@markreflex - I would have included more info on them, but according to the 'techs' at micron/crucial, the sub timings are board-specific. I honestly don't know if what I have set in my system are considered "optimal" or not.
graysky, OC'd to 3.0Ghz and using OCCT, my load temps never breached 52C on any core. Thing is, Probe lists my Vcore as 1.33v and CPUID lists is at 1.245 to 1.25.
@schizoid - yeah, the asus suckware... er software really sucks. I would uninstall it if I were you. CPUZ rocks for vcore.
hmm so is the easy overclock to 333 mhz fsb for all of you because of your p35 motherboards? It seems the people without bear lake chipsets have to raise the vcore to 1.3 and up in order to OC to 333 while those with p35s can do it at 1.2.
markfw900 Moderator, sorry, see below
markfw900 Moderator, sorry, see below
@blanketyblank - I don't think it's the P35. My "old" q6600 (B3 stepping) does 9x333 no problems on my P5B-Deluxe (965 based chipset). Anything beyond 355 or so on the FSB requires a lot of vcore which I think is the limit of this chip.
ok. I just thought since the newer boards were designed to run at 333 for penryn chips at stock they were able to OC the older 266 mhz chips to this level more easily.
I'm having the best results overclocking my Q6600 G0 with a multiplier of 8x (up to 3.2 Ghz, 400 Mhz FSB) or 9x (up to 3.6 Ghz, 400 Mhz FSB), always with a tRD of 6 and a 1:1 memory divider, memory set at stock 4-4-4-15 timings, 2.1V. I found no upside to relaxing and overclocking the memory.
I found it crucial to disable all Gigabyte "helpful" BIOS settings, e.g. set Performance Enhance to "Standard", Loadline Calibration to "Disable", before I could even begin to experiment with overclocking with any success.
After reading many articles here, especially those by Kris Boughton, I found it handy to make various spreadsheets to guide and track overclocking experiments. Working tunings backwards, determining the FSB speed from a given cpu clock speed, it is the product of the cpu multiplier and tRD that directly computes Trd, best predicting memory bandwidth. I think of these products as like rungs of a ladder; lower is better. If I can't get on the (9,5) rung at 3.4 Ghz, try the (8,6) rung, and so forth. A filtered spreadsheet and common sense nicely guides combinations to try, for each rung. Also, a table giving Vcore and Ghz combinations that work or don't work is invaluable for seeing the straight line which predicts the Vcore that will be needed at other clock speeds, and guiding further experiments to hone the data.
It's hard to assess what other overclocking results here really mean, for this is a game of chicken, and I want 24/7 stability for scientific computations that will stress the machine for days on end. Prime95 is in fact a good proxy for how I'll actually be using the machine.
At 3.2 Ghz, I see core temps of 62 C or less, setting the BIOS Vcore to 1.28125V. At 3.5 Ghz, I see core temps of 68 C or less, setting the BIOS Vcore to 1.4V. I've been to 3.6 Ghz on air, where I chickened out. It's clear from what others tolerate for temps and voltage that I could have gone higher. I'm less concerned with who views 4 hours vs 24 hours of Prime95 (mprime on Linux) as stable than what voltages and temps are involved. In my experience, absolute stability for days on end is a minor tweak once one reaches stability for 20 minutes. Perhaps this is false at maximum overclocks; I never got there.
The decent power supply, roomy Antec P182 case with many fans, and the push-pull arrangement on the Thermalright U120E cooler, set on high by a fan controller, was crucial to reaching these speeds comfortably on air. I haven't lapped my cpu/cooler but I may. Since I chickened out on voltages rather than on heat, it's not clear how much lapping would help me.
I found it crucial to disable all Gigabyte "helpful" BIOS settings, e.g. set Performance Enhance to "Standard", Loadline Calibration to "Disable", before I could even begin to experiment with overclocking with any success.
After reading many articles here, especially those by Kris Boughton, I found it handy to make various spreadsheets to guide and track overclocking experiments. Working tunings backwards, determining the FSB speed from a given cpu clock speed, it is the product of the cpu multiplier and tRD that directly computes Trd, best predicting memory bandwidth. I think of these products as like rungs of a ladder; lower is better. If I can't get on the (9,5) rung at 3.4 Ghz, try the (8,6) rung, and so forth. A filtered spreadsheet and common sense nicely guides combinations to try, for each rung. Also, a table giving Vcore and Ghz combinations that work or don't work is invaluable for seeing the straight line which predicts the Vcore that will be needed at other clock speeds, and guiding further experiments to hone the data.
