtimings set at AUTO (largesystemcache disabled in Windows XP Pro SP2):

    [​IMG]

    How tight can we go still at 2.1v vdimm (largesystemcache disabled in Windows XP Pro SP2):

    [​IMG]

    Easier chart comparison:

    [​IMG]

    Important to note, when looking at the Super Pi times, is that using E6600 B1 stepping ES cpu the Super Pi times are approximately 10-22 seconds slower clock for clock than retail Core 2 Duo B2 stepping cpus.

    Edit: It seems I had largesystemcache disabled in Windows XP Pro SP2 which might account for the slower times :o I'll rerun the last configuration of 4-4-4-4 3-30-3-3-3 again with largesystemcache enabled in Windows XP Pro SP2.

    Largesystemcache enabled time is ~25 seconds faster:

    [​IMG]

    From the above Super Pi 1:1 memory divider testing, we see that AUTO subtimings set are at least faster than 6-42-10-10-10-10 (AUTO subtimings 0.955% faster) but slightly slower than manually set 3-30-8-3-6 (AUTO subtimings 0.378% slower) and 3-30-3-3-3 (AUTO subtimings 1.098% slower). I'm guessing that AUTO subtimings' values are a bit looser but close to 3-30-8-3-6 manually set values.
     
  6. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia

    Asus P5K Deluxe Timings/Subtimings
    4:5 Memory Divider


    Next up is looking at how 4:5 memory divider behaves on Asus P5K Deluxe using 0304 bios. First feature I noticed was that changing memory dividers also triggers off the the auto shutdown and auto restart process on only the bios exit in which you change the memory dividers. This is normal in order to properly set up the settings and only happens on the first change of memory dividers at exiting bios. Second item I noticed was there might be a bug in that PCI-E Frequency isn't locking properly when set to 100Mhz on first boot up after changing to 4:5 memory divider. Setting PCI-E Frequency to 102Mhz fixed the problem completely though. I've passed this finding onto Asus folks so let's see what they can do for the next bios release.

    For 4:5 memory divider testing right now, I just wanted to find out what the bios set's for AUTO for the subtimings as memset doesn't support P35 Bearlake chipset as yet there's no way to know from within windows what AUTO values are used for subtimings.

    I decided to run 2 sets of tests, one at AUTO subtimings and one at 3-30-3-3-3 manually set subtimings.

    At AUTO Subtimings (largesystemcache enabled from now on):

    [​IMG]

    Manually set subtimings (largesystemcache enabled from now on):

    [​IMG]

    Important to note, when looking at the Super Pi times, is that using E6600 B1 stepping ES cpu the Super Pi times are approximately 10-22 seconds slower clock for clock than retail Core 2 Duo B2 stepping cpus. However, even for B1 stepping cpu it seems the 32M times are off a bit from past B1 times at 4:5 on other boards, I'll have to pluck in a B2 stepping cpu latter to see what's happening.

    [​IMG]

    Again a similar pattern emerges as with 1:1 memory divider test in that manually set 3-30-3-3-3 sub timings is only 0.437% faster than AUTO subtimings configuration. In other words, AUTO subtimings values even on 4:5 divider are set pretty tight to begin with. This maybe good if you have a nice pair of Micron D9xxx based DDR2 memory with adequate vdimm memory voltage but might be too tight for other value memory.

    I'm not sure if AUTO subtiming values are set by the memory's SPD programming ? If that was the case, then how fast or slow AUTO subtimings are would be dependent on your specific DDR2 memory modules' SPD programming ??

    For value ram 4:5 divider mode, maybe it might be easier to start off with manually set subtiming values to 6-42-10-10-10 since it's alot looser and work your way down to tighter values ?
     
  7. chainbolt

    chainbolt Member

    Joined:
    Jun 27, 2001
    Messages:
    103,014
    Location:
    Tokyo
    For the lazy one: What's your conclusion? Is it performance or overclocking wise so much better that a change from 965/975 is a must?
     
