i7-820qm Turbo Boost working strange?

As mentioned HERE and HERE, the turbo boost should go to 2ghz on all four cores, but when i run OCCT and using Argos Monitor to see the boost, i only get a turbo up to 1,82ghz on all four cores and it jumps back to 1,73ghz and up to 1,82ghz again and again, but never 2ghz and when i stress only a single core it maxes out at 2,93ghz instead of 3,06ghz.
Is something wrong on my side or is this how these cpus work and Intel gave out false advertising?

 

thats not suppose to happen. although I don't hav the 820, I've seen results of it and it actrully can reach a little above 3.06.

Check your power settings. Make sure nothing is restricting your power especially in power4hydrid.

 

Yep, make sure Min/Max processor states are both set to 100% plugged in on High Performance.

 

I dont use power4hybrid and im damn sure its not the power settings, i configured everything myself in advanced power settings...

If i turn hypertherading off, it reaches 2ghz on 4 cores and a bit over 3.06ghz on a single core. Seams thats what Intel was advertising, but its just not the same with HT turned off... Kinda dissapointed. Now i see why the i7-920xm could be justified costing so much more - with HT turned on its clocks are pretty higher than a i7-820qm even if you put the unlocked multipliers aside and run it at stock clocks...

 

How did you turn off HTT?

 

Hey there,

Just did an unrelated experiment over the last few days (it was a lot of work though) and I've had some pleasnt results.

I reinstalled all of windows from scratch and installed only the very minimum for what I use my computer for. As such, power4gear has never touched my system.

Now, for the first time, intels dynamic clocks work properly. Let me explain.

Before if I set power from 5 - 100%, the computer would idle at 100%. It would only clock down if i moved the maximum down. Note that this was both with power4gear installed, and after I uninstalled.

Now, when I set it to 5 - 100 (power 4 gear never involved on this installation), the computer idles at the lowest clock speed and ramps up when it is needed, like it is supposed to.

I realize your issue is different, but I get the feeling that power 4 gear might leave behind some annoying registry enteries that mess up the cpu.

Another thing - what are your temperatures like? It won't go to max clocks if it is running too warm. HT adds a good ammount of heat, which could explain your results too...

 

thats interesting. it could be a number of things but I'm kinda interested if I'm getting max performance myself. What program is usable to test your clock speeds and temps?

 

I use the Turbo Boost Technology Meter (side gadget) and ALL CPU Meter (side gadget). First tells me what clocks second are all cores 100%. I use wPrime set at 8 threads to max out all cores.

I would really like to know how HellCry turned off HTT.

Mine willl run w/TB and all 4 cores, 720QM.

 

Turbo boost is limited by the power consumption of the processor. When you get over a predefined limit, the CPU will no longer give you any turbo boost so the CPU will drop down to the default multiplier which is 13 for a Core i7-820QM. That's the way Intel designed these CPUs to operate. The multiplier constantly changes a hundred times a second based on how many cores are in the active state so turbo boost is constantly changing.

If you run a single thread of Prime95 and go into the task manager and use Set Affinity... to lock it to a single core, this will allow that core of the CPU to reach its highest multiplier. The other 3 cores will spend most of their time in the C3/C6 sleep state.

On a Windows PC, there are constant background tasks waking up the other cores which will immediately drop the maximum multiplier. It will not stay at the full 23 multiplier for more than a few milliseconds here and there so the average multiplier will always be less than that.

Give ThrottleStop a try and post a screen shot of what it shows when running the single thread of Prime95 test. If you right click on ThrottleStop it will also show you the turbo power limits for your CPU.

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

 

I am about 100% sure because of the lack of response that HTT cannot be disabled on a G73 with an i7? Yea whatever I had no doubt, I knew it could not be done. But OP said he did so I ran with it. I would very much like to do.

Anyway guys back to the crap. Have fun. HellCry if you can do on a G73 post but I know you can't. Good luck.

 

OCCT has an option - Hypertherading that can be ticked on or off. Its that simple. If its ticked on, turbo boost wont reach 2ghz on 4 cores, but only 1.82ghz and will keep jumping to stock and back up to 1.82ghz. If its ticked off all 4 cores go to 2ghz and stay there.
My temperatures on CPU are ~45C idle with one core always 4C hotter, and at full load with OCCT and HT on, it goes to a max of 67C.
You think thats too hot and limiting turbo?

 

Instructions for disabling HT are right here.

 

"Gloria.tv: New Sodom Legalizes Same- Marriage" LOL nice find DCx

 

HellCry, when you disable hyperthreading in OCCT, does the task manager show 4 or 8 threads?

 

Should only show 4.

 

Shows all eight, but 4 arent doing anything, and processor load is exactly 50%, so all 4 cores manage to get to 2ghz (one core sometimes jumps a bit to 1,93ghz, but only for a moment).
Wish Intel explained turbo boost better, not just 2ghz on 4 cores and up to 3,06ghz on 1 core. Noone ever mentioned these turbo speeds only apply if you dont use hypertherading...

 

Shouldn't show all 8 if hyperthreading is truly disabled. Apparently it's "disabled" only for the 3rd party software you're using.

Explained better? Intel® Turbo Boost Technology in Intel® Core™ Microarchitecture (Nehalem) Based Processors

Hyperthreading is a separate technology. It does not change a turbo boosted core speed.

