Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

All We Know About Maximizing CPU Related Performance

PunkNugget said:
Hi BDF, well I just started it up the system today to OCCT TT and all I saw was FF, so I shut down and pressed the red Clear CMOS and CMOS black button and then rebooted. I went into my BIOS and loaded up the last saved settings:

Frequency/Voltage Control (main screen)
No Dummy O.C.
Target CPU - calculated value....= cpu multi x CPU freq
CPU Freq. - 160
PCIE Freq. - 100
CPU Multi - 13
QPI - 5.8666
Memory Freq. - 1333 Mhz

Frequency/Voltage Control
CPU 0 Vcore boot - 1.325
CPU 1 Vcore boot - 1.325
CPU 0 Vcore eventual - 1.325
CPU 1 Vcore eventual - 1.325

CPU VTT both boot - 1.25V
CPU VTT both eventual - 1.25V

CPU 0 DIMM Vcore - 1.65V
CPU 1 DIMM Vcore - 1.65V
CPU 0 PLL Voltage - AUTO
CPU 1 PLL Voltage - AUTO
IOH - AUTO
CPU 0 DDR PWM Freq - 800 KHz (AUTO)
CPU 1 DDR PWM Freq - 800 KHz (AUTO)

Memory Configuration
DRAM tCL - 8
DRAM tRCD - 10
DRAM tRP - 8
DRAM tRAS - 26
Command Rate - 1

But this time I entered Vcore - 1.39375 and it still FF'd on me. Then I tried yesterday's settings before it failed (after an 1hr 13min of OCCT TT'ing). That Vcore setting was: 1.3875. It still FF'd on me. Then I went back to Vcore 1.325 and still FF'd. Then just entered in the default and it started. For some reason it won't let me even try now, don't know why. I even disconnected everything (including the battery) and cleared CMOS, waited over 5 min., then reconnected everything and still the same problem... FF. So I don't know what to do at this point... :confused: Any thoughts? Thanks...

----------



Great, now I have a better idea of what you're talking about with your settings. Thanks...



Again, I was just messin'... :p
Click to expand...

When I FF, I go down in BCLK to figure out what's going on....
 
braindeadfool said:
When I FF, I go down in BCLK to figure out what's going on....
Click to expand...

Well, I went down on the signal tweak to -50 and it started right up and just did a 2 hour OCCT TT at 1.40V and it passed. So I'm doing the same test now at 1.3875V cause I really would like to see if it can pass at a lower Vcore if possible. If it passes that OCCT TT then take two notches down even further to 1.3750V. If it doesn't pass then I'll know that 1.40V will be my setting.

If I'm doing the wrong thing with reducing the IOH QPI 0 signal tweak setting to -50 instead of the -75 that you mentioned, please let me know.

Lastly, whats the true purpose of the signal tweaking anyway and how it affects things? Does having it on -75 to -93 produce a more accurate reading in the OCCT TT than having it set at -74 to +93. Whatever you can let me know would be great, thanks... :cool:

EDIT: Well it didn't pass 1.3750V OCCT TT. So 1.3875V (@ -50 IOH QPI 0 Signal Tweak) it is. Unless of course BDF you suggest otherwise. Look forward to your reply...
 
Last edited:
Okay, so far from the Phase 3 (OCCT TT) this is what I have it set to right now:

Frequency/Voltage Control (main screen)
No Dummy O.C.
Target CPU - calculated value....= cpu multi x CPU freq
CPU Freq. - 160
PCIE Freq. - 100
CPU Multi - 13
QPI - 5.8666
Memory Freq. - 1333 Mhz

Frequency/Voltage Control
CPU 0 Vcore boot - 1.3875
CPU 1 Vcore boot - 1.3875
CPU 0 Vcore eventual - 1.3875
CPU 1 Vcore eventual - 1.3875

CPU VTT both boot - 1.25V
CPU VTT both eventual - 1.25V

CPU 0 DIMM Vcore - 1.65V
CPU 1 DIMM Vcore - 1.65V
CPU 0 PLL Voltage - AUTO
CPU 1 PLL Voltage - AUTO
IOH - AUTO (do you think that I need to put it at 1.30V or just leave it as it has been the whole time?)
CPU 0 DDR PWM Freq - 800 KHz (AUTO)
CPU 1 DDR PWM Freq - 800 KHz (AUTO)

Memory Configuration
DRAM tCL - 8
DRAM tRCD - 10
DRAM tRP - 8
DRAM tRAS - 26
Command Rate - 1

Signal Tweaks
IOH QPI 0 - -50 (do you still think that I should change this or leave it as it is?)
IOH QPI 1 - -15

The machine started up no problem today and I'm doing another Phase 3 OCCT TT (P3 OTT) for another 2 hours, just to make sure everything's working fine at these settings. So where do we go from here, do I perform a 24 hour test at these settings or something different this time? Thanks... :)

EDIT: It looks like 43 min into my P3 OTT (with my current settings above) it gave me the QUACK (error). So now I'm going to go back up (just one notch) to 1.39375V and try the P3 OTT and see it that passes... :)
 
Last edited:
voyagerd said:
For IOH QPI 0 signal, I just set it to the most negative it would go before I started getting FF from a cold boot, not a reboot.
Click to expand...

