Hello,
First post on here, and I know the topic practically begs for a "use the search" reply, but I have searched extensively on both Google and this site, and cannot seem to find an up-to-date answer that doesn't look like complete hearsay. I seem to find the most accurate information here though, so I thought I'd register and ask.
I can reach 4 GHz with BLCK 200 x 20 multiplier. I can also reach 4 GHz with BLCK 148 x 27 multiplier. So which is really better? There are some points to consider, that I haven't seen reflected in any other answers I've found, but since I haven't found any actual answers, I'll have to pose them as questions;
1. Auto Multiplier
I see two sets of behavior in Windows; if I manually set a multiplier in the BIOS, Windows will (according to CPU-Z) switch between that multiplier and 12, which I guess is a low power state. If the BIOS setting is Auto on the other hand, Windows will gradually change the multiplier and thus frequency of the CPU. I should clarify, that I did not change the SpeedStep option, but rather left that enabled all along.
- Does this gradual change provide a benefit in terms of increased efficiency and lower average temperatures? As in; does it help my electricity bill, will it result in lower temperatures even while gaming by throttling to the demand?
I would be inclined to believe that this is the case? I mean, SpeedStep was invented for a reason, I'd suppose. On the other hand, even with a locked multiplier, I do see that temperatures change with load, so clearly less power is drawn by the CPU when idle, regardless of the locked frequency?
2. Higher BCLK
If my perception on the above point is misguided, then I guess this is almost self explainatory. Anyways; A BCLK of 148 (as I currently have) is an 11 % overclock on that metric. 200 would correspond to 50 %. So am I missing out on a lot of performance increase by not going for a BCLK overclock with a low multiplier? And is that performance actually relevant? I mean; RAM isn't worthwhile overclocking in my mind, as there's plenty of it in the PC, and what else is there, really?
3. Refusing OddArguments
I followed the (excellent) three step guide to overclocking i3/i5/i7 on here. It's the first time I've overclocked "methodically" and only my second overclock ever. Still; although I'm not very experienced, I sincerely don't see the logic in the statements that a BCLK overclock is bad, because it will overclock other components, introduce variables, etc. When it's isolated "as best as it can be" in the BIOS and goes up to 200 easily, what would be the harm done, then how should it suddenly become an issue to raise it by 10 more than the default value, which is what I've seen implied over and over. Even talks about USB bus speeds being affected. I would assume this argument stems from maybe other, more finnicky chipsets or from people who just try stuff and get annoyed that it fails being a bit eager to blame the tools.
Other arguments I do see a valid point in, for instance in increasing the air temperature inside the case since other components will be overclocked and contribute to the heat build up.
But probably what I most seek clarification on is; does it matter? I understand QPI, I get how the BCLK is different from FSB (direct memory control and 3 channels was the factor in my purchase of an LGA1366 board in 2008 or thereabouts), but does it really give me a noticable performance increase anyways? I should add, that my GPU seems to definitely be the bottleneck of my system, so any (worthwhile) increase on that aspect would be welcome.
I'll drop some background in, for those who are interested or in case it's needed to answer the questions
Having always been uninterested in overclocking, I happily ran my i7 920 for years and years. Best value-for-money system I ever build, by far. In the summer or autumn of last year, it was starting to show it's age in relation to new titles though, and I though I'd try and mend that by overclocking the CPU. I did, and it worked fine. I just overclocked it very simply by raising the BCLK from 133 to 166, rebooting and enjoying. It probably shouldn't even have worked, but the system performed nice and stable for 8-9 months. I didn't feel like fine tuning anything when the result was miles better than expected - I felt I had struck gold. It also drastically improved my gaming experience, most notably The Division increasing from 20 FPS to 50 FPS, effectively turning the table so that my bottleneck went from CPU to GPU.
I now replaced my i7-920 2,66 GHz @ 3,5 GHz with a Xeon X5690 3,46 GHz currently @ 4,0 GHz. Nothing was wrong with the i7, I just felt like trying an upgrade to see if it'd help a drop in FPS I got from increasing screen resolution in games from 1920x1080 to 3840x1080 due to a new monitor. As I could find a X5690 cheap, I decided to try it, but this time with the goal of overclocking right from the start. I rely on my PC for work and play, so getting "drastic" with overclocking doesn't feel right if a mistake will be very inconvinient and really costly.
