trodas

Well, that would be interesting test to try For me it just fail ... but maybe I somehow helped it to happen? Dunno. K404 - Just to know, how to help it Mr.Scott - I believe it is. It says right there: ...and I would very much like to see some of the rigs ObscureParadox - Thank you. I just scratching my head about what I did wrong... nevermind. Let's hope for K404 vid to see, how the affinity is supposed to work. I was setting PCMark04 to run on just single core and still no go Cursed grammar test

When searching into PCMark scores with my CPU, I stumbled upon this score: http://hwbot.org/submission/590463_joe_cool_pcmark_2004_core_2_x6800_%282.93ghz%29_10191_marks The thumbnail does not work. So either make gif thumbnails working, or convert all gif images into png Either way, the thumbnail should be visible. The picture works well, when clicked on the link "PCMark 2004 screenshot" in the box...

Well, good luck with proving that argument invalid: show me a CPU, with 4MB+ L2 cache, what scored on PCMark04! I looked there: http://hwbot.org/benchmark/pcmark_2004/ And it is all Semprons and P4 there in the one CPU: http://hwbot.org/benchmark/pcmark_2004/rankings?cores=1#start=0#interval=20 For the dual CPU's, there are: http://hwbot.org/benchmark/pcmark_2004/rankings?cores=2#start=0#interval=20 ...some more interesting CPU's, but all they have is 256k of L2 and 3MB of L3 in best case scenario: http://hwbot.org/submission/2610863_wolfnyght___kew_team_pcmark_2004_pentium_g3258_21128_marks Nowhere near 4MB of L2. Only the AMD Phenoms having 512kBy L2 and 6MB of L3 are getting over the 4MB, but not L2. Therefore there is no CPU that come close with the L2 size... but there is an better argument against the sheer size of L2: Core 2 Extreme X6800 http://hwbot.org/hardware/processor/core_2_x6800_293ghz/ ...and yes! There are PCMark04 scores: http://hwbot.org/benchmark/pcmark_2004/rankings?hardwareTypeId=processor_875&cores=2#start=0#interval=50 Whole 16 of them! ... And since the demise of Futuremark online scores for PCMark04, vast majority of them is not possible to verify. Let me state it clear - from the 16 scores, the wast majority - whole 11 scores - cannot be verified by any way to be real. None of the 11 have also any picture of the machine that was benched, witch is IMHO against the rules for the first places (and yes, there are 1st, 3, 4, 5 and 6th places amongs them, not counting the rest...). So we have 5 scores only left, with at least screenshot of the test in them: 2nd place: http://hwbot.org/submission/582356_mayk_pcmark_2004_core_2_x6800_%282.93ghz%29_13547_marks (no images of the machine, no CPU-Z verify link, no info about mainboard...) 7th place: http://hwbot.org/submission/590099_gianlu_pcmark_2004_core_2_x6800_%282.93ghz%29_12553_marks 9th place: http://hwbot.org/submission/2351569_evil_pcmark_2004_core_2_x6800_%282.93ghz%29_12198_marks 10th place: http://hwbot.org/submission/589864_ivan992_pcmark_2004_core_2_x6800_%282.93ghz%29_12193_marks (TRAS 1?! WTF!) 15th place: http://hwbot.org/submission/590463_joe_cool_pcmark_2004_core_2_x6800_%282.93ghz%29_10191_marks (image thumbnail is not showing for some reason) ... NONE of these scores show the rig, or give and info that could be verified now. Maybe they using older version of the PCMark04 that works? I did not say that all these results are cheats, but once again - if you want to verify them, then you come up short. There is (with the Futuremark demise of the PCMark04 support) simply no way... I trying at least insert even for Aquamarks the CPU-Z link into comments, but there is nothing. And I mean NOTHING.

Well, if there IS a piece of software, no matter how insane it is, that fail with overclock and NOT fail w/o overclock, then I would never call that overclock stable. Might be benchable, but it is not stable. ... And there is MUCH better video to watch for you: PCMark04 fail with grammer test on BOTH cores, as expected: Watch the load on the cores. Now there is the original video with director cut edition - I added the missing info and it will show you, that affinity does work for me, when done by task manager: Hopefully processing to the HD will be done quickly by YT. IMHO the core(s) load prove, that the PCMark04 does run only on Core 0, therefore the affinity setting in the task manager works for me. (especially compared to the 2 cores run it become IMHO obvious) But I await your opinion on that matter... ... So now I have another question - is there are ever a PCMark04 score on Core 2 Extreme X6800? Or any other CPU with 4MB+ cache? You probably see already, where I pointing ...