It appeared happen with me, since I started it up the next morning when it was cold. But in my case I'm needing to make sure that I am able to cold boot or a warm reboot, as I need for this thing to work no matter what in my testing, as I will be using this as a gauge for my permanent settings, when I get to that point.

BTW, thanks for the help yesterday with suggesting the Signal Tweak (ST) adjustment as it now has been working set at -50... :cool:
 
PunkNugget said:
It appeared happen with me, since I started it up the next morning when it was cold. But in my case I'm needing to make sure that I am able to cold boot or a warm reboot, as I need for this thing to work no matter what in my testing, as I will be using this as a gauge for my permanent settings, when I get to that point.

BTW, thanks for the help yesterday with suggesting the Signal Tweak (ST) adjustment as it now has been working set at -50... :cool:
Click to expand...

Just turn it off every time it reboots from the BIOS change for a few seconds. That's what I did. Once you find where it doesn't FF from booting from completely off, if it ever does get FF again, just increase the signal by one.
 
PunkNugget said:
Okay, so far from the Phase 3 (OCCT TT) this is what I have it set to right now:

Frequency/Voltage Control (main screen)
No Dummy O.C.
Target CPU - calculated value....= cpu multi x CPU freq
CPU Freq. - 160
PCIE Freq. - 100
CPU Multi - 13
QPI - 5.8666
Memory Freq. - 1333 Mhz

Frequency/Voltage Control
CPU 0 Vcore boot - 1.3875
CPU 1 Vcore boot - 1.3875
CPU 0 Vcore eventual - 1.3875
CPU 1 Vcore eventual - 1.3875

CPU VTT both boot - 1.25V
CPU VTT both eventual - 1.25V

CPU 0 DIMM Vcore - 1.65V
CPU 1 DIMM Vcore - 1.65V
CPU 0 PLL Voltage - AUTO
CPU 1 PLL Voltage - AUTO
IOH - AUTO (do you think that I need to put it at 1.30V or just leave it as it has been the whole time?)
CPU 0 DDR PWM Freq - 800 KHz (AUTO)
CPU 1 DDR PWM Freq - 800 KHz (AUTO)

Memory Configuration
DRAM tCL - 8
DRAM tRCD - 10
DRAM tRP - 8
DRAM tRAS - 26
Command Rate - 1

Signal Tweaks
IOH QPI 0 - -50 (do you still think that I should change this or leave it as it is?)
IOH QPI 1 - -15

The machine started up no problem today and I'm doing another Phase 3 OCCT TT (P3 OTT) for another 2 hours, just to make sure everything's working fine at these settings. So where do we go from here, do I perform a 24 hour test at these settings or something different this time? Thanks... :)

EDIT: It looks like 43 min into my P3 OTT (with my current settings above) it gave me the QUACK (error). So now I'm going to go back up (just one notch) to 1.39375V and try the P3 OTT and see it that passes... :)
Click to expand...

The advice on IOH QPI 0 from Shamino at EVGA is for the value to be as negative as possible and still allow the machine to boot. But the negative value required is determined by (1) what you have set IOH voltage, (2) what your set for target BCLK (CPU Freq), and (3) your indvidual board. The higher the BCLK, the lower (more negative) IOH QPI 0 needs to be. Since we are using modest BCLK with 5680 and 5690 chips, IOH and hence IOH QPI 0 don't have to be very negative. I have mine set to -75 purely for historical reasons...lol it's always worked there, but I've never tried to turn my BCLK up much past where I am.

What you pick for a "stable" test, Punk, is up to you. I chose 24 hours because when I had an unstable overclock in the past, it would often crap out at about 8-12 hours into a simulation. In general, my simulations are less demanding than IntelBurn or OCCT Linpack, so I figure if 24 hours for a torture test, I'm good. Last time I changed things around, I went 3 hours and 15 minutes in OCCT Linpack, and machine would reproducibly error at around that same point (as I recall something like 3 hours 15 minutes, 3 hours 1 minute, 3 hours 45 minutes).

based on Shamino's advice on overclocking (see the overclocking sr-2 template thread at evga.com), I think I would fix the IOH voltage, as it affects BCLK overclock stability. LOL, on the other hand the board may know better.

For CPU VCore 1.4 V should be pretty well stable for 160, plus or minus. By 165 you'll probably need 1.43 at least, if not 1.45. It's worth looking at increasing VTT to 1.35 to see if you can decrease the CPU VCore....I don't know if it will affect it or not, if you keep it under 1.35 you should be ok....if no effect on stability, then back off. I played around a little bit with that but needed to get the machine back up and running, so I'll save that for another down time.
 
