Jump to content
HWBOT Community Forums


  • Posts

  • Joined

  • Last visited

  • Days Won


Posts posted by Tzk

  1. If two categories are introduced for AL, then what happens with benchmarks which won't scale past 16 Threads (so 8 big cores + SMT)? Could a user just submit these benchmarks twice and get points for both? A good example is 32M, as it doesn't scale at all. Could i submit it to 8+8 and 8+0 category? 

    I also think that having to disable cores to participate in a certain benchmark will cause issues and faulty submissions. So we need to have a ranking for "all cores enabled" to simply have the users submit into the right category without thinking too much. I'm also not sure if a separate mixed category will help us at all...

    So i vote "AL = 16C" and done.

    • Like 1
  2. 7 hours ago, TerraRaptor said:

    So many good scores lately thanks to all the work you and other guys (Tzk, digitalbath etc) did.

    All this wouldn't be possible without the overclocking elders (OskarWu, Tictac, Polygon) which did the groundwork for this. Also the openness of recent socket 462 overclocking and sharing all findings and infos was key. So a true team effort :) And let's not forget about the new NF2 Xtreme tweaker made by @I.nfraR.ed. His work makes this a lot easier to clock NF2 and set timings in Win.

    I also believe there is more to gain, especially on 1:1 dualchannel. I'm still stuck at 263Mhz for 32M runs.

    • Like 4
  3. On 9/1/2021 at 1:17 PM, TASOS said:

    Any chance you can test with 256mb of D43 chips ?

    Been there, done that. 256mb BT-D43 1:1:


    256mb DT-D43 and BT-D43 run exceptionally well on the A7N8X and NF2 in general. My best stick of BT-D43 is mostly done at 273Mhz, though. DT-D43 doesn't clock above ~255Mhz 32M. BT-D43 runs 263Mhz 32M in Dualchannel which probably is the chipset limit of my board.

    • Thanks 1
  4. @exaberries

    The board is fully recapped with poly caps. I used 2x Panasonic FR 3300uF 16V for the 12V rail, 5x 2700uF 4V Kemet polys for the cpu and a bunch of other caps for the whole board. The full list including the positions is here:


    I also soldered an additional 4pin like Terraraptor did plus i run a wire from the 12v pin on the 20pin connector to the cpu VRM.  Here's how the board currently looks like:



  5. Bump, digging with a golden shovel after 9 years...

    Someone on the HWBot discord mentioned this neat tool. Sadly the issue with winxp is still persistent... Is there any hope that either @W1zzard fixes this or maybe the source is released so someone else can have a look at fixing this? 


    Reading the Microsoft docs on this, it seems that compiling it against PSAPI v1 instead of v2 should be enough?


    • Thanks 1
    • Haha 1
  6. On 7/24/2021 at 12:38 AM, Mr.Scott said:

    AN7 clocks better on a 6:6 ratio

    NF7 clocks better on a 5:5 ratio

    All my DFI's seem to clock better at 1:1

    😲 Seriously? That's the one thing i never tried in all those years... I usually just kept what the board used as default for synced FSB:Dram. Might be interesting to dump the chipset registers and to see if there are differences.

    • Like 1
  7. So i tested a few more cpus. My current conclusions about BCRC are:

    • if the cpu has a minor revision of 2 or 11, then it is reported as multi locked
    • if the cpu has a minor revision of 12 or 15, then it is reported as multi unlocked
    • minor revision 11 is reported as Thorton w/ 512K, all other minors are reported as Barton
    • Stepping is just a guess and mostly AQYFA or unknown

    Now, i added Antinomy's cpus to my list and sorted them by minor revision and week.

    • Revision 2 has two ranges with a gap in them: 0337 - 0349 (?) and 0405 - 0411 (locked)
    • Revision 11 was made from 0342 to 0512 (locked)
    • Revision 12 also has two ranges with gap: 0327 - 0332 (unlocked) and 0403 - 0407 (locked)
    • Revision 15 seems to be the oldest: 0318 - 0319 (unlocked)

    So i believe the timeline of the revisions is:

    15 -> 12 -> 2 -> 12/2 -> 11

    On german Hardwareluxx forums there is a user which states that AMD changed something inside the cpu die which makes them clock a lot better. That's why the mobiles from 0351 onwards clock so much better than older Bartons. I believe that it's the change to minor revision 11 which is to "blame" for this, although we need more cpus/data to prove this.

