i7 2820QM runs at 3.1GHZ under max load? bull?

according to hothardware the i7 2820qm runs at 3.1ghz while doing a cinebench. This would mean while under maximum load this CPU runs nearly 1ghz higher than its advertised frequency?

Is this too good to be true? When does it actually run at 2.3ghz.

My guess is that the processor only runs at this frequency for a short time, perhaps just enough time to finish most bench marks.

this mean all the performance figures we are seeing are not applicable in the real world because a rendering takes a lot longer than it takes for cinebench to finish and cinebench normally doesn'y heat up a cpu to stable.

I would like to see someone do a test that would determine how long SB CPU's can sustain its turbo frequencies. Its strange to me that professional outfits like anandtech and hothardware did not mention this obvious flow in their testing

http://hothardware.com/Reviews/Intel-Core-i72820QM-Mobile-Sandy-Bridge-Processor-Review/?page=3


EDIT:Actually. can someone do a full CPU load test (prime 95, rendering or handbrake) and see how long it takes for the CPU to scale back to 2.3ghz.

 

Intel implemented turbo boost 2.0 which allows the SB CPU's to go beyond their TDP for a short period of time and it could be why they saw higher frequencies. You do have a point that with extended benchmarking that heats up the CPU, the TB would scale back quite a bit.

 

What do you mean by "the real world"? What do you think is the percentage of users who would actually ever run their machines at such a high load for any significant amount of time? How many users do any sort of rendering at all?

 

maybe i should start a poll.........

seriously though. i think the question is relevant to alot of people because it;s also applicable to extended gaming and video trans-coding scenarios.

If we are going off cinebench and superpi scores we are going to be dissapointed when we code some videos for iphone or start playing a cpu intensive game.

big difference between 2.3ghz and 3.1 ghz.

i want to see someone do a prime 95 or other full CPU burn test(rendering or handbrake) and see how long until the CPU scales back to 2.3ghz

 

Macpod is right, there's a very good chance these SB synthetic scores are overinflated due to turbo boost 2.0 providing short term clock boosts beyond TDP during the time these benchmarks run. A longer duration benchmark and/or CPU intensive game would probably cause the CPU to scale back to default TDP and thus lose massive performance.

 

The Sandy Bridge Review: Intel Core i7-2600K, i5-2500K and Core i3-2100 Tested - AnandTech :: Your Source for Hardware Analysis and News

Look at these scores. the desktop SB CPU's are almost on par with the mobile variants.

This is all a big scandal. I hope anand and other respectable sites start to mention this.

you would think professional reviewer who actually make money from this stuff would be more thorough.

 

i7-2600k



i7 2820qm



common. really?

 

You're comparing an i7-2820QM to an i5-2500k, NOT an i7-28xxk.

 

i7 2600k gets 22875 3.4ghz

i7 2820QM gets 20393 2.3ghz base

It's all there.

 

actually.......whats your point? is there even an i7 2800k?

 

Yes, there's a 10% difference in the scores. Is the i7 2820QM the mobile version of the i7 2600k? Shouldn't there be an i7 2800k? Shouldn't you be comparing an i7 2630QM to the i7 2600k?

 

There is no trickery here -- it is all fully explained if you read the entire article. The Turbo Boost chart can be seen here and indeed, the 2820QM can Turbo all the way up to 3.1GHz with all 4 cores active.

The explanation of Turbo Boost is much older than that. The way it works is that as long as the machine is not overheating, it will continue to run at the Turbo frequency. If your cooling is decent, this means that you are in fact buying a machine with a base speed that is greater than advertised. For example, my desktop with a Core i5-750 is nominally a 2.66GHz machine, but in practice it always runs at 2.8GHz under load. Turbo 2.0 is exactly the same concept, but the increase in frequency is much higher.

Now, it does mean that if you try to stuff such a CPU into a small form factor machine (e.g. a 13" laptop) it may not run at 3.1GHz for long and the benchmarks will not be accurate, but as long as the cooling is there, it will run at the Turbo frequency. To be honest, I believe these chips have a lot of thermal headroom -- notice that in the desktop review, they were effortlessly overclocked beyond what Turbo Boost does with stock cooling. The Turbo Boost won't be permanent only if you try to do something extreme.

That is pretty weird. It looks like the desktop machine is working without Turbo Boost and the laptop one is with Turbo Boost (the difference is about 10% which is what you'd expect between 3.4GHz and 3.1GHz). And it's not just AnandTech -- PC Perspective gets a similar result (although the laptop score is somewhat lower, see here and here). This is something that the testers have to be asked directly.

 

Clearly the 282QM is running near 3Ghz. I don't see why that's so hard to believe.



This isn't a synthetic benchmark. The CPU encoded faster than the i7-920. It's clear.