Last edited:
braindeadfool said:
The advice on IOH QPI 0 from Shamino at EVGA is for the value to be as negative as possible and still allow the machine to boot. But the negative value required is determined by (1) what you have set IOH voltage, (2) what your set for target BCLK (CPU Freq), and (3) your indvidual board. The higher the BCLK, the lower (more negative) IOH QPI 0 needs to be. Since we are using modest BCLK with 5680 and 5690 chips, IOH and hence IOH QPI 0 don't have to be very negative. I have mine set to -75 purely for historical reasons...lol it's always worked there, but I've never tried to turn my BCLK up much past where I am.

What you pick for a "stable" test, Punk, is up to you. I chose 24 hours because when I had an unstable overclock in the past, it would often crap out at about 8-12 hours into a simulation. In general, my simulations are less demanding than IntelBurn or OCCT Linpack, so I figure if 24 hours for a torture test, I'm good. Last time I changed things around, I went 3 hours and 15 minutes in OCCT Linpack, and machine would reproducibly error at around that same point (as I recall something like 3 hours 15 minutes, 3 hours 1 minute, 3 hours 45 minutes).

1.4 V should be pretty well stable for 160. By 165 you'll probably need 1.43. It's worth looking at increasing VTT to 1.35 to see if you can decrease the CPU VCore....I don't know if it will affect it or not, if you keep it under 1.35 you should be ok....if no effect on stability, then back off. I played around a little bit with that but needed to get the machine back up and running, so I'll save that for another down time.
Click to expand...

Thanks for the input. I don't know how to say this but how you're explaining this info to me makes more sense as time is progressing and I'm also starting to now understand Tutor's way of explaining things to me through your added help.

Aside from that. I will keep it at -50 (unless that needs to change in the future). Now onto the other things you just mentioned. So far, after 3 hours 17 min into burn test I'm okay at 1.39375V at 160/100/13. I will will up it to 165 and then adjust the Vcore to 1.43 and VTT to 1.35 for a 20 min. OTT then drop the Vcore and VTT a notch and do another 20 min OTT and then one more time (still keeping it at 165/100/13) to see how well it does and where my sweet spot is for my system. Then I'll do a 24 hr OTT to see if it runs stable at that point.

BTW, what other tests are there that I need to perform, and if not where do we go from here? Again, thanks... :cool:
 
I've been using IntelBurnTest almost exclusively. It works well for testing Vcore. Is it a good test to find the minimum VTT and IOH voltages as well or is there something better for those?

Right now I think I'm around 1.36V Vcore and 1.25V VTT @ 160Bclk with my X5680s.
 
PunkNugget said:
Thanks for the input. I don't know how to say this but how you're explaining this info to me makes more sense as time is progressing and I'm also starting to now understand Tutor's way of explaining things to me through your added help.

Aside from that. I will keep it at -50 (unless that needs to change in the future). Now onto the other things you just mentioned. So far, after 3 hours 17 min into burn test I'm okay at 1.39375V at 160/100/13. I will will up it to 165 and then adjust the Vcore to 1.43 and VTT to 1.35 for a 20 min. OTT then drop the Vcore and VTT a notch and do another 20 min OTT and then one more time (still keeping it at 165/100/13) to see how well it does and where my sweet spot is for my system. Then I'll do a 24 hr OTT to see if it runs stable at that point.

BTW, what other tests are there that I need to perform, and if not where do we go from here? Again, thanks... :cool:
Click to expand...

Linx is pretty good at making otherwise stable overclocks crash....If your proposed method above doesn't work, I'd start with what you have working now, bump up CPU Frequency in 1 or 2 unit intervals and retest, bumping up VCore as you need for stability. You are going to be riding the edge on the stability at the VCore--I think, maybe I'm wrong, it just is very dependent from chip to chip.

I think i would also try a long run on a Graphics Benchmark like Unigine continuous....that will give the whole system a workout.

----------
voyagerd said:
I've been using IntelBurnTest almost exclusively. It works well for testing Vcore. Is it a good test to find the minimum VTT and IOH voltages as well or is there something better for those?

Right now I think I'm around 1.36V Vcore and 1.25V VTT @ 160Bclk with my X5680s.
Click to expand...

I think IBT, OCCN, Linx all are pretty good. I tend to like OCCN since it sort of is a one stop shop---it logs data, provides full voltage and temp readouts and is easy to configure. OCCN has two different test modules for CPU as well, I run both of those.
 