    We also need more cpus to see of the gap in revision 12 and 2 is real...

  8. On 7/2/2021 at 3:12 PM, I.nfraR.ed said:

    I've added a message box to the tweaker on app open that displays the info I currently have implemented.

    Just a thought: Wouldn't it be better if we could open that info box with a button inside the tweaker? So we don't have to close it on every tweaker launch if we want to "just" change timings. Also the frequency and FSB display is currently broken on A7N8X, right? Tweaker reports 151Mhz and Multi 11 while Cpu-Z shows the actual values (133 + 12.5x).

    Besides that the MSR values between BCRC and Tweaker match, well done!

  9. On 7/2/2021 at 3:12 PM, I.nfraR.ed said:

    There's one more value, which is "Reticle Site" and I've included it in the info. Curious to see what different cpus show.

    Perfect. So now we can use BCRC and your tweaker to read the values. Will retest the above mentioned cpus.


    @Antinomy A question regarding the forums and its editor: Is there a possibility to install a plugin for the WYSIWYG editor to include tables? The table above was just copy&paste from excel, but i can't change the table width... Not sure how you did it in your post.

  10. @Antinomy and @I.nfraR.ed shared their thoughts about superlocked cpus and cpu registers (MSR). Maybe we can identify superlocked cpus like this... They also mentioned a tool called Barton CRC (BCRC.exe). So i grabbed a few cpus and here's the result so far. 

    Example screenshot of BCRC output:


    Here's the result which BCRC reported. The first columns is what i read from the cpu sticker (orange), latter ones are output of BCRC (blue). My conclusion so far is that the minor rev is rather random and the superlock plus the Crystal marker seems to be a more or less random guess.  

    Nr Cpu Rating Stepping week Stepping ED Value CPUID Processor Code Major Minor Type Crystal Marker
    11 Athlon XP 2600 AQZFA 0349 TPMW 2306h 06A0h Barton 0 2 locked AQYFA
    12 Athlon XP 2600 AQXDA 0318 MPMW 2319h 06A0h Barton 0 15 unlocked unknown
    13 Athlon XP 2600 AQXEA 0404 TPMW 2316h 06A0h Barton 0 12 unlocked unknown
    15 Athlon XP 2600 AQYHA 0401 UPMW 2315h 06A0h Thorton w/512K 0 11 locked AQYFA
    16 Athlon XP 2600 AQXEA 0403 WPMW 2316h 06A0h Barton 0 12 unlocked unknown
    17 Athlon XP 2800 AQYHA 0409 SPMW 2315h 06A0h Thorton w/512K 0 11 locked AQYFA
    • Like 4
  11. On 6/23/2021 at 4:41 AM, Antinomy said:

    So these do look like CPU settings. I believe some of them are delays that are programmed during SIP packet.

    Might be lame, but I didn't see this info before.

    When i started looking into romsip modding i found this post by TicTac on pcper forums. See attached screenshot for content. So yes, these are indeed multi specific settings... This made me test my FSB stability at multi 7 or 7.5. If it is stable on these multis, then it'll run on any multi. However he didn't state what these settings are actually controlling. If we can find this out (also for the upper half of the romsips), then we might be able to push the fsb further.

    We also hit a hard FSB wall at about 263MHz right now, on the german Hardwareluxx Forums we got about 4 or 5 boards which won't pass 32M above 263Mhz, no matter what you do. Vdd, Vcore, Multi doesn't matter, some boards do even freeze when trying to set 264/265 MHz. No clue what causes this, might be some peripheral controller or even the chipset itself acting up. 

    broken link: https://pcper.com/forums/?346001-ROMSIP-Table-Mod-Guide#post3100010


    • Like 3
  12. On 6/23/2021 at 4:19 AM, Antinomy said:

    options ROMs with performance tweaks (3D-fire for example - modified chipset registers?)