 

He is not saying it is not, he is saying it cannot last long at that frequency due to heat. That as explained by other poster depends on cooling.

 

Anandtech was clear about what went down:

The truth was right there in the article.

 

It will be interesting to check how much variability can happen in a typical laptop. Btw many users work w/ video or 3D content creation, plus gamers so a 100% CPU push for long time is possible.

 

this raise a valid point that cooling system design will be a factor to check when buying a new notebook. Would that make one with metal chasis(those business model) potentially faster than plastic consumer model even though the spec is exactly the same ?

 

Good catch. Still, it would be interesting to see how long it will take with prime 95 or something similar to take it down to 2.3 ghz.

I guess with the SB notebooks there will need to be a section dedicated to how long turbo frequencies last with the cooling present.

Just to be clear, i understand turbo frequency and thermal limitations but in prolonged usage the turbo frequencies would be useless. Depending on thermals of course, and i would like to know what the thresh holds there...

I still stand by my initial reaction that the reviews are not clear enough as to this turbo/thermal issue. And most people reading the reviews will think the mobile parts are perfoming on par with desktop parts when in prolonged use they are not.

 

The better you can dissipate the heat, the better a CPU will run so yes, some laptops will be faster than others. The Compal they tested tops out at 2.7GHz for long-term tasks, but it's far from the best possible cooling configurations:

 

SB is no different than Clarksfield in regard to TB being directly dependent on TDP and the cooling system of the notebook. That's where companies like Clevo and Alienware which traditionally have superior cooling vs other manufacturers will excel and pull ahead. The SB CPU's at stock speed thrash the Clarksfield CPU's but not necessarily the 920xm due to unlocked TDP via Throttlestop. IF the author of TS unlocks SB's 2920xm, then SB will be a clear winner. Like I mentioned in other posts, with a 940xm and throttlestop, I can push the TDP to 105W comfortably in my M17x-R2 for 24/7 use which pushes all 4 cores to 3.7-3.8+ GHz. SB wouldn't be able to compete with that at all unless TS is patched for it.

I think SB's biggest advantage will be the increased battery life it provides and for people that have average notebooks (e.g. HP, Dell Inspiron/XPS).

 

Alienware and Asus G series just got more of my attention.

I had my eyes on the XPS17 but it seems a system with seperate heatsinks for the GPU and CPU will show significant imporvements. enough to justify the weight and bulk penalties.

what will happen to the macbook pros........oh my...

 

They will continue to be overpriced under performing crap that people will flock to buy.

 

what you don't get is, most of the time you never run the cpu at 100% even while it states that it does. why? because it can even throttle back while waiting for memory, waiting for the hdd, or what ever. it's mostly bursting as much as it can and then idling. when ever it HAS something to do, it does it as quickly as possible to idle as soon as possible again. no cpu workload is 100% cpu, so it most likely idles enough to cool down again.

AND most cpu workloads (nearly all non-bench workloads) are not using 100% of the cpu. they go to branch, then to integer math, then some floatingpoint math, then back, etc.. that means different parts of the chip heat up and cool down at different times.

the advertised speed is a min-guaranteed speed. if it's very hot, and the cpu is at 100% usage AT EVERY PART OF IT, you still get that speed guaranteed. in any reallife scenario, as visible, the cpu goes much higher.

so yes, it sorta overclocks all the time, and it's varying it depending on the temperature => does it less when hot.

but calling the stuff a scam is a ridiculous missunderstanding of the technology.


oh, and, no, if i do extended video encoding sessions, i won't, in the future, use the cpu quicksync or how it's called ftw.

 

+ they will slow down much more often as they are overheating even while idling.

 

Not like they use anything that throttling takes away.

 

Actually. If there is no dedicated GPU like the current MBP 13 and base config 15, the throttling will be even more prevalent because the thermal loading is shared between the on board GPU and CPU. Ouch.

 

think of it in another way. right now, the mbp (with or without gpus) have a termal load for their cpu, and a fixed termal load for what ever gpu they have.

in the future, they will have a shared one, so if one is not at it's limits, the other one can go beyond it's limits. now, the cpu can look at the gpu doing nothing and say "hey, i could run faster.. but i can't".

now that is Ouch: wasted potential for higher performance. sandy bridge fixes this.

it's not throttling down. it allows to finally "throttle up" when possible. something systems just don't do yet.

 

My i7-2630qm goes from 2.0ghz to 2.6ghz ....

YouTube - Sandy Bridge i7-2630qm Turbo Boost in Action

But i think it's 2 cores that boosting... if I just base it on temps... oh well.

 

Bravoexo: Could you download prime 95 or handbrake(both free and utilises all 4 cores/8threads) and test how long it take for the thermals to lower the frequency down to 2.0ghz? If at all?