Okay, I started at 165/100/13 Vcore: 1.43 / VTT: 1.35 like you suggested. Passed the 20 min OTT, then I brought it down 2 notches to Vcore: 1.41875 / VTT: 1.30 and it failed. Then went up just 1 notch on Vcore: 1.425 and left the VTT: 1.30. Still failed. Then I brought the VTT up 1 notch to: 1.325 and it passed. So now I'm doing a full burn tonight with these settings:

EDIT: I took your advice stopped the current OTT and upped my IOH from AUTO to 1.30V and adjusted my ST to IOH QPI 0 back down to -75 and it started right up no problems (yet - LOL) and I'm back on with the OTT

Frequency/Voltage Control (main screen)
No Dummy O.C.
Target CPU - calculated value....= cpu multi x CPU freq
CPU Freq. - 165
PCIE Freq. - 100
CPU Multi - 13
QPI - 5.8666
Memory Freq. - 1333 Mhz
CPU Uncore Freq - AUTO (I put it back to AUTO for this P3 OTT. Is that where it should be or should I put it back to X17?)

Frequency/Voltage Control
CPU 0 Vcore boot - 1.4250V
CPU 1 Vcore boot - 1.4250V
CPU 0 Vcore eventual - 1.4250V
CPU 1 Vcore eventual - 1.4250V

CPU VTT both boot - 1.325V
CPU VTT both eventual - 1.325V

CPU 0 DIMM Vcore - 1.65V
CPU 1 DIMM Vcore - 1.65V
CPU 0 PLL Voltage - AUTO
CPU 1 PLL Voltage - AUTO
IOH - 1.30V
CPU 0 DDR PWM Freq - 800 KHz (AUTO)
CPU 1 DDR PWM Freq - 800 KHz (AUTO)

Memory Configuration
DRAM tCL - 8
DRAM tRCD - 10
DRAM tRP - 8
DRAM tRAS - 26
Command Rate - 1

Signal Tweaks
IOH QPI 0 - -75
IOH QPI 1 - -15

We're getting closer to honing in on what this thing can do more and more. Thanks again... :cool:
 
Last edited:
OTT failed on me on above settings, so I am now back up to Vcore 1.43125V / VTT: 1.35V and go for a 24 hr OTT and see what happens... :cool:
 
i tried to ZDNet overclock with the 5355 processors and Mac Pro 1,1 - no dice. do i need to update rhe efi?
 
PunkNugget said:
Okay, I started at 165/100/13 Vcore: 1.43 / VTT: 1.35 like you suggested. Passed the 20 min OTT, then I brought it down 2 notches to Vcore: 1.41875 / VTT: 1.30 and it failed. Then went up just 1 notch on Vcore: 1.425 and left the VTT: 1.30. Still failed. Then I brought the VTT up 1 notch to: 1.325 and it passed. So now I'm doing a full burn tonight with these settings:

EDIT: I took your advice stopped the current OTT and upped my IOH from AUTO to 1.30V and adjusted my ST to IOH QPI 0 back down to -75 and it started right up no problems (yet - LOL) and I'm back on with the OTT

Frequency/Voltage Control (main screen)
No Dummy O.C.
Target CPU - calculated value....= cpu multi x CPU freq
CPU Freq. - 165
PCIE Freq. - 100
CPU Multi - 13
QPI - 5.8666
Memory Freq. - 1333 Mhz
CPU Uncore Freq - AUTO (I put it back to AUTO for this P3 OTT. Is that where it should be or should I put it back to X17?)

Frequency/Voltage Control
CPU 0 Vcore boot - 1.4250V
CPU 1 Vcore boot - 1.4250V
CPU 0 Vcore eventual - 1.4250V
CPU 1 Vcore eventual - 1.4250V

CPU VTT both boot - 1.325V
CPU VTT both eventual - 1.325V

CPU 0 DIMM Vcore - 1.65V
CPU 1 DIMM Vcore - 1.65V
CPU 0 PLL Voltage - AUTO
CPU 1 PLL Voltage - AUTO
IOH - 1.30V
CPU 0 DDR PWM Freq - 800 KHz (AUTO)
CPU 1 DDR PWM Freq - 800 KHz (AUTO)

Memory Configuration
DRAM tCL - 8
DRAM tRCD - 10
DRAM tRP - 8
DRAM tRAS - 26
Command Rate - 1

Signal Tweaks
IOH QPI 0 - -75
IOH QPI 1 - -15

We're getting closer to honing in on what this thing can do more and more. Thanks again... :cool:
Click to expand...

CPU Uncore auto will go to 20x. I would keep it at 17x, or drop it to 16x even.
 
braindeadfool said:
Linx is pretty good at making otherwise stable overclocks crash....If your proposed method above doesn't work, I'd start with what you have working now, bump up CPU Frequency in 1 or 2 unit intervals and retest, bumping up VCore as you need for stability. You are going to be riding the edge on the stability at the VCore--I think, maybe I'm wrong, it just is very dependent from chip to chip.
Click to expand...

It looks like I have to up my Vcore incrementally one notch at a time as you said. Just curious, what did you mean by "... riding the edge on stability at the Vcore..."? Am I at risk of jackin' up my CPUs if I go higher on the Vcore? Let me know as I don't want to risk it and may need a suggestion for lower freq.; like maybe 162/100/13. If I'm incorrect on this and there's another way at the current setting (165/100/13) please let me know, thanks...
 
