My CPU does not reach the max turbo boost frequency, even run single thread app?

My laptop is Toshiba P50.
CPU: i7-4700MQ
Memory: 16G
Graphic: dedicated NVidia 740M
OS: Windows 7 Enterprise


I run SuperPI 1.9, that is single thread program. The CPU only reaches to about 3 Ghz. The superPI 1M is 13.9s.
The laptop was plugged in/ chargering.

Is it supposed to bump up to 3.4 G hz?




Thanks.

 

Well, you do have an OS running there as well, you will never attain the max single threaded frequency in my experience, simply because you are running the OS as well and it will split the load to two cores (one for superPI) and one or more for Windows. You could also behitting the max TDP a tad under 3.4 GHz and that would limit the max frequency as well.

In any case, the bottom line is that if you are at ~100 MHz or so from max frequency, it's about as good as it's gonna get.

 

If you want to attain (as close as possible to the) max single speed, you have to kill as many processes as feasible. I think I had to bring the processes down to below 60 before mine would reluctantly run the max single core frequency, and even then it would fluctuate 100 MHz or so.

 

Thanks.
After closed all of applications, the CPU reached the max frequency.

 

13.9 seconds is pretty bad for that CPU, what time do you get now?

Here's one I did on the 4700MQ, you can use for comparison.

 

^dude that's a new record for 4700MQ, should totally submit that to HWBot.

 

Funny you should mention that, because that used to bother me too for a long time. Then one day I logged on as administrator and notice after being idol for a bit it would cycle right to max and hold there. Running some diagnostic or virus scan I suppose. Otherwise, it always just under the max during HD operations, but only by one point.

 

Keep in mind the turbo function is NOT intended to run continuously. The thermal design won't allow for it. It will run at turbo when ever all the factors that allow for it are met, and one of those factors is cpu temp.

Turbo is not intended to make single core apps faster, it is intended to temporarily increase cpu speeds for a short duration when ever cpu utilization spikes. The ideas is it wants to try and muscle through a process before it feels waking up the second core (which in turn draws more power, and time) to get the current workload done.

 

As long as the CPU is not reaching the thermal throttling temperature at 100C, Turbo Boost will be running continuously. If the CPU goes over the TDP limit then the amount of Turbo Boost will get throttled back but the CPU will still be using some Turbo Boost. Intel designed their 4th Gen CPUs so they rapidly cycle Turbo Boost on and off just enough to keep the CPU under its TDP limit.

Turbo Boost makes single threaded apps run much faster and even when a 4700MQ has all 8 threads fully loaded with the most demanding of applications, a 4700MQ will continue to use some Turbo Boost.

I have always used the SuperPi 1.5 mod XS version like Dufus is using. Give that a try so your results are comparable.

Download Super PI Mod v1.5 | techPowerUp

While you are at TechPowerUp, try downloading RealTemp.

Downloads | techPowerUp

RealTemp uses high performance timers within the CPU so it can accurately determine what the multiplier is really doing. RealTemp also accurately reports the CPU load to 1 digit after the decimal place. When benchmarking, you need to make sure your CPU load is consistent before starting. Windows 8 has a lot of background baggage at times so controlling this when testing is critical.

http://i.imgur.com/DW82y1H.png

Getting the idle load down under 0.5% is a good place to start.

While SuperPi mod 1.5 is running, the maximum multiplier is easy to see. Like Dufus, I have enabled the +2 bins of overclocking that the 4700MQ supports so during single threaded benchmarks, the 36 multiplier is being used for the vast majority of the time. There are always going to be Windows background tasks that will reduce the maximum multiplier because they need to activate additional core or cores but still, your CPU multiplier should be much higher and your times should be much better.

http://i.imgur.com/f6HhJ1Y.png

Dufus has Dual Channel memory and has overclocked his BCLK a couple of MHz so his final times are about 0.4 seconds better than the 10.2 that this 4700MQ runs. The RealTemp screen shot shows the full 36 multiplier being used during the benchmarking. If you are not overclocking, you should see the full 34 multiplier during this test. Something is not right with your laptop so you will need to do some research to find out why.

Wise advice. I think I told Dufus the same thing when I first saw that score.

 

I agree with Unclewebb here, stepped turbo is for running single threads faster since power draw on the package and primary plane will be lower allowing more headroom. How many cores are utilized at one time is down to the OS, not the CPU and a single threaded application will never be able to use more than one core at any point in time. Any other cores woken up are a result of the OS concurrently running other applications.

Did send that SPi to HWBot a couple of months ago but wasn't registered for some unknown reason. I thought it was because the first time it was run under W8 but then I made a second run under W7 (~0.016 seconds quicker than W8) but no go. It doesn't matter, the time is not that much better from what is already there.