  8. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia

    Asus P5K Deluxe Timings/Subtimings
    2:3 Memory Divider


    Some observations from 2:3 memory divider testing:

    • In general this Super Talent memory doesn't like CAS4 much without alot of vdimm voltage. The same occured on Asus P5B Deluxe and Asus 680i Striker Extreme.
    • But flies at CAS5. I tried 4-4-4-4 at higher than 525mhz memory at 2.35v with AUTO subtimings and got some errors in memtest86+ v1.70.
    • Then I tried subtimings manually set to 6-42-10-10-10 and got alot less errors so I suspect AUTO subtimings are too tight but without memset in windows to see what's being set, decided to use 5-4-4-4 main timings and had no problems at all. So higher 2:3 memory divider tests are being done with 5-4-4-4 main timings.
    • Further proof of tight AUTO subtimings, in 2:3 memory divider @570Mhz 5-4-4-4 at 2.4v with AUTO subtimings there was alot of memtest errors. But at manual set subtimings 6-42-10-10-10 only 6 errors. But at @570Mhz 5-5-5-15 at 2.35v vdimm AUTO subtimings set the memtest bandwidth was same as @5-4-4-4 with 6-42-10-10-10 subtimings. So AUTO subtimings set with 5-5-5-15 main timings made up for the looser set main timings.
    • At 600mhz 5-5-5-15 it with AUTO subtimings the Super Talent memory only needed 2.35v vdimm which is the same as when I tested the memory on Asus P5B Deluxe with 1101 bios.

    Like the 4:5 memory divider tests, I'll run 2 sets of tests, one at AUTO subtimings and one at 3-30-3-3-3 manually set subtimings.

    At AUTO Subtimings (largesystemcache enabled from now on):

    [​IMG]

    Manually set subtimings (largesystemcache enabled from now on):

    [​IMG]

    Important to note, when looking at the Super Pi times, is that using E6600 B1 stepping ES cpu the Super Pi times are approximately 10-22 seconds slower clock for clock than retail Core 2 Duo B2 stepping cpus.

    [​IMG]

    Again the same pattern as with 1:1 and 4:5 memory dividers in that AUTO subtiming values in 2:3 divider mode are already set pretty tight out of the box. Auto subtiming values are only 0.251% slower than manually set subtimings of 3-30-3-3-3.
     
  9. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    Just doing some quick screenies at 7x501FSB 1:1 and also checking out Super Talent PC2-8000 memory with the new 5:8 divider available on Asus P5K Deluxe @614mhz 5-5-5-5 at 2.45v

    7x501FSB 1:1
    [​IMG]

    5:8 divider @614mhz 5-5-5-5 at 2.45v
    [​IMG]

    That's 9mhz better than on Asus P5B Deluxe with same memory @605mhz 5-5-5-15 6-42-11-10-10-10 at 2.45v IIRC!

    Important to note, when looking at the Super Pi times, is that using E6600 B1 stepping ES cpu the Super Pi times are approximately 10-22 seconds slower clock for clock than retail Core 2 Duo B2 stepping cpus.
     
  10. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    Asus P5K Deluxe Transaction Booster​


    Next feature to look in bios is Transaction Booster. This function seems to boost memtest86+ v1.70 reported memory bandwidth. By default Transaction Booster is set to AUTO, but when Enabled or Disabled a sub menu appears in the form of levels. When enabled, Transaction Booster gives you an option to select 2 Boost levels of 0 or 1. When disabled, Transaction Booster gives you an option to select 3 different Relaxed levels between 0-3 - level 0 at least looks the same as AUTO set Transaction Booster in memtest86+ v1.70 bandwidth.

    [​IMG] [​IMG]
    [​IMG] [​IMG]

    Using 9x400FSB 1:1 4-4-4-4 settings below:

    Transaction Booster Disabled - Relax level 0

    [​IMG]

    Transaction Booster Enabled - Boost level 1

    [​IMG]
     
  11. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    Now how Transaction Booster AUTO vs Enabled Boost level 1 compare for Super Pi 32M

    AUTO set = 14m 53.813s
    [​IMG]

    Enabled Boost Level 1 set = 14m 45.531s
    [​IMG]

    Transaction Booster Enabled Level 1 is 0.915% or 8.182 seconds faster than when set to AUTO!