Last edited:
PunkNugget said:
It looks like I have to up my Vcore incrementally one notch at a time as you said. Just curious, what did you mean by "... riding the edge on stability at the Vcore..."? Am I at risk of jackin' up my CPUs if I go higher on the Vcore? Let me know as I don't want to risk it and may need a suggestion for lower freq.; like maybe 162/100/13. If I'm incorrect on this and there's another way at the current setting (165/100/13) please let me know, thanks...
Click to expand...

No, that's not what I implied. I meant you are running very close to (if not below) the very minimum voltages you will need to maintain a stable system. The 5690s are rated for 1.4 V (optimum 1.35), so you aren't dramatically cranking them up. As long as your temps are reasonable, you are not likely to "fry" them. For my OVERclock (as opposed to underclock) I have to run at 1.45 for 165. That's two notches above what I need for a very stable 164. YMMV, as you are underclocking---although to get the same amount of work out of the processor there are some basic laws of thermodynamics that have to be applied. You MIGHT be able to run at slightly lower voltages to get a stable underclock setup (no one has every conclusively proven that to my satisfaction), but the 5690 will require a high amount of power when cranked to 4500+ MHz, it's unavoidable.

I believe my shortlived adventure with underclocking, in which I found the "underclock" to be unstable was probably a failure to recognize this. So my frequent lockups/panics with underclocking were probably due to a VCore set too low (and possibly a VTT to set too high).

FWIW, I run at 164 24x7. Uptimes were as long as 28 days and I never have KPs. Usually only reason to reboot now is to play around a bit with the setup :)
 
braindeadfool said:
No, that's not what I implied. I meant you are running very close to (if not below) the very minimum voltages you will need to maintain a stable system. The 5690s are rated for 1.4 V (optimum 1.35), so you aren't dramatically cranking them up. As long as your temps are reasonable, you are not likely to "fry" them. For my OVERclock (as opposed to underclock) I have to run at 1.45 for 165. That's two notches above what I need for a very stable 164. YMMV, as you are underclocking---although to get the same amount of work out of the processor there are some basic laws of thermodynamics that have to be applied. You MIGHT be able to run at slightly lower voltages to get a stable underclock setup (no one has every conclusively proven that to my satisfaction), but the 5690 will require a high amount of power when cranked to 4500+ MHz, it's unavoidable.

I believe my shortlived adventure with underclocking, in which I found the "underclock" to be unstable was probably a failure to recognize this. So my frequent lockups/panics with underclocking were probably due to a VCore set too low (and possibly a VTT to set too high).

FWIW, I run at 164 24x7. Uptimes were as long as 28 days and I never have KPs. Usually only reason to reboot now is to play around a bit with the setup :)
Click to expand...

I know you just mentioned that "...no one has ever conclusively proven that to my satisfaction..." when it comes to UC'ing, but here is Tutor's BIOS set up here (on Post #70):

http://www.insanelymac.com/forum/index.php?showtopic=233891&st=60


"Here are my some of my underclocking results for my 24-7 settings of:

1. No changes in Signal Tweaks yet.

2. No changes in Memory Configuration yet.

3. Frequency/Voltage Control

Both CPU's - Without VDroop

CPU 0 Vcore boot - 1.2750
CPU 0 Vcore eventual - 1.3000
CPU 1 Vcore boot - 1.2675
CPU 1 Vcore eventual - 1.2875

CPU VTT both boot - 1.325
CPU VTT both eventual - 1.350

CPU 0 DIMM Vcore - 1.65
CPU 1 DIMM Vcore - 1.66
IOH 1.35

Everything else at default or auto for now.

4. Frequency/Voltage Control
No Dummy O.C.
Target CPU - 2.17
CPU Freq. - 167
PCIE Freq. - 103
CPU Multi - 13
QPI - 4.8
Memory Freq. - 1066

Everything else at default or auto for now.

5. CPU Configuration

Everything is Enabled and/or set to Auto, including the newly exposed C1E, except MaxCPUID which is disabled."

But just like you said (as well as Tutor) YMMV... That's the thing, I'm always thinking "...am I doing something wrong?" Since it's working so well for Tutor (regardless of my milage varying). I have the same equipment now (with the exception of the X5690; which I thought would be just slightly better than the X5680's when it comes to performance).

- Same OS (Mac OS X 10.6.7)

- Same 48GB of RAM 2000 MHz / 9-11-9-11 / 1.65V - it's a matching set of Mushkin Model #993991 (I'm also a Mushkin reseller so I got a better deal and their reviews have been top notch on this particular set of RAM, that's why I got it and they are luke-warm to the touch in their OCCT CPU: Linpack TT (which BTW, I will call from now on OCLTT). I had the Dominator Corsair GT's Red Tops (2000 MHz / 9-10-9-11 / 1.65V) but it failed the MemTest86 as one of the sticks were bad and the tech support said that you need to get matching set of RAM and doing that would have been more than twice the price. I tried using the same RAM as Tutor's Corsair XMS3's but during the Stress Tests they ran hot to the touch (as they are low profile RAM with no real heat sinks to dissipate really any kind of heat).

