S_A_V

Thanks! WBC enabled, buffer flushing disabled. I tried different versions of AS SSD - read access time always fails with 3x SSD (1x / 2x SSD - no problem). But it gives overall score anyway. Don't know why, maybe "division by zero" somewhere in the calculations

Video Encoding tweaks allowed for other versions of PCMark? It gives 2x boost (from 464 to 983 fps in my case) in PCMark04 WMV Video Compression subtest: Can I submit this score?

http://hwbot.org/submit/benchmark/as_ssd_benchmark/

Yes, RevoDrive was designed with internal RAID (4xR0 for X2, not 2x), but it performs closer to 1xSSD (up to 700-800 points) than 4xSSD (2000+ points) and it works as single PCI-E unit, that's why I posted it as single SSD. I don't see any AS SSD submissions rules for now and not sure what is better in this case. Try to ask moderators about it. If they consider RevoDrive must be posted in RAID category - I will move it to 4x.

WMV registry / PowerToy (including 32x32 encoding size) gives +4K to overall score. Internet Explorer tweaks gives another +1K...+2K (this is for Win7 ~45 pages/s , wait when Win8 compatibility issues will be fixed you will see ~60 pages/sec). Uncapped XP Startup - also up to +4K to overall score (at least for 2xSSD R0). So, if we allow PowerToy and IE Tweaks - it will be enough to make old scores easy to beat, even if we will keep XP Startup cap. That is why I think now if the best moment to remove the cap. Now or never. Both way new scores with new rules will be around 50K.

Storage system may looks not so informational in software. By default it may be something like "Array_001", "Array_002"... Besides of risk to crash extreme overclocked rig while run all this applications and windows for the screen, if we will require a lot of things include in screen - more people will make screens in huge desktop resolution, they will not fit in 300kb limit, so we will see a lot more highly downsized hardly-readable crap... I think better way is include a rule to always attach real picture of the system with storage used for the score. But not for all, only for those who scores high enough in HDD subtest (over 220 MB/s in XP Startup, for example). Or as alternative - picture only for top20, no matter how much in hdd subtests.

Keep PCMark05 and the Rules as it is. This is not the case. Current rules not compatible with current real moderation of this benchmark. Scores done with powertoy (wmv codec registry settings) blocked all the time, but powertoy still not added to the submission rules as forbidden tweak. Please add this to the rule or stop blocking such scores. Otherwise current rules looks confusing. The same with new browser tweak found by Gluvocio and his blocked score. I can confirm such high web page rendering can be done with only browser settings. There is no info in the rules that browser settings can not be changed. Keep PCMark05 but make new , improved up to date Rules, whats allowed and whats not. Yes, rules must be updated with list what is not allowed. Keep PCMark05 update the Rules as anything and everything goes and remove the 220xp start up cap. 220 MB/s XP Startup cap was stupid idea from the beginning. But if we remove it now - a lot of old hard earned scores will be easily beatable. Such action may looks disrespectful for "storage guys" who benched PCMark05 for ages with a lot of different hardware. Personally, I don't have so much PCMark05 scores, so removing the cap is ok for me. May be we need start separated vote about this topic - keep or remove cap? But if we decide to remove - it must be removed completely. Not changed to the new cap (300-400-500-1000 MB/s, etc). I got over 500 MB/s XP Startup with current generation SSD in RIAD0. Something like RevoDrive X3 or next gen SSD will raise it even higher. Software RamDisk gives over 3000 MB/s in every HDD-subtest, this is way more than current hardware storage systems, so it can be clearly visible in results. Of course, this numbers can be done with current generation hardware, not something ancient. On older platforms results will be proportionally smaller. Remove PCMark05 from hwbot. PCMark05 still support validation system (ORB). HIGHLY tweakable, same as SuperPI 32M and 3DMark01. It's not need 4xGTX580 for Top20 score. Two SATA3 SSD in R0 on integrated controller is enough for 40K+ score (I prove it already). It measures overall system performance, not only CPU or RAM or GPU. For me PCMark05 is still best version of all PCMarks. PCMark04 is buggy (grammar check), discontinued (no more official patches) and lack of support multicore CPUs. PCMark Vantage way too long (more LN2 waster than Unigine!), way less tweakable than PCMark05 and needs to install OS to storage that it measures. PCMark7 light versions of PCMark Vantage, almost the same. Less depends on storage subsystem performance than both PCMark05 and PCMark Vantage.