    [​IMG]
     
  12. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    Intel Core 2 Quad Q6600 Kentsfield
    +
    Corsair Nautilus 500 H20 cooled 3/8" ​


    Alot of folks have asked what about kenstfield quad core cpus on Asus P5K Deluxe ? So decided to swap out my E6600 B1 ES cpu for my Q6600 QUPT B3 ES cpu.

    But I know that the external water cooling unit Corsair Hydrocool 200EX with 1/4" OD tubing I'm currently using won't be able to handle Kentsfield quad core. So changed the cpu cooling to a Corsair Nautilus 500 3/8" ID tubing based external water cooling unit which was kindly supplied by the folks at Corsair :)

    [​IMG] [​IMG] [​IMG] [​IMG] [​IMG]

    Corsair Nautilus 500 coolant vs Corsair Hydrocool 200EX coolant

    [​IMG]

    To illustrate the tubing sizes from left to right:
    Corsair Nautilus 500 3/8" ID vs Corsair Hydrocool 200EX 1/4" OD vs Clearflex 3/8" ID vs Clearflex 1/2" ID tubing

    [​IMG]



    Let's compare stock Q6600 temperatures on P5K Deluxe with Corsair Nautilus 500 3/8" tubing with Q6600 on P5B Deluxe with Swiftech G4 Storm/Thermochill HE120.E/MCP650 1/2" tubing setups.

    Q6600 stock clock/vcore with P5B Deluxe
    Swiftech G4 Storm + Thermochill HE120.3 (3x120x38mm Panaflo 103cfm) + MCP650 ​


    Room temps = 21.0C - 24.5C
    Idle temps = 22.0C - 27.0C
    Load temps = 37C - 43C
    Load room delta temps = 16C - 22C
    Load cpu load - cpu idle delta temps = 15-21C

    Idle:
    [​IMG]

    Load:
    [​IMG]

    vs


    Q6600 stock clock/vcore with P5K Deluxe
    Corsair Nautilus 500 with fan set on high speed ~1800rpm​


    Room temps = 27.5C - 28.5C
    Idle temps = 43.0C - 47.0C
    Load temps = 56C - 62C
    Load room delta temps = 28.5C - 34.5C
    Load cpu load - cpu idle delta temps = 13C - 19C

    Idle:
    [​IMG]

    Load:
    [​IMG]

    Looks like Nautilus 500 is behind my custom DIY - Swiftech G4 Storm setup by at least 12C. This is as expected considering the huge difference in ambient room temperatures, size of radiator and fan setups. Swiftech G4 Storm setup has a triple fan radiation Thermochill HE120.3 and 3x 120x38mm thick Panaflo 103cfm fans versus Corsair Nautilus 500's single low cfm fan and much smaller radiator. To illustrate the difference here's Q6600 @3500Mhz at 1.472v idle/1.408v load on P5B Deluxe using Swiftech G4 Storm setup with 25-27C room ambients resulted in 31-37C idle and 60-68C load temps.




    Q6600 Kentsfield @3600Mhz
    +
    Corsair Nautilus 500 H20 cooled 3/8" ​


    My Q6600 B3 ES Kentsfield quad core cpu 24/7 stable clock on Asus P5B Deluxe is 3600Mhz at 1.456v with Swiftech G4 Storm/Thermochill HE120.3/MCP650 with 60-68C coretemp reported load temps. From above comparison between the Storm/Thermochill HE120.3/MCP650 vs Corsair Nautilus we can see that at 3600Mhz temps will be alot higher on the Corsair Nautilus. How high ? Let's find out :)

    Q6600 @3600Mhz - 9x400FSB 1:1 at 1.475v bios set vcore with CPU Voltage Damper enabled gave me windows idle temps of

    Idle @1.475v bios set

    [​IMG]

    However, the 2nd core = core1 failed quad Prime95 v25.1 smallFFTs load testing even at 1.475v or 1.4875v.