- Same 4890 HIS 4890 Turbo OC'd GPU that produced artifacts as a result of it running very hot, (I think something happened as a result of the tests or maybe the card just failed). Now I'm using a Water Cooled Sapphire 2GB 4890 GPU which now runs warm to the touch during OTT. So I know that I have an almost identical setup, but enough of that now as I'm just venting...

I did take your advice and I'm now upped my Vcore to 1.45V and left the VTT alone at 1.35 (as you suggested), and so far I'm an hour into my test and no QUACKING. My temps are now around - 30C Low to 76C Peak in it's testing. Also for the OCLTT. Yesterday I was used the "Infinite" test setting, as opposed to the "Automatic" test setting. I don't know if that has anything to do with the testing, but I just switched back to the (manual) "Automatic" and input 24 hours (for it's testing). I hope 76C is at an okay level as I've never seen it go that high before. The SR-2 mobo is reading 69C and the Dual Bay Res is reading 32.3C at their peak. Again, if this is okay and I can keep this test going, then let me know.

Lastly, I have 165/100/13. Should the PCIE Freq. during this P3 OCLTT stay at 100 or should it go up to 101, 102 or 103? If it just needs to stay at 100 for this particular test then I'll just leave it be. Again, thanks... :cool:

EDIT: The OCLTT just failed at 1 hr 43 min, so do you think I should up the Vcore to 1.46 and keep the VTT still at 1.35V? Remember my temps are up to 76C. If you think that bringing the CPU Freq down a notch to 164/100/13 and keep it the Vcore at 1.45 and the VTT at 1.35V then I can try that. I'll wait for your reply. Thanks again... :)
 
Last edited:
PunkNugget said:
I know you just mentioned that "...no one has ever conclusively proven that to my satisfaction..." when it comes to UC'ing, but here is Tutor's BIOS set up here (on Post #70):

http://www.insanelymac.com/forum/index.php?showtopic=233891&st=60


"Here are my some of my underclocking results for my 24-7 settings of:

1. No changes in Signal Tweaks yet.

2. No changes in Memory Configuration yet.

3. Frequency/Voltage Control

Both CPU's - Without VDroop

CPU 0 Vcore boot - 1.2750
CPU 0 Vcore eventual - 1.3000
CPU 1 Vcore boot - 1.2675
CPU 1 Vcore eventual - 1.2875

CPU VTT both boot - 1.325
CPU VTT both eventual - 1.350

CPU 0 DIMM Vcore - 1.65
CPU 1 DIMM Vcore - 1.66
IOH 1.35

Everything else at default or auto for now.

4. Frequency/Voltage Control
No Dummy O.C.
Target CPU - 2.17
CPU Freq. - 167
PCIE Freq. - 103
CPU Multi - 13
QPI - 4.8
Memory Freq. - 1066

Everything else at default or auto for now.

5. CPU Configuration

Everything is Enabled and/or set to Auto, including the newly exposed C1E, except MaxCPUID which is disabled."

But just like you said (as well as Tutor) YMMV... That's the thing, I'm always thinking "...am I doing something wrong?" Since it's working so well for Tutor (regardless of my milage varying). I have the same equipment now (with the exception of the X5690; which I thought would be just slightly better than the X5680's when it comes to performance).

- Same OS (Mac OS X 10.6.7)

- Same 48GB of RAM 2000 MHz / 9-11-9-11 / 1.65V - it's a matching set of Mushkin Model #993991 (I'm also a Mushkin reseller so I got a better deal and their reviews have been top notch on this particular set of RAM, that's why I got it and they are luke-warm to the touch in their OCCT CPU: Linpack TT (which BTW, I will call from now on OCLTT). I had the Dominator Corsair GT's Red Tops (2000 MHz / 9-10-9-11 / 1.65V) but it failed the MemTest86 as one of the sticks were bad and the tech support said that you need to get matching set of RAM and doing that would have been more than twice the price. I tried using the same RAM as Tutor's Corsair XMS3's but during the Stress Tests they ran hot to the touch (as they are low profile RAM with no real heat sinks to dissipate really any kind of heat).

- Same 4890 HIS 4890 Turbo OC'd GPU that produced artifacts as a result of it running very hot, (I think something happened as a result of the tests or maybe the card just failed). Now I'm using a Water Cooled Sapphire 2GB 4890 GPU which now runs warm to the touch during OTT. So I know that I have an almost identical setup, but enough of that now as I'm just venting...