When overclocking the X5690 I (for the first time) followed a proper procedure and learned that when my multiplier is set to 20, the BCLK of the remainder of my system will do 200. I didn't bother testing further, 200 is fine for me. With a 27 multiplier that would come to 5,4 GHz on air cooling anyways, so clearly that wasn't going to happen. I increased the multiplier and without much experimentation, 4,3 GHz seems to be a practical limit for my cooling. At this stage, the CPU will rapidly shoot up to 80° C when benched, although it does stay there and never exceeded 82°. My system is cooled with a decent air cooler, and the CPU isn't the limiting factor in my system anyways, so I feel 4...4,5 GHz would be a realistic expectation and value stability over chasing those last few, insignificant MHz. I decided to calculate which BCLK would give me the same frequency with the maximum non-turbo multiplier for the CPU and applied that (148x27) to achieve the same "result" (maybe?) but allowing for automatic speed stepping.
System
Intel Xeon X5690 3,46 GHz 6C12T (at 4,0 GHz currently)
ASUS P6T DELUXE LGA1366
ASUS nVidia GTX 660
Samsung EVO 850 SSD + some older drives
24 GB of whatever the memory is (fastest set I could find for non-exorbitant sums of money, I believe it's Corsair)
Samsung CHG90 32:9 monitor @ 3840 x 1920
If anyone has any insights as to if BCLK can allieviate my graphic card situation a bit, or any other suggestions for helping the situation a little, it'd be appreciated. Some games I play, that have in-game stat reporting shows that my CPU usage is around 30-40% max while my GPU usage seems stuck on 97-99 %.
I realize the correct option is to buy a new card, but I simply don't feel right about the current pricing due to mining; spending really big money on a GPU that's already years old seems wrong, especially in the light of a new generation supposedly being released in July along with hopefully a fix for the mining situation. I'd like something that would make this system last me till around Black Friday time maybe, and then I'll review my upgrade options then, but if I'm overlooking the obvious, feel free to point it out.
I appreciate any and all insightful input,
Nicolai
First post on here, and I know the topic practically begs for a "use the search" reply, but I have searched extensively on both Google and this site, and cannot seem to find an up-to-date answer that doesn't look like complete hearsay. I seem to find the most accurate information here though, so I thought I'd register and ask.
I can reach 4 GHz with BLCK 200 x 20 multiplier. I can also reach 4 GHz with BLCK 148 x 27 multiplier. So which is really better? There are some points to consider, that I haven't seen reflected in any other answers I've found, but since I haven't found any actual answers, I'll have to pose them as questions;
- Is it preferred to keep the multiplier on "Auto" in the BIOS, thus allowing the OS to ramp the frequency up and down as needed, to achieve a higher efficiency and lower average temperatures?
- Is it preferred to raise BCLK, even at the expense of having to lock the multiplier, to achieve a faster overall system speed?
- Am I correct in refusing the arguments against BCLK biased overclocking that I see the most, which is "introducing too many variables", "overclocking components that are not meant to be overclocked", etc.?
1. Auto Multiplier
I see two sets of behavior in Windows; if I manually set a multiplier in the BIOS, Windows will (according to CPU-Z) switch between that multiplier and 12, which I guess is a low power state. If the BIOS setting is Auto on the other hand, Windows will gradually change the multiplier and thus frequency of the CPU. I should clarify, that I did not change the SpeedStep option, but rather left that enabled all along.
- Does this gradual change provide a benefit in terms of increased efficiency and lower average temperatures? As in; does it help my electricity bill, will it result in lower temperatures even while gaming by throttling to the demand?
I would be inclined to believe that this is the case? I mean, SpeedStep was invented for a reason, I'd suppose. On the other hand, even with a locked multiplier, I do see that temperatures change with load, so clearly less power is drawn by the CPU when idle, regardless of the locked frequency?
2. Higher BCLK
If my perception on the above point is misguided, then I guess this is almost self explainatory. Anyways; A BCLK of 148 (as I currently have) is an 11 % overclock on that metric. 200 would correspond to 50 %. So am I missing out on a lot of performance increase by not going for a BCLK overclock with a low multiplier? And is that performance actually relevant? I mean; RAM isn't worthwhile overclocking in my mind, as there's plenty of it in the PC, and what else is there, really?