There seems to be a problem with older clocking apps written without an understanding of the new architectures...

 

The threads arent disabled, but they are not used, so its basically the same thing. And hyperthreading - using all 8 instead of 4 threads does affect turbo boost. Proove me wrong if you can with a screenshot of all 8 threads running at 100% and argos monitor showing max turbo boost on either a 720qm/820qm/920xm.

 

That's not possible. The two hyperthread threads do not run truly parallel for a given core (in fact they don't run in parallel at all).

This is my point in my post....there's several issues including reporting software not correctly reporting usage % and load software not able to truly load cores. So unless one knows exactly how reporting software is implemented it's really not worth worrying about it.

If you really want to call out Intel on their claims you'll need to design your own system around the processor at least down to the BIOS level. A Windows based PC isn't going to cut it - Windows is in control of thread scheduling.

Hyperthreading needs to be disabled at the BIOS level. I don't recall seeing that option on the G73's BIOS like it is on my ASUS desktop motherboard's BIOS.

 

i wish you could disable hyperthreading with this laptop.

personal preference.

 

The fact is, if i enable HT in OCCT, the cores wont go over 1,82ghz (in fact they spend more time at 1,73ghz) and if i disable it, the cores go to 2ghz (and they stay at 2ghz). Its consistent and perfectly repeatable - the results - watching frequencies or turbo boost multipliers in argus monitor are always the same. The only explanation i can come up with is that hyperthreading prevents the cpu from reaching the max specified turbo frequency, since it has to run at 100% with all threads running unlike with hyperthreading off, where half of it is freed up and its got room to go to a higher frequency.
Unless you can offer a better one i stand by this explanation.

 

How it works isn't really relevant, but at the risk of being pedantic again, turning off hyperthreading does not free up "half" of anything - It just changes the way thread scheduling is handled by the OS and possibly the monitoring apps you use.

I'm not trying to make a point about how it works...my point is just that the software reporting the speeds is inaccurate - the OS gets in the way - there's no easy way to completely eliminate all other threads and max out one core. It's close enough, but you're not going to see exact numbers. See unclewebb's posts for why. Or, study the Intel whitepapers and write your own CPU usage monitor - you'll see things are changing too rapidly to exactly monitor in real time.

 

I guess i just got used to my previous D900Fs desktop i7-950 which always had a constant frequency.
In the end as long as it delivers enough performance to run all i need i really dont have to go into it any further, i was just a bit thrown off when i saw how turbo boost on these mobile i7s works... So dynamic, multipliers just jumping up and down faster than you can track it...

 

DCx thanks for the link, very helpful.

MarkS you sound like you have a good grasp I get what you are saying.

HellCry run wPrime set on 8 threads. Oh and you just are not getting HTT and how it works. At best the application (OCCT) is not taking advantage of. Consider if this app actually produced a quantity? If it did you would find that the CPU running faster but on less load produces less "work" than slower with higher usage. If you were truly disabling HTT you would run with 4 cores and 100% usage.

As with MarkS I don't really care about the whys just the "is".

At this point no one has explained how to disable HTT on a G73, even the ridiculous juvenile Google link.

 

Moved.
10char

 

Ok, I'll bite, just WHY would you want HT off?

Are you really worried about the 1-2% overhead for hyperthreading?

Honestly, it just does not make much sense...

 

LMFAO
Yeah, well the link is supposed to be this... with a let me google this for you.

But you know. Whatever. There's the information.

 

I think the idea is that HT uses system resources that could be allocated to real cores, rather than logical ones. And since we've got 4 cores, why put them off to a logical core, right?

Well, it's not that simple. HT is like having two funnels into a single tube. So when one core is being fed information from one tunnel, it's a single core. Two funnels into one core is HT, and two funnels into two different cores is Dual Core.

So, by one theory, HT would use some TDP/TDW and lower the overall turboboost. By another theory, it's just a scheduling "aide". Since the Core I series are OOE processors HT is just helping it along by using un-used processing power while waiting for info to be written to the register.

AT LEAST, that's what I hear on the streets.

 

For the vast majority of users and usage, there is a gain to having it on.

There's no "1-2% overhead" if processes are getting 20%-30% more CPU cycles to work with (which is what I see in the multimedia software I write). The gain exceeds the overhead. It takes some very specific or incorrect multithread programming for hyperthreading to be detrimental to performance.

The only valid reasons I know for turning HT off are
1) Specific software is being used that is known to perform poorly with HT on. I can't imagine what that software would be....serious number-crunching software or old software that messes with thread affinity I would suspect.
2) An older version of Windows(pre-Vista) is being used, but even on XP it takes some poorly written software to run worse with HT enabled.

 

I did some timedemo testing with a 920 (i7) @ 4GHz with HyperThreading enabled Vs. disabled in COD4. I found a solid 10% average FPS hit having it enabled.

Testing was done with an i7 920 @ (200x20) using the 10x memory multiplyer (@ CL8). 5850 @ 1000/1300.

COD4 was a 2 and a half min timedemo taken from a full TDM server. 1920x1200 4xAA trilinear, full AF (16x if i remember right) using fraps.


I still play alot of COD4. Thus, It is relevant to my interest.