I did take your advice and I'm now upped my Vcore to 1.45V and left the VTT alone at 1.35 (as you suggested), and so far I'm an hour into my test and no QUACKING. My temps are now around - 30C Low to 76C Peak in it's testing. Also for the OCLTT. Yesterday I was used the "Infinite" test setting, as opposed to the "Automatic" test setting. I don't know if that has anything to do with the testing, but I just switched back to the (manual) "Automatic" and input 24 hours (for it's testing). I hope 76C is at an okay level as I've never seen it go that high before. The SR-2 mobo is reading 69C and the Dual Bay Res is reading 32.3C at their peak. Again, if this is okay and I can keep this test going, then let me know.

Lastly, I have 165/100/13. Should the PCIE Freq. during this P3 OCLTT stay at 100 or should it go up to 101, 102 or 103? If it just needs to stay at 100 for this particular test then I'll just leave it be. Again, thanks... :cool:

EDIT: The OCLTT just failed at 1 hr 43 min, so do you think I should up the Vcore to 1.46 and keep the VTT still at 1.35V? Remember my temps are up to 76C. If you think that bringing the CPU Freq down a notch to 164/100/13 and keep it the Vcore at 1.45 and the VTT at 1.35V then I can try that. I'll wait for your reply. Thanks again... :)
Click to expand...

Keep the VTT at 1.35, drop CPU Freq down to 164; you will have 2 cores at 4592 and 4 at 4428 with that, that's pretty slick. Your temps are already a bit higher than I'd feel comfortable running 24x7...given the ridiculous cost of these things.

I can't find any convincing evidence that overclocking PCIE helps anything
 
braindeadfool said:
Keep the VTT at 1.35, drop CPU Freq down to 164; you will have 2 cores at 4592 and 4 at 4428 with that, that's pretty slick. Your temps are already a bit higher than I'd feel comfortable running 24x7...given the ridiculous cost of these things.

I can't find any convincing evidence that overclocking PCIE helps anything
Click to expand...

Tell me about it. I paid $1,489 each for these bad boys and don't want to jack them up. I know I'm under warranty, but these are matching CPUs and I don't want to mess with that. I will do what you mentioned; keep the Vcore at 1.45 & VTT at 1.35V. Before I do that though, I have another 120 Rad that I can fit in my case, and want to use to create a Push/Pull situation and reboot it again to see there's a temp difference that I will hopefully receive. If it gives me 5C less then I'll be happy. If not at least I tried and then I'll take it out. Thanks for the input and I'll let you know how it goes, later... :cool:
 
PunkNugget said:
Tell me about it. I paid $1,489 each for these bad boys and don't want to jack them up. I know I'm under warranty, but these are matching CPUs and I don't want to mess with that. I will do what you mentioned; keep the Vcore at 1.45 & VTT at 1.35V. Before I do that though, I have another 120 Rad that I can fit in my case, and want to use to create a Push/Pull situation and reboot it again to see there's a temp difference that I will hopefully receive. If it gives me 5C less then I'll be happy. If not at least I tried and then I'll take it out. Thanks for the input and I'll let you know how it goes, later... :cool:
Click to expand...

The 120 might or might not...I have 4 x 480s cooling mine, but use low speed fans to do it. I've been running at about 1.45 volts 24x7 for 4 months now, so far so good...
 
braindeadfool said:
The 120 might or might not...I have 4 x 480s cooling mine, but use low speed fans to do it. I've been running at about 1.45 volts 24x7 for 4 months now, so far so good...
Click to expand...

LOL !!! Only seeing 1C to 2C difference; not much change. My Rad config is now a 480, 360 & 120 Push/Pull at full 2000 RPM for the OCLTT. When I'm done with the OCLTT and figure out what I'll be setting up for my UC and I'll be lowering the RPM down to 1200 to 1300 at a much more comfortable level for me. For what you have though - WOW !!! 4 x 480's ?!?! That's a lot !!!

BTW, I just reduced the CPU Freq to 164 and it's still pulling the same exact temps 30C - 75C. So do you think I should keep it at 165 and do a 24 hour OCLTT or just keep it at 164?

EDIT: After 4.5 hrs it failed at 164. So where to go from here? Maybe I need another Rad, but only a 240 would fit and another pump. Man the Phase 2 OCLTT was so much easier. Any suggestions?
 
Last edited:
PunkNugget said:
LOL !!! Only seeing 1C to 2C difference; not much change. My Rad config is now a 480, 360 & 120 Push/Pull at full 2000 RPM for the OCLTT. When I'm done with the OCLTT and figure out what I'll be setting up for my UC and I'll be lowering the RPM down to 1200 to 1300 at a much more comfortable level for me. For what you have though - WOW !!! 4 x 480's ?!?! That's a lot !!!

BTW, I just reduced the CPU Freq to 164 and it's still pulling the same exact temps 30C - 75C. So do you think I should keep it at 165 and do a 24 hour OCLTT or just keep it at 164?

EDIT: After 4.5 hrs it failed at 164. So where to go from here? Maybe I need another Rad, but only a 240 would fit and another pump. Man the Phase 2 OCLTT was so much easier. Any suggestions?
Click to expand...

I've got a lot of radiators, but it's so it will be quiet.