It's hard to assess what other overclocking results here really mean, for this is a game of chicken, and I want 24/7 stability for scientific computations that will stress the machine for days on end. Prime95 is in fact a good proxy for how I'll actually be using the machine.
At 3.2 Ghz, I see core temps of 62 C or less, setting the BIOS Vcore to 1.28125V. At 3.5 Ghz, I see core temps of 68 C or less, setting the BIOS Vcore to 1.4V. I've been to 3.6 Ghz on air, where I chickened out. It's clear from what others tolerate for temps and voltage that I could have gone higher. I'm less concerned with who views 4 hours vs 24 hours of Prime95 (mprime on Linux) as stable than what voltages and temps are involved. In my experience, absolute stability for days on end is a minor tweak once one reaches stability for 20 minutes. Perhaps this is false at maximum overclocks; I never got there.
The decent power supply, roomy Antec P182 case with many fans, and the push-pull arrangement on the Thermalright U120E cooler, set on high by a fan controller, was crucial to reaching these speeds comfortably on air. I haven't lapped my cpu/cooler but I may. Since I chickened out on voltages rather than on heat, it's not clear how much lapping would help me.
Well, I can tell you that @ 1.2625 V running my b3 q6600 @ 9x333 lapping helped to the tune of 7-10 °C which is 12-15 % reduction.
Lapping Q6600 IHS
Lapping the Ultra-120 Extreme
this thread is FTW and helped me get some things strait in my head before i did my first Quad Core OC.
Might be useful to add some information about strappings and stability holes.
I don't mean strappings as in regards to running asynchronous memory , but as regards to the CPU. I didn't learn about this till later, but it seems that there is a fsb hole where it's almost impossible to stabilize. However once you can go beyond that apparently you can do a higher fsb with a higher strapping.
It seems this was how I was able to actually get into windows at 400 mhz once, but couldn't boot over 350. After reading through some forums, it seems that at that level there is a 1333 strapping which allows for these speeds though there is a bandwidth cost.
Another interesting thing I found out with my overclocking experience is that if I run my memory at 2:3 while I overclock at 300 mhz my perfomrance in games actually goes down. That is games like GoW get choppy even though at 1:1 it runs smooth. I'm not quite sure why, but my theory is that taxing my mobo by OCing both the cpu and memory at the same time results in less bandwidth for my video card or that it's affecting how my RAM somehow.
I don't mean strappings as in regards to running asynchronous memory , but as regards to the CPU. I didn't learn about this till later, but it seems that there is a fsb hole where it's almost impossible to stabilize. However once you can go beyond that apparently you can do a higher fsb with a higher strapping.
It seems this was how I was able to actually get into windows at 400 mhz once, but couldn't boot over 350. After reading through some forums, it seems that at that level there is a 1333 strapping which allows for these speeds though there is a bandwidth cost.
Another interesting thing I found out with my overclocking experience is that if I run my memory at 2:3 while I overclock at 300 mhz my perfomrance in games actually goes down. That is games like GoW get choppy even though at 1:1 it runs smooth. I'm not quite sure why, but my theory is that taxing my mobo by OCing both the cpu and memory at the same time results in less bandwidth for my video card or that it's affecting how my RAM somehow.
Thanks.
My "other" box is a B3 Q6600 on a Bad Axe 2 motherboard, still at stock with a Zalman 9700 cooler in an Antec P150 case. What a difference these choices make! Before overclocking, my new rig used less power and ran cooler at full load than this "other" box runs at idle. This has been my "production" server while I tune the new box. Any overclocking of this box will surely be heat-limited.
I'll make my first lapping experiments with this B3 Q6600, moving to an Antec P182 case I picked up on sale, and upgrading the cooler to a push-pull Thermalright U120E. Lapping sounds fool-proof if one is careful (e.g. don't get the cpu wet), but I should be so lucky as to destroy this B3 cpu, I'd have to buy a G0 replacement. I'm guessing that in years to come, people will be fondly recalling the G0 Q6600 for the amazing overclocking beast that it is.
Hmm, I wasn't sure where to put this but I have a stability issue that isn't related to the CPU or memory, at least I don't think so. My system is Orthos Blend test stable, but when I run Atitool and Blend simultaneously, the system freezes within 10 minutes. The freezing occurs in gaming as well but I've never had a crash in Windows doing something not related to the GPU like writing this post. So what is the problem in my system? Is there a voltage I should increase or some other tweak. I haven't tried stressing with stock CPU yet since it doesn't seem to be the culprit. I did revert to default GPU clocks but that didn't help. Any ideas?
Dunno for sure, but it sounds like a graphics card issue... what if you lower the cpu back to stock and repeat?