The IE vs FF discussion is mute - no one I know have the ability to patch PCMark04 exe that way, it can use another browser, not to mention (since we all know, how M$ is upholding standards) that it might not be even possible, after all. What calls are used for IE, might not even work for FF, Opera, Safari or whatever browser will render fastest the page(s) during the test. So let's drop that discussion, unless there is patch like that somewhere... In theory, the CPU overclock *might* affect some settings and some instructions, but since it is not my first, nor my second, not my 50th overclocked CPU, then I'm fairly certain, that when I say that ATM a 286x11 settings are stable for my ASRock 775i65G, then they are. Proven once again... Now the services that are need for the affinity settings - that could be productive discussion that might lead somewhere. But the video - that is even BETTER idea. And I have another idea... 1) I made a video, witch show the PCMark04 in action, eg. how grammar test fail with the affinity set (hopefully well?) by task manager. I only forget to show the CPU-Z screen to prove that the CPU is NOT OVERCLOCKED at the time of the test... There you go: Do I done it right? 2) Now the video only lack a CPU-Z proving that the CPU is *NOT* overclocked at the time of the video... 3) ...but it also lack - and that is what idea I got now - showing the task manager window with the cores load to determine, if the PCMark04 grammar test works on one core only! In fact, I think that we can see proof that affinity does work - or not. The load from the Bandicam (m-jpeg, 15fps) could / should be (hopefully) possible to differenciate from the PCMark04 test and if the PCMark04 does not exceed 50% of CPU load, then affinity works well, because it was not able to use more that ONE core...! ... And is there some point I trying to made? YES, there it goes: maybe the grammar test fail because the CPU have 4MB cache also, not only the 2 cores is the problem...! So maybe this error is unrecoverable... Time to make another vid...

Well, I don't exactly see why, because Firefox is usually slower the IE in the pure rendering speed... but whatever. Different versions of IE could also give different speed results, but that it outside of the scope of this thread, because that would require modificatons to the PCMark04, witch is something that require a great deal of knowledge and hacking skill - eg. nothing I could do. Back to your idea about default speed. Ok, speed in bios set to auto, as CPU-Z prove to me: As you can see, it again did not work at all. Before 9:01h I set the affinity to the PCMark04 will run at time 9:01h on single core, then from 9:03h back on dual core(s). Nothing happend. I hope you now recognize that I was right and overclocking withing stable limits does NOT cause the affinity is not working. Right? ... Aside this, rather personal question, take a good look at the Winblows CPU info. Yes, that is kinda... missing out there, right? There is not much bellow the memory information. I checked on my Win XP SP3, what it does tell about my P4 650, and again there is no CPU type, but at least it says that the PA mode (extension of physicall address) is ON: No idea, if that could be some key to understand why the affinity does not work, but I just reporting that, what I find. Also - could not some BIOS settings caused that? See: http://forum.hwbot.org/showthread.php?t=134923 The "Intel Virtualization Technology" is now completely ghosted and I remember that on P4 it was possible activate it...

For some old school fun and testings I sometimes overclock really old hardware, like FX 5600XT or FX 8600GT, FX 6800GS, FX 6800GT... Now for the FX 5600XT card I use old beta Forceware 45.28 that I modifield slightly and using these Coolbits let me improve the performance by overclocking the card: REGEDIT4 [HKEY_LOCAL_MACHINE\SOFTWARE\NVIDIA Corporation\Global\NVTweak] "CoolBits"=dword:0000000f "NvCplEnableHardwarePage"=dword:00000001 "NvCplEnableAdditionalInfoPage"=dword:00000001 "NvCplEnableAGPSettingsPage"=dword:00000001 (the only important part for the clock adjustment is the "CoolBits"=dword:0000000f. The performance scale nicely and all is good... except some more modern 3DMarks require more modern drivers to operate. Surprisingly using the modern drivers AND Coolbit overclock result in sluggish performance. For example 3DMark03 goes from 2551 to 563 marks: http://hwbot.org/submission/2885484_ http://hwbot.org/submission/2885483_ Yes, on the Forceware v93.71 drivers I have a slightly different O/C (365/301 vs 370/305 on the old drivers), but nowhere near to "justify" such performance drop. Now this is not specific on the 3DMark, all 3D apps go sluggish after using the Coolbits on more recent drivers. For example Aquamark3 - Forceware 61.76, stock 234/200: 13341 points: http://hwbot.org/submission/2885479_ Overclocked to 365/301 by Coolbits: 6640 points: http://hwbot.org/submission/2885477_ So there IS a dramatic slow-down, when I use more recent Forceware drivers (tried v61.76 and v93.71) AND Coolbits. Same I observed on other mentioned videocards, like the FX 6800 and 8600 one, so it is not tied to the FX 5600XT only. Therefore I look for more recent Coolbits, found there: http://downloads.guru3d.com/download.php?det=815 Again, I used them at first completely, then removed them - for example that easily it could be done: REGEDIT4 [-HKEY_LOCAL_MACHINE\SOFTWARE\NVIDIA Corporation\Global\NVTweak] ...so I could strip them to the bare important for clock changes: "CoolBits"=dword:ffffffff Now this is different value that the 0000000f used before, so I hoped... and it caused the exactly same slow-downs. That make me a "bit" puzzled, so I started searching and I found third version of Coolbits: "CoolBits"=dword:00000003 http://en.wikipedia.org/wiki/Coolbits It probably won't be any surprise now, when I write that this once again enabled the O/C possibility, but caused sluggish performance. ... So I give up on Coolbits and used Riva Tuner. It does work w/o reinstalling on new Windows, the O/C is limited to 355/300 (600 effective) and once again - after overclocking, the performance go bad. Slow. Terrible. Even some image quirks, at 355/300, when 370/306 works nicely with old drivers. ... Now could anyone tell me, what is going wrong there? Why Coolbits / or any O/C in general, cause slow down of graphic? I hoped I could find someone else who had the same problem years and years ago, I manage to find some threads where people appear to having the similar troubles: http://www.pcreview.co.uk/threads/coolbits.2806125/ "I just overclocked the GPU a tiny bit, the frame rate and graphics slow down and become juddery" ...but that is too few results from my search... so perhaps I doing something wrong, like overoptimizing by disabling the nVidia help service or the startup programs that was put there by the nVidia drivers install...? Once again - if I did not touch the overclocking, the performance is like it should be. After Coolbits = sluggish speed. Or could it be, that any WHQ nVidia drivers lock-out the overclock by intentionaly slowing things down and I could only use beta drivers for overclock?! Could that be? And where to get some old beta drivers anyway? There are all WHQ ones, except one Quadro Beta Driver: http://www.nvidia.com/object/winxp-2k_archive.html