3. Refusing OddArguments
I followed the (excellent) three step guide to overclocking i3/i5/i7 on here. It's the first time I've overclocked "methodically" and only my second overclock ever. Still; although I'm not very experienced, I sincerely don't see the logic in the statements that a BCLK overclock is bad, because it will overclock other components, introduce variables, etc. When it's isolated "as best as it can be" in the BIOS and goes up to 200 easily, what would be the harm done, then how should it suddenly become an issue to raise it by 10 more than the default value, which is what I've seen implied over and over. Even talks about USB bus speeds being affected. I would assume this argument stems from maybe other, more finnicky chipsets or from people who just try stuff and get annoyed that it fails being a bit eager to blame the tools.
Other arguments I do see a valid point in, for instance in increasing the air temperature inside the case since other components will be overclocked and contribute to the heat build up.
But probably what I most seek clarification on is; does it matter? I understand QPI, I get how the BCLK is different from FSB (direct memory control and 3 channels was the factor in my purchase of an LGA1366 board in 2008 or thereabouts), but does it really give me a noticable performance increase anyways? I should add, that my GPU seems to definitely be the bottleneck of my system, so any (worthwhile) increase on that aspect would be welcome.
I'll drop some background in, for those who are interested or in case it's needed to answer the questions
Having always been uninterested in overclocking, I happily ran my i7 920 for years and years. Best value-for-money system I ever build, by far. In the summer or autumn of last year, it was starting to show it's age in relation to new titles though, and I though I'd try and mend that by overclocking the CPU. I did, and it worked fine. I just overclocked it very simply by raising the BCLK from 133 to 166, rebooting and enjoying. It probably shouldn't even have worked, but the system performed nice and stable for 8-9 months. I didn't feel like fine tuning anything when the result was miles better than expected - I felt I had struck gold. It also drastically improved my gaming experience, most notably The Division increasing from 20 FPS to 50 FPS, effectively turning the table so that my bottleneck went from CPU to GPU.
I now replaced my i7-920 2,66 GHz @ 3,5 GHz with a Xeon X5690 3,46 GHz currently @ 4,0 GHz. Nothing was wrong with the i7, I just felt like trying an upgrade to see if it'd help a drop in FPS I got from increasing screen resolution in games from 1920x1080 to 3840x1080 due to a new monitor. As I could find a X5690 cheap, I decided to try it, but this time with the goal of overclocking right from the start. I rely on my PC for work and play, so getting "drastic" with overclocking doesn't feel right if a mistake will be very inconvinient and really costly.
When overclocking the X5690 I (for the first time) followed a proper procedure and learned that when my multiplier is set to 20, the BCLK of the remainder of my system will do 200. I didn't bother testing further, 200 is fine for me. With a 27 multiplier that would come to 5,4 GHz on air cooling anyways, so clearly that wasn't going to happen. I increased the multiplier and without much experimentation, 4,3 GHz seems to be a practical limit for my cooling. At this stage, the CPU will rapidly shoot up to 80° C when benched, although it does stay there and never exceeded 82°. My system is cooled with a decent air cooler, and the CPU isn't the limiting factor in my system anyways, so I feel 4...4,5 GHz would be a realistic expectation and value stability over chasing those last few, insignificant MHz. I decided to calculate which BCLK would give me the same frequency with the maximum non-turbo multiplier for the CPU and applied that (148x27) to achieve the same "result" (maybe?) but allowing for automatic speed stepping.
System
Intel Xeon X5690 3,46 GHz 6C12T (at 4,0 GHz currently)
ASUS P6T DELUXE LGA1366
ASUS nVidia GTX 660
Samsung EVO 850 SSD + some older drives
24 GB of whatever the memory is (fastest set I could find for non-exorbitant sums of money, I believe it's Corsair)
Samsung CHG90 32:9 monitor @ 3840 x 1920
If anyone has any insights as to if BCLK can allieviate my graphic card situation a bit, or any other suggestions for helping the situation a little, it'd be appreciated.
Some games I play, that have in-game stat reporting shows that my CPU usage is around 30-40% max while my GPU usage seems stuck on 97-99 %.I realize the correct option is to buy a new card, but I simply don't feel right about the current pricing due to mining; spending really big money on a GPU that's already years old seems wrong, especially in the light of a new generation supposedly being released in July along with hopefully a fix for the mining situation. I'd like something that would make this system last me till around Black Friday time maybe, and then I'll review my upgrade options then, but if I'm overlooking the obvious, feel free to point it out.
I appreciate any and all insightful input,
Nicolai