Very valuable information from the OP, thanks! I havent been able to read through all 20 or so pages of this thread but I just got a Q6600 for my Asus P5B Deluxe along with some of this memory: http://www.newegg.com/product/...x?item=N82E16820231145
Upgraded from a E4300 and 2GB of OZC DDR2 800 Platinum rev 2.
And have been attempting to maximize overclocking on the new processor and memory. Obviously there are FSB holes in this board and I was able to use 390 x 9 multiplier but I can also do 460 x 8 multiplier. I cant get it past 460 no matter what I try for some reason. I went by this guide here: http://www.thetechrepository.com/showthread.php?t=41
And the memory timings I cant get lower than 6-5-5-17 @ 1:1 when using the 8x multiplier. Can anyone help me out here if I am doing something wrong or am I just hitting the wall in this board here? I thought that the P5B Deluxe was able to hit 500 pretty easily. I have tried a lower multiplier and it doesnt matter I still cant get past 460.
BTW: Case is Antec 900 and CPU Cooler is 9700 Zalman, PSU Corsair 520W. CPU Voltage never is higher than 1.475 and very safe temperatures I have never seen higher than 50C at full load.
Thanks in advance!
Upgraded from a E4300 and 2GB of OZC DDR2 800 Platinum rev 2.
And have been attempting to maximize overclocking on the new processor and memory. Obviously there are FSB holes in this board and I was able to use 390 x 9 multiplier but I can also do 460 x 8 multiplier. I cant get it past 460 no matter what I try for some reason. I went by this guide here: http://www.thetechrepository.com/showthread.php?t=41
And the memory timings I cant get lower than 6-5-5-17 @ 1:1 when using the 8x multiplier. Can anyone help me out here if I am doing something wrong or am I just hitting the wall in this board here? I thought that the P5B Deluxe was able to hit 500 pretty easily. I have tried a lower multiplier and it doesnt matter I still cant get past 460.
BTW: Case is Antec 900 and CPU Cooler is 9700 Zalman, PSU Corsair 520W. CPU Voltage never is higher than 1.475 and very safe temperatures I have never seen higher than 50C at full load.
Thanks in advance!
500 FSB? Maybe... but quads are more demanding that duals. 460 is pretty damn good for a Q6600 dude, you might just be as high as you can go with it. Also know that 1.475 V is pretty much near the redline for that chip as per Intel as well. Dunno what to tell you about your 6-5-5-17 timings... if you look at your chips, are they high density or low density? Pull one out and look at the sub partnumber. If it starts with a .8 you have high density memory which the P5B deluxe really hates. I had to hand-pick out the .16 (low density) ones for my board.
See this thread for more on this as well as a picture of the DIMM so you can see what I'm talking about. Also, what BIOS are you using?
See this thread for more on this as well as a picture of the DIMM so you can see what I'm talking about. Also, what BIOS are you using?
Thanks for the quick reply! Well thats good to know that its about where I should be with the CPU and MB, I ended up dropping it down to 450 FSB 8x multiplier and 1.425vcore for saftey. The BIOS is 1226, I believe thats the latest version from asus. I will check the memory here shortly.
I just took them out and looked at them, wasnt sure what subpart number there was so I just took a few pictures. It appears as though they are two-sided just from looking at them on the side, but who knows. Here are the two picks I took of the DIMMs and of the box they came in:
http://www.drakedavid.com/DSC_0267.JPG
http://www.drakedavid.com/DSC_0268.JPG
I mean they are rated at 5-5-5-15 and I only have them running at 450x2 = 900 so they should be able to run fine right? I wonder if I should contact newegg for another pair or G.Skill....
here's a quick observance.....
TheDrake a few posts up referenced a guide at the techrepository. in that thread, post #3 went into a little more detail on the FSB straps, and what is happening with the memory controller itself. ok, IN that post, Tony referenced this link:
http://www.thetechrepository.c...thread.php?p=32#post32
it goes into even further detail on the straps. in post number 2, i was quite stunned at how the actual clock of the NB can vary drastically from what we assume when we change from the default multiplier!
i too have a q6600, which defaults to a multiplier of 9. TheDrake, if you have yours set to a multiplier of 8, then based on that, at a FSB of 460, your REAL northbridge frequency should be 517.5 Mhz. a higher FSB would only drive this higher.
might this perhaps be the reason for some instability? maybe it's not the 450x2=900 that is the issue with the G.Skill RAM?
i have mine running at 333 FSB now, at a multiplier of 9, 1:1, and it does perfectly fine. but if i try ANYTHING above that, it does not post at all. from that techrepository guide, apparently i would have to jump straight to 400 FSB to avoid the "hole". but that seems to mean that unless i'm going to attempt 3.6 Ghz, i'd have to lower the multiplier to 7 or 8. this would in effect raise this new issue of seriously overclocking the northbridge. --> 9/8 X 400 FSB = 450 Mhz and 9/7 X 400 FSB = 514 Mhz.
alright, i really apologize if this post is too long, but PLEASE feel free to tell me if i am reading all this correctly, and if this might be an issue with stability with different multipliers.