I list here all that I know can be used for better pcmark05 result. If you consider some of the following as cheat (or forbidden tweak that must not be allowed) - please add it to the new edition of pcmark05 submission rules: 1. Internal windows tools (regedit.exe, taskmgr.exe, msconfig.exe, services.msc) 2. Aero Tuner - small third-party application to tweak Windows Aero settings. 3. Intel RST driver and RAID options like stripe size and write-back cache. 4. Moving the mouse during the test 5. Changing desktop resolution (800x600, 1024x768, etc). 6. PowerToy (yes, it's nothing than front-end GUI for WMV-encoder registry keys): all allowed, all disallowed, or all allowed except "Encoding Width/Height" options. Only "Encoding Width/Height" options gives huge boost to Video Encoding subtest (8000-9000), all other options gives very small increase of result. 7. Easy accessible options from internet browser's menu (disable/remove plugins, pictures, html-formatting, scripts, etc). 8. Windows Shell (explorer.exe) and Internet browser (iexplore.exe) - advanced settings in registry: HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\Explorer\ HKEY_CURRENT_USER\Software\Microsoft\Windows\CurrentVersion\Explorer\ 9. Easily accessible video driver setting (Catalyst Control Panel, NVIDIA ForceWare Control Panel). 10. Advanced video driver setting (RivaTuner's Power User settings, video driver registry keys, etc.) 11. Usung GPU-accelerated rendering/encoding/decoding/etc (OpenCL, DirectCompute, CUDA, etc.). Unsure if it helps to the current versions of codec and browser, but this situation can be changed in any time - with new versions/generation of software/drivers/hardware. 12. Unreleased OS, alpha/beta builds leaked to public (Windows 8). 13. Changing PCMark storage partition on-the-fly during the test (subst, diskpart, self-made scripts and/or applications). Using different storages for different hdd subtest. For example, general usage on ACARDs with discrete controller (600+ MB/s) and in the same pcmark run - virus scan on SSD with integrated controller (1300+ MB/s virus scan). 14. Using third-party video/audio codecs and internet browsers (I mean anything not developed by Microsoft).

Improved IE9 settings... Thanks Gluvocio

Ok, I understand virus scan may depends on amount of memory used for write cache, not only raw storage performance. That's why I not report you score and not starting any thread about it. I listed it here only as example that not all scores are equal on subtests. Yes it's possible. I use Sapphire Radeon 6950 2Gb based on reference design unlocked to 1536 shaders and it gives the same performance in pixel shader subtest as Radeon 6970. Easiest way to unlock is flashing 6970 bios to 6950 but after that you can't raise voltage in any software (MSI Afterburner, Sapphire TriXX, etc.). Better way is mod bios of your card with special PHP-script you can download from here: http://www.mediafire.com/?6yifeiyqcdk1pk1 1. Read your current bios with ATIWinflash to file original.bin 2. Place original.bin to same folder with unlocking script and run the script (run.bat). modded bios will be created in the same folder with name modded.bin 3. Flash modded bios (modded.bin) to your card with ATIWinflash That may work with any 6950, including non-reference and 1gb models. And after this you still can raise voltage with afterburner.

thanks guys. memory access is gambled like car low in 3dmark01. sometimes it gives 32 but most time - 26-27. I set 8gb of ram in hope that intel RST driver will use more ram for write-back cache. not sure it helps or not. also I'm tired rebuild and manual TRIM'ing SSD's between EVERY pcmark run. benching ssd raid with pcmark and ln2 will be pain in the ass... that's why I still on water with this setup.

IE9 is default browser for now. 12 may 2011 Microsoft updates their distributives of Windows 7 with integrated SP1 and IE9.

It's ok for 68X0 on air. I got 2621 fps with 6850 @ 910/1210 (not published) and 3960 fps with 6870@1250/1200: http://hwbot.org/submission/2121515_ 6970 are better for pixel shader. Or highly clocked 5870. Yes, 5870 are better than 5850 flashed into 5870. Flashing 5850 into 5870 not helps to unlock more shaders. More shaders - better result in pixel shader test.

http://en.wikipedia.org/wiki/Subst

Really? If so, why PCMark05 submissions rules still not updated about PowerToy? http://hwbot.org/news/881_application_9_rules/