I wouldn't add more radiators, most of these overclock failures are not heat related. Try turning down Uncore multiplier another notch or two, but not below 15x. If that doesn't work then turn down the QPI frequency to 4.8

Another alternative would be to just live with what you have, unless whatever task you'll be doing with it is really as intense as 4 hours+ of Linpack on 24 cores. That's pretty uncommon even in rendering, etc. You may have no problems. I had nearly 2 weeks of uptime on an overclock that failed 10 minutes into it's first Linx run....
 
braindeadfool said:
I've got a lot of radiators, but it's so it will be quiet.

I wouldn't add more radiators, most of these overclock failures are not heat related. Try turning down Uncore multiplier another notch or two, but not below 15x. If that doesn't work then turn down the QPI frequency to 4.8

Another alternative would be to just live with what you have, unless whatever task you'll be doing with it is really as intense as 4 hours+ of Linpack on 24 cores. That's pretty uncommon even in rendering, etc. You may have no problems. I had nearly 2 weeks of uptime on an overclock that failed 10 minutes into it's first Linx run....
Click to expand...

You got it. I will try it and see how it goes. Thanks... :cool:
 
Hi BDF, Tutor, (also Voyagerd and anyone else who contributed):

I don't want to put the cart before the horse but I wanted to share some encouraging news. Well, I used these BIOS settings:

Frequency/Voltage Control (main screen)
No Dummy O.C.
CPU Freq. - 165
PCIE Freq. - 100
CPU Multi - 13
QPI - 5.8666
Memory Freq. - 1333 Mhz
CPU Uncore Freq - X17

Frequency/Voltage Control
CPU 0 Vcore boot - 1.4500V
CPU 1 Vcore boot - 1.4500V
CPU 0 Vcore eventual - 1.4500V
CPU 1 Vcore eventual - 1.4500V

CPU VTT both boot - 1.35V
CPU VTT both eventual - 1.35V

CPU 0 DIMM Vcore - 1.65V
CPU 1 DIMM Vcore - 1.65V
CPU 0 PLL Voltage - AUTO
CPU 1 PLL Voltage - AUTO
IOH - 1.35V
CPU 0 DDR PWM Freq - 800 KHz (AUTO)
CPU 1 DDR PWM Freq - 800 KHz (AUTO)

Memory Configuration
DRAM tCL - 8
DRAM tRCD - 10
DRAM tRP - 8
DRAM tRAS - 26
Command Rate - 1

Signal Tweaks
IOH QPI 0 - -75
IOH QPI 1 - -15

and then enabled everything that I needed to get Mac OS X working to see if things would not only start up, but also see if all the rendering apps would work without any issues (lockups and whatnot), Well I did a GB and our new high score is 35,176 !!! To me this is a HUGE VICTORY !!! Foolish as this may sound, but my heart was racing with anticipation, hoping that thing wouldn't freeze while testing things out; praying that it would reach over 35,000 and it did !!! Also my Cinebench score is Open GL - 41.41 / CPU 22.37 !!! To me this is also amazing !!! My temps during the tests didn't go over 62C (pretty much most of the time it stayed between 30C - 57C). Only once did I see it go to 62C. I realize that I can't keep the VCore at 1.45V forever, so with some more tweaking I will work at bringing it down to 1.35V at least.

I can't thank you enough Tutor and BDF for all the time you spent (Tutor - MANY MONTHS OF PATIENCE WITH ME !!! & BDF - DAY & NIGHT THE WHOLE WEEK !) helping me through all of this stuff. I just hope that those that are willing to take the dive (challenge) into learning this, will be able to reap the rewards. Now that I understand things better when it comes to the Target CPU - calculated value....= cpu multi x CPU freq and how other things in the BIOS are affected by that has been such a big benefit for me learning this and again, I hope this helps others in the process to learn through my mistakes from this experience.

If there is anything else that you see that I need to change/tweak, please let me know. Thanks... :)

EDIT: I just did another GB & CB test (with the same BIOS; except I just upped the CPU Freq to 167) and for the new GB score: 36,085 & CB score: OpenGL: 43.64 / CPU: 22.55. Again, this is encouraging. I tried to up the CPU Freq to 170 and the apps either quit out on me or the system locked up. So now I just brought everything down a bit to this:

No Dummy O.C.
CPU Freq. - 164
PCIE Freq. - 100
CPU Multi - 13

Frequency/Voltage Control
CPU 0 Vcore boot - 1.43125V
CPU 1 Vcore boot - 1.41875V
CPU 0 Vcore eventual - 1.44375V
CPU 1 Vcore eventual - 1.43125V

Everything else is the same. I will be bringing down the Vcore when I do some more testing. BTW, when I bring down the Uncore Multi, am I not to bring it down below the CPU Multi x CPU Freq? Again, I really want to thank the both of you for all your help and everyone else as well for your contributions to making this work. I look forward to your reply with anything else you can suggest. Later... :cool:
 
Last edited:
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back