    Yes. You basically use the PCI registers of the chipset and set the desired settings via ISA or PCI option rom, right before the OS is loaded. There's two ways of doing this:

    a) you can just hardcode some settings

    b) you can write some data from modded bios settings to the CMOS register (bios settings are stored here) and load these cmos values from the option rom. This way you can add additional bios settings without hacking the whole bios. We did this on at least 3 or 4 boards (a7n8x-E, a7n8x v2.0, Abit, Epox?) and it works great. 

    the cleanest solution would be to reverse engineer the bios and add those options in a native way. However i've already spend a bunch of hours looking into this on my A7N8X and even comparing different bios versions hasn't got me anywhere. Luckily Asus made 2 bios versions where each introduced a new bios setting, so the changes inside the code are clearly visible. However it looks like the hex strings which are used to define the bios items and the code which is controlled by them are stored right next to each other. So if there's an option added, then some offsets and pointers will break and the bios won't work anymore. 

    My conclusion was that if we ever want to add completely new options into the bios which don't "recycle" unused bios items, then it's a ton of work. That's the point where i gave up... It's probably better to spend more time upon improving and understanding the romsips first. 

    • Like 3
  13. I'd like to add that there's actually two types of nvidia specific code for the ram controller inside the bios, depending on the board and (maybe?) chipset revision. We got NVDAMC and NVMM. NVDAMC is the memory controller firmware up to version 3.19. Nvidia then renamed it to NVMM and introduced version 4.x. Note that these versions are not compatible with each other. So if a board runs on 3.x (examples are Abit NF7, DFI Infinity/LP B, A7N8X), then you can't update to 4.x. Board won't boot. Some newer MSI boards (K7N2?) run on NVMM 4.35 and obviously won't boot on 3.19. I see no reason to use an older version than 3.19. It's the latest available NVDAMC version before NVMM got introduced and runs great in terms of stability and performance. 

    Also note that the NVMM versions (ex: v4.62) extracted from Intel Nforce boards won't work on any socket 462 board. At least that was the conclusion when we tried it in the past.


    If you want to swap the BPL, you can do it with a regular hex editor as it isn't LZH compressed. Thus you don't need modbin to compress it and calculate the checksum.

    • Like 3
  14. @TerraRaptor

    Yes it is. Atlan1980 (member of hardwareluxx.de) reworked the whole Vdimm feedback loop and developed a Vtt mod for the Asus board. He introduced smaller resistors for the FB loop, swapped the mosfets and coils in the Vdimm regulator and removed a cap in the feedback loop which basically eliminated the Vdimm drop under load. We're now at <50mV drop under load at high vcore. I haven't tried it myself, but it looks like we can push much further now, even with TCCDs. @digitalbath managed to pass a 260Mhz 32M run with 2x512mb TCCDs :D 

    Link, have a look at the attached PDF on this post:


    • Like 1
  15. 12 hours ago, Sparks.nl said:

    Those Geil refuse 1t on nf2 and nf4

    I've run Geil Ultra-X (2x512mb TCCDs) at 250MHz 3-3-3 1T on NF2 through 32M ;) The A7N8X doesn't seem to like CL2.5 with Merlin Romsips... I've had 2.5-3-3  250MHz 24/7 stable on DFI Infinity years ago. It's achievable for sure :) Booting them above that isn't an issue, but getting it stable enoughfor 32M is... TCCDs are very picky and the NF2 just doesn't like them at all.


    • Like 1
  16. I got Rev. F TCCDs with production date of week 42 2005, so there were newer chips made. However my Geil Ultra-X (unknown date, probably 2004) are by far better. The Corsair 3200XL v1.2s can't even reach their spec on Socket A (250MHz 2.3-3-3) but fail to pass 240Mhz 3-3-3. That's disappointing für TCCDs. The Ultra-X at least reach 250Mhz 2.5-3-3 primestable. I guess the NF2 just isn't doing well with these Samsung chips.

    Same for 1gb Infineon CE-6 sticks and 512mb singlerank. Those fail to pass 210Mhz on my NF2 boards, regardless of voltage and timings. However 256mb Infineon BT-5 ( Corsair 3200C2PT v4.x) reach 250MHz 3-3-3-8 without issues.

    The Nforce is indeed quite picky when it comes to certain ram chips. Winbond works incedibly well, Hynix too. Haven't tried Micron or some other type of Infineon yet.


  • Create New...