i7 Turbo - Tool to test for Turbo throttling

[unclewebb]

Edit: I think the i7 Turbo bug that HW Monitor was exposing in Core i7 systems has finally been fixed.
Check out version 6.0 below to test for any problems.

[unclewebb]

http://www.fileden.com/files/2008/3/3/1794507/Turbo.zip

-Added the operating system activity to the CPU Load % results.

-This reduces or eliminates the interference that some third party monitoring programs were causing to the i7 Turbo reported results.

[rge]

Realtemp 3.30 (with newest turbo code) works perfectly for always giving the correct multi, even while running other utilities that dont share well, and it catches turbo throttling as soon as it occurs.
http://www.fileden.com/files/2008/3/...mpBeta_330.zip

Using the old Gigabyte bios F3, only GB bios that has TDP limits for turbo, you can watch turbo throttling as soon as it begins.

Early throttling caught by RT, not yet by cpuz down to 23.8 from 24. My i950 starts to throttle almost right at 100W tdp.



Again throttling to 23.6, not yet caught by cpuz.



Complete throttling/turbo disengaged by 120W, now shown by both cpuz and RT

[Papagaj]

Please can anyone help me if it's possible to run more TDP on rampage 2 extreme? Yesterday i found out that in linx its changing my multiplier from 26 to something like 25.425....

[xpatar]

I've got the same problem, ocing a 920 C0/C1 @ 4.2GHz throttling begins:

@ idle


Multis stays @ 21 while cpu consumption is @ 30.88 W


@ full load running prime small ffts


Multis drop @ 20 while cpu consumption is @ 125.50 W


Turbo log file:


Running on ASUS P6T DELUXE V2 bios v.0504

[Papagaj]

lol nice load temps

[xpatar]

lol nice load temps

Temps really worry me, today I just received my Watercool HK CPU LGA1366 Rev. 3.0 LT and XSPC RX360 Triple Radiator. I'm gonna connect them and see how well they perform.

[Papagaj]

well that was kinda sarcasm m8.... 92 is way too much

[unclewebb]

xpatar: You're screwed. The Asus P6T is the Turbo throttling champion.

Turbo throttling can be caused by excessive core voltage, excessive current or when your core temperature gets too high. It looks like you have all three problems.

Even if you get your core temperatures 10C lower, there are still the other two factors that are likely going to cause Turbo throttling at the 4.2 GHz speed that you would like to run at.

It doesn't seem that Asus is ever going to fix this issue so if you want to run over 4 GHz then you will have to get another motherboard. The next board I buy won't be an Asus and I know there are a few more users thinking the same thing because of this issue.

Papagaj: I thought there was a setting in the bios to turn off Turbo throttling on the rampage 2 extreme.

[xpatar]

Thanks for the reply unclewebb, let's hope that this issue is going to be fixed soon, otherwise I'm going for a new mobo.

[003]

Papagaj: I thought there was a setting in the bios to turn off Turbo throttling on the rampage 2 extreme.

There definitely should be. It was added in a BIOS update iirc, so he should update to the latest BIOS.

[pangingIII]

Praz: I wrote a small tool to try to help out with the DFI 12 multi issue. It's called MultiBoost.

http://www.fileden.com/files/2008/3/...MultiBoost.zip

I was just looking for some feedback. When your CPU is stuck at 12, does MultiBoost do anything to your multiplier when you are running Prime? Try it with EIST enabled or disabled in the bios. You could also try toggling C1E with i7 Turbo while MultiBoost is running.

Any feedback I get might help me find a fix for this problem if it's possible.

Edit: One user had some success.
http://i4memory.com/115485-post45.html

Thanks unclewebb, the tool is actually working
without it as soon as i get to Win 7 desktop i would get stuck at multi 12x
after running your program it's able to stay at multi 21x (even 22x will pop up sometimes) for most of the time.
What is your program doing, could you explain a bit about it? thanks.



DFI UT X58
Thermal Management Control: Enabled
EIST Function: Enabled
CxE Function: AUTO

[unclewebb]

pangingIII: What motherboard are you using? This was mostly a DFI issue but I think some MSI boards also have this problem where the multiplier gets stuck at 12 at full load in Windows 7.

Edit: I see you have the DFI.

Run i7 Turbo to see what it reports when CPU-Z is showing the 22X multi. That might not be 100% accurate. rge and I are going to be doing some more testing in the next day or so to try to prove this one way or the other.

What is your program doing, could you explain a bit about it?

I'd rather not say.
I can't give away my secrets. I'll send you a PM.

Edit: You don't like PM messages so I'll come clean. MultiBoost changes a register within the CPU. For some unknown reason this register is getting set to 12 on the DFI boards so adjusting that one register to 21 allows your CPU to run at full speed, including its Turbo feature. A very simple patch until DFI or Windows 7 comes up with the correct fix.