 

You guys are looking at turbo all wrong. You are looking at what it does, not WHY it does it.

Here is intels official list of factors;

Type of workload
Number of active cores
Estimated current consumption
Estimated power consumption
Processor temperature

Who knows what the first one is looking at but the middle 3 are a big deal in notebooks. Just because your cpu is cool does not mean it will always hit up the turbo.

 

Well, looking at it differently perhaps rather than "all wrong".

It would be more correct in my case and probably Unclewebb's as to "why it doesn't do it" and wrt that try and fix it.

For the list of factors Intel have provided the only one that is a "big deal" for my own laptop is the last one, temperature. Of course YMMV

The first turbo boost was actually called "Intel Dynamic Acceleration" (IDA) from the C2D days and that worked to provide an increase in frequency if one core was idle. It was also possible to run both cores full load, contrary to Intel specification, at the IDA frequency using a little trickery.

Next generation CPU's saw IDA turn into "turbo boost technolgy" followed by further enhancements such as TDP and TDC throttling as well as some others. Infact the number of controls over turbo boost seems to be ever growing.

 

You are looking at it wrong.

Power management or silent fan settings might lock the turbo. But if you stress the CPU off-leash and it stays within its thermal design turbo mode will run continuously.

- Samsung Series 9 Core i7-3537U which never throttles

 

The Turbo Boost feature was designed by Intel to be used ALL the time. Whether your CPU is running a single threaded application or your CPU is running flat out and fully loaded, a 4700MQ will be using as much Turbo Boost as possible. It the CPU exceeds the TDP limit, the amount of Turbo Boost is reduced but it is still using as much Turbo Boost as possible. The only factor in your list that will completely disable Turbo Boost is if you reach the thermal throttling temperature which is not until a peak core temperature of 100C.

Both of those statements are wrong. A properly designed cooling solution for a laptop is supposed to allow a 47 Watt TDP CPU to run at the full 47 Watts indefinitely, without overheating or ever reaching the thermal throttling temperature. That is the Intel guideline. A properly designed cooling solution lets the 4700MQ use Turbo Boost indefinitely. At 47W, the 4700MQ in my Lenovo Y510P settles in at about 85C so it can continue to use Turbo Boost regardless of what type of load it is running. The amount of Turbo Boost can and does vary based on the type of load but it does not disable Turbo Boost. Same thing with current, power consumption or the number of active cores. All of these factors can limit Turbo Boost but they do not disable it.

Here is some interesting info about modern CPU design.

Power Optimization – a Reality Check
http://www.cs.berkeley.edu/~krioukov/realityCheck.pdf

Intel's 4th Gen CPUs follow this "Race to Sleep" philosophy. They transition from the low power C States up into full Turbo Boost faster than any previous generation and they stay in Turbo Boost as long as possible.

Your ideas about Turbo Boost are wrong. They are opinions that are not based on any testing or facts.

 

Turbo boost works continuously because the temps are low.

 

Yeah. When I select the high performance power plan, my CPU sits at about 2.8 GHz.

 

Intel Turbo Boost is designed to keep working even when the core temperatures are nearing the thermal throttling temperature.



3 threads of Prime95 is enough load to keep all 4 cores active. Even with 1 core at 97C, just 2C shy of the 99C thermal throttling temperature, this CPU is still using full Turbo Boost. The default multiplier for this CPU is 32 when 4 cores are active. This CPU is not only using that but at 97C, it continues to use the +2 bins of overclocking that Intel built into this CPU. At 97C it is running at the maximum 34 multiplier. This CPU is running exactly as Intel designed it to work.

When the CPU reaches the thermal throttling temperature it will slow down just long enough for the CPU to drop 1C. At that precise moment, the CPU immediately goes back to full speed with full Turbo Boost. It is incredible technology when you take the time to test it out.

If Turbo Boost is being disabled before that, your laptop has a problem.

 

That 4700MQ is a beast...

 

I'll nip this in the bud right there. Turbo exceeds the 47 watt tdp design. Like wise turbo defeats all standard tdp ratings. That is why it is turbo and not simply the rated speed of the cpu.

And I am glad you have so much faith in your notebook cooling solutions. That is why we have so many threads about laptops throttling and people wondering what cooling pads to buy. -_-

 

Yes it is. I love my i7 2760QM, but that 4700 blows it away.

 

No you just don't get it. The CPU is designed to turbo indefinitely at any power consumption up to its rated TDP, assuming it doesn't reach its throttling temperature first. If it exceeds TDP running a burn test like Prime95 or Linpack, it will continue to do so for a set amount of time before scaling down clock speed just enough (but not disabling Turbo Boost entirely) to stay within TDP.