    Idle @1.4875v bios set

    [​IMG]

    Load @1.4875v bios set

    [​IMG]

    Failed core1 @1.4875v bios set

    [​IMG]

    Post Load Idle @1.4875v bios set

    [​IMG]

    Roasty at upto 86C coretemp reported load temps! Might have to try remounting Corsair Nautilus 500 waterblock a few times to make sure it's making proper contact :)
     
    Last edited: May 14, 2007
  13. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    Still @3600Mhz at 1.4875v bios set vcore, it manages to complete Systool Quad 16M Pi benchmark fine though. This is a good indicator of being able to complete Super Pi 32M at least even though not Quad Prime95 v25.1 small FFTs stable :)

    Systool Quad 16M Pi Halfway mark
    [​IMG]

    Systool Quad 16M Pi Completed
    [​IMG]



    Q6600 B3 ES @3600Mhz - 9x400FSB 1:1​


    While Q6600 @3600Mhz at 1.475v bios set wasn't prime95 stable, it still managed to complete single, dual and quad 32M probably due to the lower heat output compared with prime95 smallFFTs.

    System
    • Q6600 QUPT B3 Step 7
    • Corsair Nautilus 500 3/8" ID
    • Asus P5K Deluxe 0304 bios
    • 128MB Gainward FX5200 PCI
    • 2x1GB Super Talent T1000UB1G5 Singles Micron D9GKX
    • 74GB WD Raptor 8MB on SATA 1 port
    • Pioneer DVR-08 Burner
    • Sony FDD
    • 1KW PCP&C SLI psu
    • WinXP Pro SP2

    Super Pi v1.50 Single instances​


    1M
    [​IMG]

    8M
    [​IMG]

    32M
    [​IMG]

    Super Pi v1.50 Dual instances​


    8M
    [​IMG]

    32M
    [​IMG]

    Super Pi v1.50 Quad instances​


    8M
    [​IMG]

    32M
    [​IMG]


    Quick bench before I sleep :)

    Highest Q6600 FSB I've had to date thanks to Asus P5K Deluxe! (will try higher FSB later) :)

    [​IMG]

    [​IMG]


    Q6600 B3 ES - Max FSB​


    From above quick 8x450FSB test you can see Asus P5K Deluxe has much better Kentsfield Quad core FSB headroom. I've seen other P5K Deluxe folks with QX6700 @480-500+ FSB. So now to see what my Q6600 B3 ES can do FSB wise. :)

    I usually start off with Systool V1.0 728 build Quad 16M Pi test (uses 1.3GB of memory - the Quad 32M test uses 2.6+ GB of memory) to test max FSB for my Q6600. If it fails it usually aborts the test or reboots the system.

    With below settings max FSB was around 7x470FSB

    Change CPU PLL Voltage from 1.50v to 1.60v allowed me to boot into windows 7x480FSB otherwise with 1.50v would BSOD on entering windows. Upping NB Volts from 1.4v to 1.55v allowed 7x480FSB

    Upping Vcore to 1.4375v allowed 7x485FSB

    7x490FSB is turning out to be tricky to achieve, Quad Systool V1.0 728 build 16M Pi benchmark will reboot the system even quicker if you raise vcore from 1.4375v to 1.45v. Similar thing happened on P5B Deluxe but at 400-425FSB ceiling when vcore past a certain point reduced max stable FSB.

    Q6600 @3395Mhz - 7x485FSB​


    Here's where I ended up for single and dual Super Pi runs at 7x485FSB 1:1 )

    [​IMG]

    Single Super Pi 32M
    [​IMG]

    Dual Super Pi 32M
    [​IMG]


    .
     
    Last edited: May 15, 2007
  14. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    Q6600 B3 ES @3800Mhz - 8x475FSB​


    Amazing indeed with Asus P5K Deluxe, I've pushed my Q6600 to a new high - being able to run single and dual Super Pi 32M @3800Mhz - 8x475FSB ! :eek:

    [​IMG]

    Single Super Pi 32M
    [​IMG]

    Dual Super Pi 32M
    [​IMG]

    Looks like Corsair Nautilus 500 can handle it :)
     
    Last edited: May 15, 2007
  15. chainbolt

    chainbolt Member

    Joined:
    Jun 27, 2001
    Messages:
    103,014
    Location:
    Tokyo
    ROFL .... Eva you are too fast.
     