I built MultiBoost into the latest version of RealTemp 3.30.
Just add this to your RealTemp.INI file.

MultiBoost=1

One less program to run that way.

The low core voltage in your screen shot shows that EIST is enabled and has lowered your core voltage which means it has also lowered your multiplier. I'm 99.99999% sure that CPU-Z is not accurate at idle.

[Kougar]

Unclewebb, I'm totally loving your utilities. I've been trying to find a tool to monitor CPU throttling but none seemed to work with either this CPU or this 64bit OS. Thanks for another great program!

I do have a quick question, or rather two: What is the throttling limit for Core i7 D0's, I had thought it was 80c but is it actually 100c? Intel's page says 67.9°C, but that's not the core temp value they are giving.

If the CPU begins to throttle back will your i7 Turbo utility detect this? I'm thinking Realtemp would under the Thermal Status field, but after forcing temps to 82c the multi is still at 21 as reported by your tool and no throttling is being reported either. Just wanting to make sure I'm following everything correctly.

[unclewebb]

Unclewebb, I'm totally loving your utilities.

You're most welcome. I was just working on a slightly different version of my i7 Turbo tool today. This one is the V8 version. It will show the calculated multiplier for each thread. After rge does some testing I'll think about releasing it. CPU-Z sometimes shows a +2 Turbo with all 4 cores enabled in the bios and I'm not sure how accurate that is. I'm hoping this new tool can be used to figure this out.

On my Q6600 it's able to show that when running a single threaded app on core 0, core 1 which is joined with core 0 has the same multi but it's two friends for core 2 and core 3 are trying to idle down. I never knew that!



The 67.9°C thermal spec that Intel released is a Tcase temperature which is a temperature measured with a probe located at the geometric center of the IHS. When running a high stress program like some Linpack testing, the theory is that as long as the Tcase stays below this temperature then a user should rarely have to worry about thermal throttling during normal operation. The core temp will approach 100C at full load when the Tcase is at this value.

This spec is designed for system builders. This allows them to design cases and select heatsinks and decide how many fans they are going to need to keep the CPU under that temperature.

Thermal throttling is all controlled by the core temperature sensors. With Core i7, they seem to start thermal throttling right at 100C. The previous Core 2 CPUs that I tested would start a couple of degrees sooner. When thermal throttling starts, that will show up in the RealTemp Thermal Status area and your multiplier will start to rapidly cycle down to 12.

Turbo throttling is different. There are a combination of factors and temperature is just one thing. I think Turbo throttling can start to happen at about 80C to 85C. I don't have enough data about this yet. There may not be an exact temperature number. It also depends on core voltage and current going through the CPU so it might depend on those two as well even when the temps are high.

i7 Turbo and RealTemp 3.30 should both show any Turbo or thermal throttling. The Calculated Multiplier should be the same in both programs now. RealTemp has been using this method since Core i7 was introduced but I recently refined it a little so it shows multiplier steps of 0.2 instead of 0.5 to make sure that it can catch any problems a little sooner.

[Kougar]

Thanks for the reply!

On my Q6600 it's able to show that when running a single threaded app on core 0, core 1 which is joined with core 0 has the same multi but it's two friends for core 2 and core 3 are trying to idle down. I never knew that!

Interesting... very interesting. I never would have suspected that either!

The 67.9°C thermal spec that Intel released is a Tcase temperature which is a temperature measured with a probe located at the geometric center of the IHS. When running a high stress program like some Linpack testing, the theory is that as long as the Tcase stays below this temperature then a user should rarely have to worry about thermal throttling during normal operation. The core temp will approach 100C at full load when the Tcase is at this value.

So suffice to say the motherboard probe under the CPU socket would be a close temperature, but not this specific one. If the Tcase temp is measured at the surface of the IHS, then methods like watercooling would slightly delay the onset of throttling as this area is cooled better vs air coolers?

Turbo throttling is different. There are a combination of factors and temperature is just one thing. I think Turbo throttling can start to happen at about 80C to 85C. I don't have enough data about this yet. There may not be an exact temperature number. It also depends on core voltage and current going through the CPU so it might depend on those two as well even when the temps are high.

Heh, doesn't get much more confusing than that! Here's some data then, no CPU no Turbo throttling at all with >80c on three cores that I can see. Already set the fans to 5V and have the overvolted/overclocked GPU dumping heat into the watercooling loop to create a worst case scenario. I've never seen it reach 83c before.



Edit: Heh, just noticed I'm using 3.20v, will have to fix that.

[gatecrasherlok]

Nice tool Thanks alot.

Never knew x58's had throttling problems with HT, will try and learn more about this subject and try and get to learn how to use this program.