For example, running Prime95 on all 8 threads at 3.2 GHz causes my 45W i7-3630QM shoot up to almost 55W. It maintains that for a while before dropping down to ~3 GHz which keeps it right around 45W.

On the other hand, in a game such as BF3, it's locked at 3.2 GHz across all threads indefinitely because power consumption is much lower at around 35W.

Actually, I can make my 3630QM run at >50W indefinitely with some changes in the BIOS, but it's useless outside of synthetic benchmarking as nothing else comes close to making the CPU draw that much power.

 

Hmm, I think I'm starting to understand why the XTU benchmark is able to give the impression that I can pull 100+W from my 4900MQ continuously:

So because XTU is essentially "power-cycling" the Prime95 processes, each core never really quite exceeds its short power max limit, and this gives the illusion that I'm running my 4900MQ at 100+W for well over a minute. I knew this was too good to be true for locked non-Xtreme CPU.

 

Octi, how are you measuring your cpu power draw?

Also like I said, turbo exceeds the tdp design, that is WHY it eventually overheats and throttles back down. Some cooling solutions are better then others so obviously this changes from laptop to laptop, but throwing that aside I find it hilarious you use the word indefinitely when there is nothing indefinite about it. On the average cooling solution, a cpu running in turbo will reach its heat threshold. You are also ignoring the intention of turbo in laptop designs, using as you said, synthetic benchmarks which are pretty useless. Nothing runs around maxing out a single thread for long periods of time, even single threaded apps. Also turbo ideology is little different in laptops. Turbo on desktops can take advantage of better cooling tech to help accelerate 1-2 threaded games and apps. Turbo on mobile processors on the other hand are combined with power management so that one core can be boosted to complete a process in a specific amount of time while attempting to keep other cores powered down. Hence why turbo on mobile processors are not intended for extended use (had to repeat this, even I got lost in what we were arguing about).

 

Measured using ThrottleStop or the CPU Package Power reading in HWiNFO64 (they're the same number).

The rest of what you're saying is just your opinion, not Intel's design principles at all.

 

That was in an intel document describing the new internal power management in the 4th gen ULV processors. It was something I read when doing research for a new laptop recently. I really need to find it again, lots of detail that was in it that I can't seem to find any where else.

 

The ULV CPU's are purposely gimped by Intel. Our full-power quads behave differently and are a lot more capable.

 

This, so much of this.

 

Running Super-Pi, a single threaded application, on one of the 4 cores on my 4700MQ with 36x multi enabled results in an Intel estimated core power draw of 12W and a package draw of 20W. Running a single thread of Linpack results in a core power draw of 24W and a package draw of 32W. Both these are well with the standard 47W limit let alone the higher limit I have set.

TDP is a specification for manufacturers to design a cooling solution not the maximum power a CPU can run at. If a manufacturer skimps on the cooling to keep costs down your likely going to be limited, however if the manufacturer supplies cooling and hardware above the minimum recommended power handling then the CPU can be pushed to operate past it's default power specification.

 

Well take for it what you will.. that is just what I read for intels current line of thought. The ULVs aren't just there to satisfy one customer, I think intel thinks all their chips are using to much power and they want to use the ULVs integrated power management on all future chips eventually.

 

Funny you should mention TDP specification. The heatsink in my P370SM is rated "only" for 57W, but through the magic of Liquid Ultra, max fans, and delid (of bottom panel), I can run a sustained 90W load and still keep it under throttling temp. Even 120W in short bursts can be tolerated, which is pretty insane since it's now approaching desktop TDP territory.

 

Well the OP was for 4700MQ not ULV, maybe this post from an Intel Engineer helps.

https://communities.intel.com/thread/19612

IOW if it stays under these and is not otherwise hampered by BIOS / Software, it will run indefinitely.

 

If Intel thinks its chips are using too much power, what the HELL is AMD thinking?!

 

^"Who cares about power, FULL SPEED AHEAD"

 

AMD is doing alot of thinking and not much else. They are doing their best to Capitalize on their console market control, but their fabrication party is stuck in neutral preventing them from doing any really new technologies.

 

No their architecture just sucks. Intel did way more at 32 nm than AMD ever could. Everybody loved Sandy Bridge and to this day there's still no reason to upgrade from it. AMD needs to scrap Bulldozer and start over which it sounds like they are already doing. Phenom II had higher IPC than FX, FX only wins because of higher clock speeds. Reminds me of Pentium 4 Northwood to Prescott and what a disaster that was.