Yes, this is a possibilty. However the same possibility is, that for cost-effective approach, it is all about just some resistors on the GPU or it is even just a bios lock. That would be easiest way to do. I remember, that some nVidia cards could be unlocked from consumer to pro versions by just two resistors and other bios - and that was it.

Pretty cool How the AMD FreeSync over HDMI demo looked in action? They used slightly (firmware) modified Nixeus NX-VUE24 monitor and slightly modified drivers to get 30 - 144Hz variable frequency Pretty cool, IMHO http://www.anandtech.com/show/9337/amd-shows-freesyncoverhdmi-concept-at-computex-2015

Dunno. It all boil down to whatever the chip do have additional components or not. You need the hi-res image of the chip structure to determine if there are additional inactive parts or not. For example GTX 980ti have inactive parts also - because activating them means that it beat Titan X without overclock, witch is what nVidia does not want for selling reasons. Also matter, how these parts of the chip are deactivated. If there was, for example, laser cut... then all is hopeless. If there is not a laser cut, but a bios or resistors/capacitors driven change on the top of the chip, then unlock is possible. ... Another good example was the double precision. For the professional AMD cards it was unlocked, for the consumers segment the real double speed precision calculations was locked down. Of course to prevent them being used as the professional cards. Also with some research it could be determined, how the precision speed is locked and unlock it

First let me clarify that I did not (yet) have the knowledge or even verified information, that this could be possible. But please bear with me and consider the information - maybe someone can spot the differences and help others improve the GPU performance. So, in recent AMD materials for APU Carizzo was pictured a GPU Tonga, witch is present in Radeon R9 285 and feature 1792 stream processors, 32 ROP units and 256bit wide memory bus: AMD GPU Tonga However the catch is, that the die size of Tonga is 359 mm², witch is only slightly smaller that his predecesor, GPU Tahiti with it's 365 mm². How does that matter? Well, there is very likely to be plenty of deactivated parts, giving the die size. Some people claimed, that AMD confirmed for them, that physicaly Tonga boast 2048 stream processors, not only 1792 active ones. But there is something even more serious: from the picture is clear, that there is 2048 stream processors (32 CU) - and that there is, in fact, SIX 64bit memory bus controllers, witch means 6x64bit = 384bit wide memory bus is possible! Now it is unclear, how many ROP units are actually on the chip: it could be 32 as in Tahiti, but it could be also 48 to simplify the memory interface - ROP units and memory controllers could be wired directly, as it is done with other chips. So all in all it means, that GPU Tonga have the possibility of same power, as GPU Tahiti PLUS it have faster tesselation and delta compression. If there is 48 ROP units, then at same clocks as GPU Tahiti, this fully unlocked GPU Tonga could give about +20% of speed...! If the number of ROP units remain at the Tahiti levels, it will give about 10% speed increase over Tahiti - eg. R9 290X card. ... The fully active GPU Tonga will probably get marketed as R9 380X and the release data is unknown, but there is another thing that is possible: unlocking the chip DIY hacking. ... Yes, we cannot change the memory bus interface (making memory bus wider is practically equivalent to creating a own card from scratch) on current R9 285 cards, but since these stream processors are deactivated likely by bios or some switch component(s) on the GPU itself, we can - theoretically - unlock the 2048 stream processors, witch could provide still a significant boost in the performance It will be different to have a 1792 or 2048 stream processors - even the number of (at least active) ROP units remain at "low" 32... Inspired by: http://www.pcgameshardware.de/AMD-Radeon-R9-285-2G-Grafikkarte-259969/News/Tonga-GPU-mit-256-Bit-Interface-und-32-CUs-1160632/