Thanks again

[Papagaj]

There definitely should be. It was added in a BIOS update iirc, so he should update to the latest BIOS.

Since Im using 975EE i have the latest bios which is 1406.... can't find it there( How is it called exactly please and where is it?

/edit never mind got it.... it was right at the main screen

[phsinc1]

is that mean throttling? min 20.996

[unclewebb]

is that mean throttling? min 20.996

Usually not. These timers are very precise but sometimes they are not 100% accurate. The best thing to do is to turn on the logging feature and then run Prime95 or LinX and when you're done go back and have a look at the log.

The 20.996 likely occurred during a quick transition from idle to full load.

Kougar: Less core voltage equals less chance of Turbo throttling. Watts is proportional to the square of voltage so a 5% increase in voltage is a more than 10% increase in CPU power consumption and will get you closer to the Turbo throttling point faster twice as fast as a 5% increase in MHz.

rge has been doing some testing and it looks like the newest version of i7 Turbo is going to show when one core jumps up to a +2 Turbo ratio even with 4 cores enabled. Stay tuned for some interesting results.

[rge]

unclewebb and I have been trying to determine when +2 turbo (25 on i950 or 22 on i920) is active. Turning off all cores but 1 in bios is one way of seeing +2.

Another way to get +2 turbo is enabling c3/c6 in bios with all cores active. (C1e/eist can be enabled or disabled). And at least on my bios/mobo, you have to run at stock mhz for this to work.

By running spi, can prove the multi is indeed 25 when all 4 cores are active in bios just asleep via c3/c6 except active one, here are pics with unclewebbs V8 turbo.

pic 1, all cores active in bios with c3/c6 enabled, the loaded core goes near 25, and in proportion to the load.


Using task manager to force spi to stay on thread 0,1 ie core 1, it tends to stay around 24.5 to 24.6...but dont know if v8 program running is inhibiting other cores from fully sleeping, cpuz will bounce back between 24 and 25. I tried running v8 on thread 0 and spi thread 1, but never saw full 25, but still get up to 24.6.


Here is with 1 core active in bios, still c3/c6 enabled....locks on 25.


spi proof of 25 multi at stock with all 4 cores active in bios and c3/c6 enabled
25x133 (1 core active in bios) spi with c3/c6 enabled for apples to apples comparison. 4 runs spi 1m
12.449
12.449
12.480
12.449


Next up 24x133, ie, turbo running spi with all cores active in bios, c3/c6 disabled, so turbo is limited to 24.
12.808
12.824
12.807
12.823
12.808


Next up, 24x133, ie turbo running but with c3/c6 ENABLED, to allow 25 multi to show up...and spi proves it is indeed same speed as proven 25 multi with 1 core active in bios.
12.449
12.386
12.433
12.434


So definitely getting 25 multi with all cores active in bios as long as stock mhz and c3/c6 active.

One other point, enabling c1e, eist, or c3/c6 does slow down spi slightly which is why you have to run 1 active core in bios with c3/c6 enabled for fair comparison.

[ben805]

xpatar: You're screwed. The Asus P6T is the Turbo throttling champion.

+1

It doesn't seem that Asus is ever going to fix this issue so if you want to run over 4 GHz then you will have to get another motherboard. The next board I buy won't be an Asus and I know there are a few more users thinking the same thing because of this issue.

ditto on that man.

[pangingIII]

pangingIII: What motherboard are you using? This was mostly a DFI issue but I think some MSI boards also have this problem where the multiplier gets stuck at 12 at full load in Windows 7.

Edit: I see you have the DFI.

Run i7 Turbo to see what it reports when CPU-Z is showing the 22X multi. That might not be 100% accurate. rge and I are going to be doing some more testing in the next day or so to try to prove this one way or the other.

I'd rather not say.
I can't give away my secrets. I'll send you a PM.

Edit: You don't like PM messages so I'll come clean. MultiBoost changes a register within the CPU. For some unknown reason this register is getting set to 12 on the DFI boards so adjusting that one register to 21 allows your CPU to run at full speed, including its Turbo feature. A very simple patch until DFI or Windows 7 comes up with the correct fix.

I built MultiBoost into the latest version of RealTemp 3.30.
Just add this to your RealTemp.INI file.

MultiBoost=1

One less program to run that way.

The low core voltage in your screen shot shows that EIST is enabled and has lowered your core voltage which means it has also lowered your multiplier. I'm 99.99999% sure that CPU-Z is not accurate at idle.

Thanks for the info, and again for your programs.

[xpatar]

Has anyone any info about turbo throttling on D0s.
I mean, when does it start (cpu voltage, cpu temp)?

[Brama]

+1




ditto on that man.

Saying the truth, Asus fixed it on Rampage II Extreme but seems to refuse fixing it in P6T series.