 

I don't think that they will ever scale back to base frequency when under stress, there is always some turbo boost.
With my 820qm when running full stress 4 or more threads the frequency stays locked at 2.00GHz, (2 bins of turbo boost).
Im sure the new SB CPUs with turboboost 2.0 will be able to match and far exceed this full time boost.

 

that would be awsome. Ideally something like an alienware M17x would be locked at max turbo or just below most of the time.

Which means an i7 2720QM would be quite a bit faster than the i7 2630QM because the max quadcore frequency of the i7 2630QM is only 2.6ghz vs 3.1ghz for the 2720QM

 

O.k., anybody know any "CPU-intensive" game? One that would actually tax one of these CPUs?

 

Well, yeah, but that requires people to read it, rather than just looking at the pretty pictures...

 

BC2, GTA IV, any new RTS (e.g. sc2).

 

Actually, two tests are needed: with IGP enabled, and one without the IGP enabled. Theoretically, without the IGP taking up a portion on the TDP as it does in this review, the CPU should be able to sustain higher frequencies longer.

Of course, this does mean that Optimus could show yet another weakness in some systems...

 

add Metro 2033 although its GPU intensive too..

Panther214

 

This is crazy! We shall see how all this ends up.

 

I could be wrong here, but the CPU will never clock itself down to the base core frequenzy which is 2.3 GHz if you are doing demanding tasks.

If you look at this picture

You can clearly see that just a tiny bit of the top is speed greater than the TDP, and is like mentioned heat dependent. If you have a very good cooling system it will stay on 3.1 GHz (Quad) and fall down after a while. If you have a crappy thermal system it will fall down from 3.1 very fast.
Like mentioned from Anantech it went from 3.1 to 3.0 etc etc down to 2.7 and SETTLING there.
That is probably the limit of which the TDP ends. Meaning it will upclock from 2.3 GHz to 2.7 WITHIN TDP, and from 2.7 to 3.1 GHz ON TOP of TDP. So it will never fall beyond 2.7 GHz, it will stay there like in previous Clarksfield CPUs.

Now like mentioned earlier the CPU will almost never stay at 100% for a very long period of time, and therefor it will have the ability to fully access and utilize the speed beyond TDP for normal tasks. And not all tasks require quad core. If the CPU run on Dual core or Single core it will use less watt, and therefor have much more "room" for it to have a longer run with speeds beyond TDP. And from the tests at hothardware and Anantech you could also see that the CPU does draw very little power compared to i5-750 and i7-920. That again speaks in the SB favour.

At full load, even scaled to over 3.1GHz with all eight threads pegged at 100%, the Core i7-2820QM system didn't even break the 90 Watt level. When we fired up a more graphics intensive app like Cinebench's OpenGL test, we found the Sandy Bridge consumed about 56 Watts under heavy graphics loads that call less upon the chip's CPU engines

 

one of my issues with sb ive mentioned before is that the igp could potentially hold back the cpu because of tdp.
my theory is that these chips are really rated higher than the specs. but the specs are what they will constantly perform at with igp enabled.

 

Nice post Cloudfire. Try running Linpack with AVX and move some data through those 256bit registers to get that power level up.

 

Ok, tried out Prim95 64-bit

First minute the multiplier was 26, then it throttled back to 24 for another minute, then it played around 22 and 23 for the next 6 minutes... and kept at it... til I stopped it 10 minutes later. It never went back to 2.0ghz and the temps for core 0-1 was 87-88C max, while core 2-3 was 84-83C max

 

HP's horrid cooling strikes again! :/ Apparently this thread at least has a bit of a point...

 

I would love to see how GTA IV performs on one of these Quad CPU's.

 

thanks for that explanation. it all makes more sense now. I had thought at full load the i7 720qm always ran at base frequencies or just a bin or two higher. SB seems to improve on this.

would be interesting to see if better cooling allows for frequencies to stay at max turbo or if it will be limited TDP and not cooling.

 

Tried it again... w/ video this time, 5 minute run.

YouTube - Sandy Bridge i7-2630qm + Turbo Boost + throttling w/ Prime 95

Temps up to 90C this time (10C from maxTj whoossh, ha ha), and it did briefly went down to 2.0ghz, but it keeps throttling back up to 2.2 and 2.4ghz.

 

You have to run it for at least an hour to get reliable results, but I think that we can already see how that would go.

 

Yeah, but I don't want torch this laptop .... especially that MicroCenter removed the 3 other stocks they had if I were to return this one due to melting it down. ha ha...

 

Nice lappy Bravoexo, thanks for the posts + videos!

 

@PlasmaBomb, I am happy to entertain the requests... and will be even happier once everyone else gets a sandy bridge laptop.

 

this behavior makes sense -- the TDP includes a budget for the IGP.

ie, so it should tend to stay above the base clock rate as long as it is not working the IGP at the same time.