Looks cool, people seems to having fun... so far, so good Enjoy mate!

basic weapons will do fine

Stop, we are scared...! We give you our LN2 willingly, just don't hurt us! ... BTW, nice room you have. I wish you luck in the competition

K404 - Well, what could I have done wrong? The software is pretty straight-forward. I set the app, then set the time and what operation should happend and that it is. As you can check from these two screens: As you can see, in 9:45:00 nothing happend... Could this software work on basic (Win32) XP SP3 after all? Or a 64bit WinXP are need? Both cores are used to the max, IIRC. True. However I can even overclock UP into unsafe zone and ALL tests (except Gramar and Web page render) are passed: So it is NOT the overclocking In fact, I can pass other benches at higher, unstable, settings: http://hwbot.org/submission/2878614_trodas_3dmark_99_max_radeon_9600_xt_28951_marks http://hwbot.org/submission/2878610_trodas_3dmark2000_radeon_9600_xt_21768_marks So this is out of the possibility. Look, I know what I doing and I know what parts of CPU will fail soonest when overclocked. Not the core functions, but ram bus controll functions, as they are physically located on the corner of the chip, so Prime 95 memory stress test is best way to check for stability. Damn! Yes, that is the test that failed... So, there is no way to convince PCmark14 to use default browser instead, right? Then I cannot produce the score (if not reinstalled Win) and we can only concentrate on the Affinity not working QuickFast - Except the Affinity you are right. I installed the WMEncoder9 and WMPlayer9 to get things running. I failed on the Affinity and on the Internet Explorer thing...

K404 - Good point. It is NOT in the history. In fact, nothing even get it into the history part of window. Dunno why. Perhaps because the software does not work? One minute before I set the Affinity. How could it confirm that? In Czech version of Windows XP SP3, there is no Affinity mentioned anywhere. I can only force program to some of the core(s), that it is. Nothing else. Mr.Scott - All test and benches are working fine at my stable settings. If SuperPi 32M test or wPrime 1024M tests works fine - why should not few minutes of PCmark04? I can, of course, test at stock. But I can guarantee you, that there will be NO difference. It again fail to produce the cursed damned score. K404 - So, do you get the score? That is the only one thing I care about right now. QuickFast - Internet Exploder uninstalled

Good night at Taipei Looks like you will have a great time there... So try not gain too much weight

Somebody should *PATCH* this bloody lame PCmark04 to work Manual clicking game is not funny at all Even when I set Affinity to 1 core for the WHOLE PCmark04, I did no get the score no matter how hard I try: So on Core 2 Extreme X6800 it does not work at all. Period.

Congratulations - hopefully it will be a good show! Pls report back with some fresh pics and infos AMD are just teasing us with supershort previews: ...and GTX 980ti are (unsurprisingly) 5% bellow Titan X, so... with a little O/C it is a Titan X speed for GTX 980 price: http://www.hwbattle.com/bbs/board.php?bo_table=hottopic&wr_id=175&ckattempt=1

Great, thanks! Much better now

not exactly santa

2498 submissions for such nice mobo: http://hwbot.org/hardware/motherboard/p5e3_premiumwifi_apan/ ...and no image? Fix is need: :celebration:

Probably the executable get on D drive infected too, witch lead me to believe that the source was infected? But better check things twice, you never know I was kinda desperate, 3 or 4 times reinstallin winblows and no help, virus is still there? WTF! So if someone checked, then I hope it was just a false positive and I must get in to the testing computer by other means. Hmmmm.

Christian Ney - Thank you - looking much nicer now! Kojima45 - Damn shame. You are sort of setting the exception for Japansee. Mostly we percieved them as the ones that take too much pictures... You know the joke: Man shortly kidnapped a bus full off Japansee tourists. Now the police have over 5 000 images of the man... Interesting. If mine are the same, then I could get higher... Thank you! Sadly the CPU and the board was between the stolen goods from me, so I have no idea if I can even get hold of it again. It might be interesting how far can I go on air, watter... or maybe with some more intense cooling methods? We see But hey - I definitively love japansee caps - Nichicons Good quality

Much nicer! Thank you very much