Good Transparent Windows run is 13к for Win7 (2600K@5600MHz) and 16.5k on Win8 (2600K@5300MHz). "Vista > Win8 > Win7" for Transparent Windows subtest but overall score is "Win8 > Win7 > Vista". Transparent Windows is not only one "tweakable" pcmark05 subtest. There is some other scores with "anomalies": Virus Scan: http://hwbot.org/submission/2130814_freakezoit_pcmark_2005_core_i7_2600k_37031_marks http://hwbot.org/submission/2129940_freakezoit_pcmark_2005_core_i7_2600k_36991_marks 2000+ MB/s with "only" two iRAM's? It's two scores with different settings, so it seems repeatable and may not bugged. But I don't see anyone close to this in whole top20... Web Page Rendering: most people got 21-22 pages/s with IE9 and Win7, but there is also scores with 26-32 pages/s. 26 pages/s (subzero): http://hwbot.org/submission/2186476_subzero_pcmark_2005_core_i7_2600k_42059_marks 32 pages/s (pro): http://hwbot.org/submission/2188240_pro_pcmark_2005_core_i7_2600k_43116_marks'>http://hwbot.org/submission/2188240_pro_pcmark_2005_core_i7_2600k_43116_marks I can get so high only with Win8, but not with Win7. Video Encoding: 1xxx MB/s without WMV PowerToy (or directly editing registry setting) and 2xxx MB/s with it. 3990 MB/s (pro): http://hwbot.org/submission/2188240_pro_pcmark_2005_core_i7_2600k_43116_marks 3D Pixel Shader: 3xxx untweaked and 4xxx tweaked. Already discussed here: http://www.hwbot.org/forum/showthread.php?t=26580&highlight=Pixel+Shader

Congrats for the win of K10 stage So what the secret of >6K grammar check runs? I tried affinity, parallel loading, lowering CPU multi... anything is working up to 6k for me and no more And why you don't use Windows 8? It gives faster web page rendering and video compression.

CPUZ can run on Windows 8 with ACPI disabled in ini file but shows wrong BCLK on Sandy (dont tried with other CPUs yet). That's why I used ThottleStop instead of CPUZ to see/show CPU clock. 3Dmark03-05-06 and PCMark05 can run on Windows 8 with -nosysteminfo key. Also you can try to add Windows 8 "support" to video drivers - just replace NTx86.6.1 to NTx86.6.2 (for x86) or NTamd64.6.1 to NTamd64.6.2 (for x64) in nv_disp.inf or CWxxxxx.inf/C7xxxxxx.inf files.

I always keep Grammar Check not higher than 6.0-6.1 Kb/s and it's not failing. Just slow down CPU with something (wprime for example) while running Grammar Check. There is some way to get Grammar Check over 6.0-6.1 but I don't know it. I seen some results with 7-12 Kb/s posted to HWBOT OC Challenge June 2011 by OC Forums and Classicplatforms.com members here: http://hwbot.org/competition/hoc_jun11/stage/224_pcmark04:_amd_k10/ Try ask them how it's done.

Some subtests needs all cores enabled for better result. No, it's not for performance. I added Throttlestop on screenshot for informational purposes, because CPUZ shows wrong BCLK in Windows 8.

Yes, аnd 90% of my water submissions done the same way. 58 pages/s in Web Page Rendering speaks for Win8 performance. Win8 will be the fastest OS for ANY version of PCMark when it will be released. Current builds of Win8 still very unstable and bugged. But 7955 can be used for PCMark04 and PCMark05 with some restrictions.

do you have some unofficial patch for pcmark04.exe that solve grammar check error? how do you get over 6 kb/s in this subtest?

Vgpu-OCP-mod: pencil R699 resistor or solder 100k VR between pin25 (IMAX) of uP6204B and ground Vgpu-mod: VR between pin20 (FB) of uP6204B and ground Vmem-mod: VR between pin6 (FB) of uP6101B and ground

Could be better, could be worst. I only had 2-3 liters for 3dm01 after wprime runs. Anyway, it's fastest 3dm01 score with SB and non evga or gigabyte board for now

There is no way to detect what score is "tricked" with CPU speed and what is not. Anyone who can play with affinity/priority and some tools for slow down cpu can boost score by 1%-10%-100%-... with it and you will never be sure unless score will be too big for the clocks/ranks. All that you can - adding rule for all new submissions/screenshots to show CPU speed (system info window -> processor speed or details window -> processor speed). http://i.imgur.com/tCNCy.png I think both 3dmark99 and 3dmark2000 must be completely removed. Or make "wrapper" for it to monitor cpu speed/usage. Are you sure about 3DM01? I think only 3dmark99 and 3dmark2000 uses cpu speed for fps/score calculation. No, it's not speedhacking software. Just slow down cpu by any application. System Info component of 3dmark99/00 measures cpu speed at start of the benchmark, then uses this number as multiplier in score/fps calculation formula. For example, if your cpu speed detects as 3 ghz normally and gives you 80k score, then if you "tricked" cpu speed twice (7 ghz) score will boost proportionally to 160k. This is error in development and such things needs to be patched, but unfortunately old benchmarks discontinued.