  16. H[]R3NDUS

    H[]R3NDUS Member

    Joined:
    Dec 2, 2005
    Messages:
    962
    Great stuFF

    Wow Eva you have to much time on your hands ..Thats bloody awsum work M8.:thumbup:

    How does the board feel fast compared to p5b series ?? or much the same ??

    Cheers Hori
     
  17. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    :lol: would of been faster if not for the 8 second set vB flood control on the forums :D

    P5K Deluxe can end up faster in memory bandwidth intensive tasks than 965P by virtue of the new dividers available 5:6 and 5:8 opens up alot more FSB/MEM combinations :)

    So far i can say, P5K Deluxe is all the good features of Asus P5W DH, Asus P5B Deluxe, Asus 680i Striker extreme, Asus Commado and Asus P5W64 WS Pro rolled into one :) Asus took alot of feedback into consideration when they designed P5K Deluxe it seems (some suggestions i put forth :D )

    Pros:
    1. CPU Voltage Damper fixes the old vcore droop issue!
    2. High memory clocking like P5B Deluxe/Commando with upto 2.55v vdimm max :)
    3. High Kentsfield quad core fsb - folks been able to get their QX6700 to 480-500FSB compared with P5B Deluxe 400FSB or 450-480fsb with volt mods or 440-480FSB on Asus Commando :)
    4. Very good failed oc recovery like Asus 680i Striker extreme, less need for clearing CMOS on failed overclocks :D
    5. Cleaner cpu socket area for phase change or extreme cooling users like Asus Commando (my Commando just arrived on Friday so another adventure starting soon - this is for phase change :) )
    6. Stock auto performance is very fast on par with moderately tweaked P5B Deluxe in terms of memory bandwidth/super pi. Handles tighter subtimings for memory than P5B Deluxe but performance similar - so maybe loosened up a bit by default for P5K Deluxe.

    Cons:
    1. Lack of PS/2 mouse port - for KVM switch PS/2 users like me it mights it a bit more complicated to use. I use setup a spare USB mouse next to my PS/2 mouse and by passed my KVM switch PS/2 board but kept KVM PS/2 keyboard/monitor plugged in.
    2. No power/reset/clear cmos buttons like the Striker Extreme boards
    3. Bios chip it isn't replaceable but soldered/fixed on the board it seems. See page 2-3 of manual for the board diagram here. The bios chip is supposedly located at the centre point in between the last PCI3 and PCI-E 2 slot and the red IDE connector location. Directly in alignment with the battery. That would make the bios chip to be the soldered on square chip to the right of the 7 pin header pictured here ? But it has Crash Free 3 bios feature see page 4-7 of manual. Shove the included driver dvd disk the system and boot it up and it should attempt to do a recovery bios flash.
    4. Added May 17th, 2007: Folks over at hardforum.com have highlighted some possible layout concerns in that if 2x double slot video cards are used in the 2x PCI-E video card slots, it will render 2 out of 3 of the PCI slots unusable and the location of the front USB/Firewire connectors and IDE connector.

    But those are preliminary conclusions.. only had the Asus P5K Deluxe and P5K3 Deluxe board for 4 1/2 days so far :)

    [​IMG]

    No DDR3 memory yet for P5K3 Deluxe, so concentrating on Asus P5K Deluxe and Asus Commando for now
     
    Last edited: May 17, 2007
  18. RodneyJM

    RodneyJM (Banned or Deleted)

    Joined:
    Mar 20, 2005
    Messages:
    1,916
    Location:
    Engaged in Silicon Wars !
    Good stuff eva2000 as usual, if this is the result from P35 series, what X38 going to be like? :eek: :)
     
  19. OP
    OP
    eva2000

    eva2000 DDR1/DDR2/DDR3 Addict

    Joined:
    Jun 27, 2001
    Messages:
    21,902
    Location:
    Brisbane, Australia
    Yeah if P5K Deluxe is anything to go by, the X38 will be very :cool: :)
     
  20. Tasty_Vito

    Tasty_Vito Member

    Joined:
    Sep 9, 2003
    Messages:
    957
    Location:
    Tasmania

Share This Page