TheDrake a few posts up referenced a guide at the techrepository. in that thread, post #3 went into a little more detail on the FSB straps, and what is happening with the memory controller itself. ok, IN that post, Tony referenced this link:
http://www.thetechrepository.c...thread.php?p=32#post32
it goes into even further detail on the straps. in post number 2, i was quite stunned at how the actual clock of the NB can vary drastically from what we assume when we change from the default multiplier!
i too have a q6600, which defaults to a multiplier of 9. TheDrake, if you have yours set to a multiplier of 8, then based on that, at a FSB of 460, your REAL northbridge frequency should be 517.5 Mhz. a higher FSB would only drive this higher.
might this perhaps be the reason for some instability? maybe it's not the 450x2=900 that is the issue with the G.Skill RAM?
i have mine running at 333 FSB now, at a multiplier of 9, 1:1, and it does perfectly fine. but if i try ANYTHING above that, it does not post at all. from that techrepository guide, apparently i would have to jump straight to 400 FSB to avoid the "hole". but that seems to mean that unless i'm going to attempt 3.6 Ghz, i'd have to lower the multiplier to 7 or 8. this would in effect raise this new issue of seriously overclocking the northbridge. --> 9/8 X 400 FSB = 450 Mhz and 9/7 X 400 FSB = 514 Mhz.
alright, i really apologize if this post is too long, but PLEASE feel free to tell me if i am reading all this correctly, and if this might be an issue with stability with different multipliers.
I think I see what you mean. So instead of the memory being at 460 x 2 it is actually 517 x 2. I mean I understood the whole FSB thing from the other thread I just didnt realize that it directly affected the memory too, dummy me. With the setting of 450 FSB that I have now and the timings above I have only been able to reach about 7GB speeds via sisoftware Sandra Memory bandwidth and according to it thats only 43% efficiency.
What I may try to do is just up the multipler to 9 and lower the FSB so that I can get some better timings out of the memory and see how it does. Thanks for the help!
What I may try to do is just up the multipler to 9 and lower the FSB so that I can get some better timings out of the memory and see how it does. Thanks for the help!
@thedrake - sorry man, I thought I read that you had crucial ballistix memory but I can see from your pics that I was mistaken. Forget my advice about the looking up the sub partnumber. If you look at that thread I linked, you'll see what I meant about the little sticker though.
@TheDrake:
based on what those articles indicate, i really do think that is what's happening. now, i do not know for sure, as i am not an IT tech.
but it would make sense, if behind the scenes, the memory was really running at that actual NB frequency, and therefore was entirely too overclocked. also, since you have the Sandra suite, one of those posts indicated that Sandra gave the true FSB speed in it's system specs display. look at that above highlighted link, post #2, at his screenshots. (the "mainboard" tab in Sandra, as compared to what CPU-Z displays.) i'm curious as to what your's reads.
i'd do it too, but since i rebuilt my system late 07, i've not played with Sandra at all.
i ran some numbers last night too. if you do use frequencies between say around 350 to 399 for the FSB, it would seem to work mathematically if you used a lower multiplier than 9. for example, 9/8 X 366 = 411.75 Mhz. could this perhaps be why a system using this FSB will work, but at a multiplier of *9*, NOTHING will make mine POST at 366?
i am itching to experiment!
based on what those articles indicate, i really do think that is what's happening. now, i do not know for sure, as i am not an IT tech.
but it would make sense, if behind the scenes, the memory was really running at that actual NB frequency, and therefore was entirely too overclocked. also, since you have the Sandra suite, one of those posts indicated that Sandra gave the true FSB speed in it's system specs display. look at that above highlighted link, post #2, at his screenshots. (the "mainboard" tab in Sandra, as compared to what CPU-Z displays.) i'm curious as to what your's reads.
i'd do it too, but since i rebuilt my system late 07, i've not played with Sandra at all.
i ran some numbers last night too. if you do use frequencies between say around 350 to 399 for the FSB, it would seem to work mathematically if you used a lower multiplier than 9. for example, 9/8 X 366 = 411.75 Mhz. could this perhaps be why a system using this FSB will work, but at a multiplier of *9*, NOTHING will make mine POST at 366?
i am itching to experiment!
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter