All We Know About Maximizing CPU Related Performance

[Tutor]

From what I've read you don't adjust the base clock on sandy bridge it messes up PCI frequency.

I was at about 1.4v to get there after the stress test is over I'll push it a bit more..I'd like to get to a nice stable 5Ghz but I have a feeling that I'm going to want water for that.

5Ghz should get me about 17,500 geekbench It's the cinebench I wonder about..

On some Sandy Bridge i 2700 builds, you can lower the base multiplier and raise the turbo multiplier at the same time. So, if your bios permits, lower the base multiplier from 3.5 GHz or 35 to 3.3 GHz or 33. Also, there's usually 5%-8% above and below 100 BCLK that you can adjust without adversely affecting PCI. So you could raise BCLK to 105 to 107. Then, for example, have cores allocated to turbo at 4.3, 4.4, 4.5, and 4.7, depending on your matching of their thermal capacities, yeilding, e.g.,:

(43) 4.3 x 1.07 = 4.601
(44) 4.4 x 1.07 = 4.708
(45) 4.5 x 1.07 = 4.815
(47) 4.7 x 1.07 = 5.029

Just food for thought.

I use both Geekbench 2 and Cinebench 11.5 to bench and tune my systems.

 

[GermanyChris]

From what I've been reading from other forum websites from other programers that use a variety of CPUs and Mac OS X, weather using the Mac Pro or a Hackintosh setup (and I could be incorrect), utilizing Cinebench as your rendering "standard" for scoring isn't as reliable anymore as Maxon hasn't made any updates to ver. 11.5 since November 23, 2009. Whereas GeekBench has been dedicated to their own updates as a result of OS software changes from Mac OS X (SL to Lion) and Windows (XP and 7). It still may be used for overall testing, but I tend to default on GB as my rendering standard over Cinebench. This is just what I've been reading and asking others out there have also have SR-2 setups. If I'm incorrect on this Tutor (and you may know better), what's your thoughts on this if you have any?

Here is the GB and CB @4.9…

This is as far as I'm going on air, it may be as far as I'm going..I dunno yet

 

[SR2Mac]

Here is the GB and CB @4.9…

This is as far as I'm going on air, it may be as far as I'm going..I dunno yet

Wow!!! You're getting 17,537 and that's a 4 core. I'm OC'ing with my W3680 (6 core) and I'm reaching 18,500+. Can't wait to start UC'ing with this CPU to see if I can reach higher GB scores... But first I have to get everything working fine with my SR-2 setup first...

BTW Tutor, any thoughts to my last comment when it comes to using Cine 11.5 and it's possible reliability issues when it comes to render testing? Just curious on your thoughts on this... Thanks...

 

[GermanyChris]

On some Sandy Bridge i 2700 builds, you can lower the base multiplier and raise the turbo multiplier at the same time. So, if your bios permits, lower the base multiplier from 3.5 GHz or 35 to 3.3 GHz or 33. Also, there's usually 5%-8% above and below 100 BCLK that you can adjust without adversely affecting PCI. So you could raise BCLK to 105 to 107. Then, for example, have cores allocated to turbo at 4.3, 4.4, 4.5, and 4.7, depending on your matching of their thermal capacities, yeilding, e.g.,:

(43) 4.3 x 1.07 = 4.601
(44) 4.4 x 1.07 = 4.708
(45) 4.5 x 1.07 = 4.815
(47) 4.7 x 1.07 = 5.029

Just food for thought.

I use both Geekbench 2 and Cinebench 11.5 to bench and tune my systems.

I've backed back down to 4.7..

I'm going to give this some thought..maybe a test or two later on this week my only concern is I boot of a PCI SATA card and I'd like to not corrupt my very nicely set up SSD..

----------

Wow!!! You're getting 17,537 and that's a 4 core. I'm OC'ing with my W3680 (6 core) and I'm reaching 18,500+. Can't wait to start UC'ing with this CPU to see if I can reach higher GB scores... But first I have to get everything working fine with my SR-2 setup first...

BTW Tutor, any thoughts to my last comment when it comes to using Cine 11.5 and it's possible reliability issues when it comes to render testing? Just curious on your thoughts on this... Thanks...

can you imagine 16 sandy bridge cores..wow..The new MP is going to be fast..

i7 2700k drop in geekbench was 13,000ish time 4 for the MP

 

[SR2Mac]

I've backed back down to 4.7..

I'm going to give this some thought..maybe a test or two later on this week my only concern is I boot of a PCI SATA card and I'd like to not corrupt my very nicely set up SSD..

----------



can you imagine 16 sandy bridge cores..wow..The new MP is going to be fast..

i7 2700k drop in geekbench was 13,000ish time 4 for the MP

From what I'm gathering for info from others about using the SR-X (if you're referring to the new Dual CPU from EVGA and the newer CPUs that can be used with it) won't give you a ton of advantage over the SR-2, about 10%-17% boost. But we'll see what happens when someone builds a hackie using and SR-X setup and their CPUs that will go with that system...

 

[GermanyChris]

From what I'm gathering for info from others about using the SR-X (if you're referring to the new Dual CPU from EVGA and the newer CPUs that can be used with it) won't give you a ton of advantage over the SR-2, about 10%-17% boost. But we'll see what happens when someone builds a hackie using and SR-X setup and their CPUs that will go with that system...

Here is 5Ghz 1.5v We have most defiantly hit diminishing returns. It looks like 17875 is the fastest 2700K right now.

 

[Tutor]

Here is 5Ghz 1.5v We have most defiantly hit diminishing returns. It looks like 17875 is the fastest 2700K right now.

And that's really quite a 4-core achievement, especially without H2O or LN.

----------

...

BTW Tutor, any thoughts to my last comment when it comes to using Cine 11.5 and it's possible reliability issues when it comes to render testing? Just curious on your thoughts on this... Thanks...

Cinebench is as real world as you can get when it comes to benching rendering prowess.

 

[Tutor]

Here is 5Ghz 1.5v We have most defiantly hit diminishing returns. It looks like 17875 is the fastest 2700K right now.

It appears that you're quicking reaching the overclockers' barrier. What I'm suggesting is that you try to underclock and turbo bias your Sandy Bridge. This is what I've done to all of my Westmeres. The more you can underclock it the less Vcore you'll have to feed it and if you can turbo bias it, it will outperform a straight overclock, yeilding the best of both worlds, because you can get fewer cores to run faster that you can get them all to run, with less heat (yeilding longer CPU life and reduced total electricity charges). Hopefully, that 1.5 Vcore will be a thing of the past and both your Cinebench 2 and Geekbench scores will shoot out of the roof.

 

[SR2Mac]

And that's really quite a 4-core achievement, especially without H2O or LN.

----------



Cinebench is as real world as you can get when it comes to benching rendering prowess.

Okay. Again, that's what I've seen by others sharing all the ins and outs of how conflicting Cinebench works on their own rendering testing, but you would be the person that I would default to understanding that better. That's why I asked...

BTW, I PM'd you...

 

[GermanyChris]

I'm going to spend this week in windows so I can get the processor, video card and ram tweaked with proper tools.

The goal is 18,200 or so..

 

[GermanyChris]

Here we are 17,793 1.455v 5Ghz.

It stays under 90 on air with intel burn..

I think we'll add some water, to keep it cool and go with it..

 

[SR2Mac]

Here we are 17,793 1.455v 5Ghz.

It stays under 90 on air with intel burn..

I think we'll add some water, to keep it cool and go with it..

That's great. I'm currently using Corsair's H100 at 4.2GHz 24/7 and it's at 31C in a 79F room... If you were using an H80 or H100 I bet you it would be around 55C-60C tops...

Keep up the great work !!!

 

[GermanyChris]

That's great. I'm currently using Corsair's H100 at 4.2GHz 24/7 and it's at 31C in a 79F room... If you were using an H80 or H100 I bet you it would be around 55C-60C tops...

Keep up the great work !!!

I just broke 18,000 setting my RAM up to 1866

All my low power setting work and 18000, I'll hit 18,200 on water no problem.

 

[SR2Mac]

I just broke 18,000 setting my RAM up to 1866

All my low power setting work and 18000, I'll hit 18,200 on water no problem.

This boy is on a roll !!! You go Chris !!!

 

[GermanyChris]

This boy is on a roll !!! You go Chris !!!

is (yes> /dev/null) a good stress test?

the one stress test I can find crashes the system no matter the clock.

----------

On some Sandy Bridge i 2700 builds, you can lower the base multiplier and raise the turbo multiplier at the same time. So, if your bios permits, lower the base multiplier from 3.5 GHz or 35 to 3.3 GHz or 33. Also, there's usually 5%-8% above and below 100 BCLK that you can adjust without adversely affecting PCI. So you could raise BCLK to 105 to 107. Then, for example, have cores allocated to turbo at 4.3, 4.4, 4.5, and 4.7, depending on your matching of their thermal capacities, yeilding, e.g.,:

(43) 4.3 x 1.07 = 4.601
(44) 4.4 x 1.07 = 4.708
(45) 4.5 x 1.07 = 4.815
(47) 4.7 x 1.07 = 5.029

Just food for thought.

I use both Geekbench 2 and Cinebench 11.5 to bench and tune my systems.

I tried this last night, I cant adjust individual clocks..

The nice part is though when there is no load the CPU drops to 1.6Ghz but it would be nice if there was a step between 1.6 and max clock

 

[Tutor]

is (yes> /dev/null) a good stress test?

the one stress test I can find crashes the system no matter the clock.

----------



I tried this last night, I cant adjust individual clocks..

The nice part is though when there is no load the CPU drops to 1.6Ghz but it would be nice if there was a step between 1.6 and max clock

I'm suggesting setting the individual turbo limits which most Sandy Bridge motherboards that I've seen, allow. Can you post pic(s) of turbo features found in bios?

 

[SR2Mac]

is (yes> /dev/null) a good stress test?

the one stress test I can find crashes the system no matter the clock.

Well, you can use it. In fact that is really the only reliable stress test that you use for Mac OS X. In my past experience using Prime95 for Mac over stresses certain parts on your Mac (usually the RAM) and so that is why I prefer to use this link:

http://macfidelity.de/2009/05/10/mac-how-to-stress-test-your-cpu-in-mac-os-x/index.html

Going over this with Tutor, his thoughts are to still use Prime95 BUT only use it in Windows when you're stress testing as he has found this to be the best route.

If you don't have Windows and still need to stress test your CPU using Mac OS X with Terminal typing in: yes> /dev/null for Mac OS X for stress testing is the best route, as I can do this to my current Xeon W3680 without lockups or failures.

Again, if you choose this route, you are going to have to open up as many pages per "virtual" core to achieve the full stress load for your CPU. What I mean by this is, if you have let's say 6 cores (12 virtual) on your CPU, then you're going to have to open the Terminal app and type in: yes> /dev/null on each page, then hit enter, open a new Terminal App page and do it for a total of 12 times.

DO NOT go over the amount of Terminal Pages needed for your virtual cores as it will impede the full stress load and actually lower the stress load test. I found this out by using iStat to test this theory out and it holds true, you can't go beyond adding more terminal app pages using: yes> /dev/null to achieve more stress; only the virtual core number for your CPU.

Again, if you have no choice but to use Mac OS X, this is your best choice for now as I have yet to find any app out there that is reliable enough to stress test your CPU using Mac OS X.

But Tutor's input and experience is more reliable if you want to stress test your system doing it in Windows using Prime95...

 

[GermanyChris]

I'm suggesting setting the individual turbo limits which most Sandy Bridge motherboards that I've seen, allow. Can you post pic(s) of turbo features found in bios?

I can read them, type in a value but it doesn't change anything..

I used all setting except disable turbo to try..

It's a P67 Mother Board, I wonder if my BIOS is old?

I'll take a picture when I get back to the apartment and hopefully it's still running it's been 4hrs of yes> /dev/null

----------

Well, you can use it. In fact that is really the only reliable stress test that you use for Mac OS X. In my past experience using Prime95 for Mac over stresses certain parts on your Mac (usually the RAM) and so that is why I prefer to use this link:

http://macfidelity.de/2009/05/10/mac-how-to-stress-test-your-cpu-in-mac-os-x/index.html

Going over this with Tutor, his thoughts are to still use Prime95 BUT only use it in Windows when you're stress testing as he has found this to be the best route.

If you don't have Windows and still need to stress test your CPU using Mac OS X using: yes> /dev/null for Mac OS X for stress testing is the best route, as I can do this to my current Xeon W3680 without lockups or failures.

Again, if you choose this route, you are going to have to open up as many pages per "virtual" core to achieve the full stress load for your CPU. What I mean by this is, if you have let's say 6 cores (12 virtual) on your CPU, then you're going to have to open the Terminal app and type in: yes> /dev/null on each page then hit enter open a new Terminal App page and do it 12 times total.

DO NOT go over the amount of virtual cores as it will impede the full stress load and actually lower the stress load test. I found this out by using iStat to test this theory out and it holds true, you can't go beyond adding more terminal app pages using: yes> /dev/null to achieve more stress; only the virtual core number for your CPU.

Again, if you have no choice but to use Mac OS X, this is your best choice for now as I have yet to find any app out there that is reliable enough to stress test your CPU using Mac OS X.

But Tutor's input and experience is more reliable if you want to stress test your system doing it in Windows using Prime95...

Yes, I've got 8 running..

I messed my Windows install up and don't much feel like fixing it right now and to top it is boot up is rather slow I actually spend more time booting than adjusting and testing..

I'm trying to back voltages down a bit, to keep the fans from becomming obnoxious..

1.425v for 5Ghz wasn't really where I wanted to be, especially since I can run 4.8 @ 1.35v after playing some more. Ideally I'd like to be at 1.38-1.39 @ 5Ghz

 

[SR2Mac]

I messed my Windows install up and don't much feel like fixing it right now and to top it is boot up is rather slow I actually spend more time booting than adjusting and testing..

Sorry to hear that. How long is your boot time taking? Shouldn't be any more than a minute...

 

[GermanyChris]

Sorry to hear that. How long is your boot time taking? Shouldn't be any more than a minute...

I never times it..but it boots to the SATA cards first..so it needs to cycle there first.

One of the SATA cards carries a RAID 0 set, and another carries my boot drive, but my home folder is on a second drive on that same card..

When I was playing with BLCK the first night I corrupted the SSD/HDD combo balance and I got it put back together but I don't love the command line..

So now all Apple drives come out completely when tweaking in Windows I need those drives stable..

 

[SR2Mac]

I never times it..but it boots to the SATA cards first..so it needs to cycle there first.

One of the SATA cards carries a RAID 0 set, and another carries my boot drive, but my home folder is on a second drive on that same card..

What ??? Sorry, I was just thrown off there a bit by reading your set up. I had to read it a few time to fully understand it. Maybe I'm missing out on the benefits of your current config, but why is it set up this way?

When I was playing with BLCK the first night I corrupted the SSD/HDD combo balance and I got it put back together but I don't love the command line..

You don't have to love it, it just has to work.. Tutor any thoughts? I really don't know much about how to work out a solution to this type of config as my setup is pretty typical - 1 SSD as my main boot drive and my 600GB Velociraptor along with my 2TB HDs as my backup drives.

 

[Tutor]

I can read them, type in a value but it doesn't change anything... . I'll take a picture when I get back to the apartment
... .

Don't mean to be a pain, but can you place the focus on those values and increase them with the plus and minus keys? I only mention this option because in instructing others over time whose bios setting aren't directly in front of me I've learned that was an option that they could have used, but failed to employ, after their having told me initially that the values couldn't be changed. Some others of your may know what I mean from first hand experience.

 

[GermanyChris]

What ??? Sorry, I was just thrown off there a bit by reading your set up. I had to read it a few time to fully understand it. Maybe I'm missing out on the benefits of your current config, but why is it set up this way?



You don't have to love it, it just has to work.. Tutor any thoughts? I really don't know much about how to work out a solution to this type of config as my setup is pretty typical - 1 SSD as my main boot drive and my 600GB Velociraptor along with my 2TB HDs as my backup drives.

LOL OK so here we go

I have a total 10 SATA ports 6 SATA III and 4 SATA II

on the 4 SATA II ports are 3 optical drives and one WD Green 3TB

2 SATA III ports are on the board 1x1TB Seagate (Windows) 1 2TB Black

4 More SATA III via PCI on the x8 is the SSD and 1TB Black that has my home folder the other card on the x1 lane has a striped RAID set of 2x 250 Hitachi's it serves a scratch/junk collector/.dmg safe house etc. they're the computer equivalent of a junk drawer..

I have a lot of drives that I might start minimizing into fewer but they work and new drives cost money, and a fair amount of it right now.

----------

Don't mean to be a pain, but can you place the focus on those values and increase them with the plus and minus keys? I only mention this option because in instructing others over time whose bios setting aren't directly in front of me I've learned that was an option that they could have used, but failed to employ, after their having told me initially that the values couldn't be changed. Some others of your may know what I mean from first hand experience.

OK, I got them changing but I can only change 3 of the 4..which puts the max 4 core turbo back to 3.9..Strange

 

[Tutor]

OK, I got them changing but I can only change 3 of the 4..which puts the max 4 core turbo back to 3.9..Strange

Of the three that you can change, what can you set them to? Does it vary based on the setting for the base frequency multiplier?

 

[GermanyChris]

Of the three that you can change, what can you set them to? Does it vary based on the setting for the base frequency multiplier?

Nope I can set them all to 50, until I get to the 4th then it ends..

From what I've been reading this is normal for p67 gigabyte motherboards

I hope I'm not stuck with a trigger, So I'll do some more research over the weekend