trodas Posted July 4, 2015 Share Posted July 4, 2015 Usually, when I overclocking, I find that there is the maximum overclock (FSB/CPU) that let me boot into Windows. Then, about 5 to 10MHz lower on the FSB (usually over 200MHz for the CPU clock) is the machine semistable for benching. And real stability happens futher down the road, about 15 to 20MHz on the FSB and 300+MHz on the max. CPU clock. A good example is my recent experiments on stock (not ever recapped, sadly, yet) ASRock 775i65G R3.0 mobo: default X6800: http://valid.canardpc.com/d6wa6r - 2.93GHz (FSB 266) stable: http://valid.canardpc.com/rjg1fw - 3.14GHz (FSB 286) benchable: http://valid.canardpc.com/i68sc4 - 3.25GHz (FSB 296) max OC: http://valid.canardpc.com/557qcn - 3.29GHz (FSB 300) To put it shot, veriety benchmarks need settings around 292 to 296MHz FSB to sucesfully run. Still is possible to boot at 300MHz FSB to Windows. And the 100% stability is only guaranteed at way down to 286MHz. Usually every board I was overclocking was like that. Wide margin of semistable opration before hiting the upper limit. I believe that this is normal situation. ... What lead to this thread was my recent experience with ASRock 775Dual-VSTA. W/O any mods (the board is *SCREAMING* to get added caps where they are stolen!) I put the crappiest CPU from the 775 socket line that I have into it (Celeron D 336) and the board surprised me. It cannot change Vcore, so I thought that around 150, 155MHz will be the absolute limit. It get me all the way to 164MHz. Okay, 4 phases, better design - so be it. Good. But what make my head spin is, that at 165MHz it cannot boot to Windows. However at 164MHz it not only can boot to Win, but it can perform Super Pi 32M test: http://hwbot.org/submission/2912173_trodas_superpi___32m_celeron_336_46min_24sec_125ms wPrime 1024M test: http://hwbot.org/submission/2913115_trodas_wprime___1024m_celeron_336_54min_7sec_374ms ...and everything I thrown at it: http://hwbot.org/submission/2911685_trodas_pcmark_2004_celeron_336_4526_marks http://hwbot.org/submission/2913095_trodas_3dmark2001_se_radeon_9600_xt_11811_marks ...including Windows XP SP3 install... That is very surprising to me, since I never ever experienced overclocking on board, witch have zero unstability margin. 165MHz = cannot perform Win boot, 164MHz = everything works just fine! ... So I would like to ask others, if anyone ever experienced something like that. Or if that is unique experience. Or what are your thoughts about this. PS. sadly I must admit that I find something that crash at FSB 164 - when I run the 3DMark 2003 and 2005 at default settings, eg. including the CPU tests, these CPU tests fail, complaining about memory address xxx cannot write to address xxx... So it is not stable and it is not because of the mobo or something, since dramatically lowering the FSB to 158MHz fix this. Maybe I could still tweak this by tweaking the "as fast, as they can post and pass memtest" ram settings, tough. Still, SuperPi 32M and wPrime 1024M tests do pass repeatedly at 164MHz and at 165MHz I cannot boot into Windows. Such short margin I never experienced yet. Or could be, that this is related to the not very demanding Celeron D 336 CPU and with more demanding, normal CPU like the Core 2 Duo X6800 Extreme it will return back to "normal" situation, where there is a margin of semistable operation? Quote Link to comment Share on other sites More sharing options...
K404 Posted July 4, 2015 Share Posted July 4, 2015 Sounds like some kind of limiter at the 166FSB strap and there is either a spread in the FSB or the BIOS is incorrectly marked. Could it be a block hole? Have you tried 170-175FSB, something like that? Quote Link to comment Share on other sites More sharing options...
trodas Posted July 5, 2015 Author Share Posted July 5, 2015 (edited) Well, I did not. Upon your suggestion I tried SetFSB in Win to force higher FSB from the "basic" 164MHz: Tried 170: http://valid.canardpc.com/vyznij ...and then 175: http://valid.canardpc.com/fxw94m And after a while at FSB 175 the BSOD visited me Luckily I managed to save the screens and CPU-Z validation before So yep, there is probably something in the bios, that is preventing 165MHz to work. Maybe 166, 167 or 168 will boot to Win, just not be entierly stable? As the 3DMark go, even 3DMark 2006 CPU tests worked just fine at 164MHz: http://hwbot.org/submission/2913960_trodas_3dmark06_radeon_9600_xt_713_marks ...so the problem is not in CPU, but rather (as the BSOD info sometimes visible) in the ATI drivers / AGP settings: The most interesting part is, that the test finished and at the end, when all was left to do is switch the screen / resolution - then the error happen. Not before. Never before. So it is not about CPU stability, as the other tests tend to suggest, but something into the AGP settings. Sadly these settings are too complicated and I never seen settings like that ever: Suggestions welcome. Tried incrasing the AGP voltage, but no luck. Enabled (from Auto) the AGP 3.0 Calibration and DBI Function (whatever that means?) and no change. Tried enable FastWrites and seen polygon errors during the CPU test only (?!) and same crash at the end of the CPU test, as before. Back to disabled state Edited July 5, 2015 by trodas Quote Link to comment Share on other sites More sharing options...
K404 Posted July 5, 2015 Share Posted July 5, 2015 (edited) The CPU strap is set during POST, so booting at 164 and OCing in windows to 175 means that the strap is definitely being pushed to its limit.... same behaviour as Asus P5B.... Boot at 395, then OC in windows... 410-415 FSB will be about all you will get because the strap needs to change. The board will obviously do 500+, but it needs the strap and sub-strap timings to be correct. Just after a strap change, efficiency will drop, but the new headroom means you will still come out better off. Basically, it's like driving a car. When you go up a gear (strap,) the revs drop (efficiency) but you can hit better speeds (literally) and the revs climb as you do. Did BIOS setting of 175 not work? Edited July 5, 2015 by K404 Quote Link to comment Share on other sites More sharing options...
trodas Posted July 5, 2015 Author Share Posted July 5, 2015 (edited) You really think that I should push that hard? After all, it is just a testing Celeon... I'm affraid that I end up clearing the bios )) Not to mention I should probably flash the latest beta bios, because it add support for 4G ram and 45nm CPU's, such as E6300/E7600 ;-) http://pctreiber.net/asrock-bios-downloads?did=103 That would be probably worth to try hard. And the crashing CPU test (at the end) of 3DMark 03 and 05 (but not 99, 00, 01 and 06) is what puzzle me more ATM that possibility to reach high clocks... PS. tried the 170MHz FSB on the latest official bios - WinXP booted just fine, then flick and crash to BDSD. Win32 something dll So no, probably it cannot get higher with that CPU. Flashed the beta 3.19a bios, no visible changes (except added option in the CPU page, but that it is). Edited July 5, 2015 by trodas Quote Link to comment Share on other sites More sharing options...
K404 Posted July 5, 2015 Share Posted July 5, 2015 Are you pushing hard, or learning about the board behaviour? Quote Link to comment Share on other sites More sharing options...
trodas Posted July 5, 2015 Author Share Posted July 5, 2015 I did not want to scare the poor board just yet... :-) ...but on serious note, w/o increasing the Vcore, we should not realistically expect that the Celeron will run at 3.6GHz stable, when it is just a 2.8GHz chip. Quote Link to comment Share on other sites More sharing options...
K404 Posted July 5, 2015 Share Posted July 5, 2015 (edited) By co-incidence of CPU and board, you are testing more than one variable at a time..... Edited July 5, 2015 by K404 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.