2720QM vs 2820QM vs 2920XM

I think it would be nice to have a thread where we can post the differences between these Sandy Bridge CPUs and to make the choice between these CPUs easier.

Specifications:


Price:
* Intel Core i7-2920XM - 2.50GHz - $1096
* Intel Core i7-2820QM - 2.30GHz - $568
* Intel Core i7-2720QM - 2.20GHz - $378
(* Intel Core i7-2620M - 2.70GHz - $346)
(* Intel Core i5-2540M - 2.60GHz - $266)
(* Intel Core i5-2520M - 2.50GHz - $225)

Performance:
Note: 2820QM is not included in these tests.





Power consumption
Note: 2720QM or 2820 is not included. But 2630QM and 2720QM will probably be similar.



Source:
Review Intel Sandy Bridge Quad-Core processors - Notebookcheck.net Reviews
Intel Core i7-2820QM - Sandy Bridge For Notebook PCs - Intel Sandy Bridge Comes to Laptops! - Legit Reviews

To be updated more...

 

i7-2630QM ^^

-- it's gonna be somewhere in the neighborhood of $350, and no markup unlike the 2720 and the XM.

 

You probably won't see the 2630QM much, at least initially, since it will be strictly to ODMs/OEMs. I think when it's on eBay, the other CPUs will also be found for fairly cheap too.

 

could you also post max ghz and how long sustained and power consumption when idle and low/mid level usage?

 

and also max sustained ghz w/ and w/o igp usage.

 

I see myself with a 2920XM or 2940XM if I get a Sandy Bridge laptop, it'd be interesting to know how the IGP affects their overclocking.

 

Although great for midrange systems, the igp of an sb laptop could prevent the cpu from sustaining max clocks in a performance laptop.
In that case, i think it would be better to have switchable graphics, with the option to shift to discrete gpu when doing cpu intensive tasks

 

Incorrect. When doing CPU intensive tasks, the GPU will be power gated and use 0W.

 

Still waiting for more reviews out there that can show us real results because Intel has not given us any spesific numbers on how this dynamic turbo works.
I can only quote what Anandtech said in their review of the 2820QM.

Finally, we’ll wrap up this discussion by looking at maximum CPU loads. In the Cinebench test, quad-core Turbo is interesting to watch; running the CB11.5 SMP benchmark, at first all of the cores start at the maximum 3.10GHz speed—blisteringly fast for a notebook! About 11 seconds in to the test, the core speed drops to 3.0GHz, where it remained until 39 seconds; then it dropped to 2.9GHz, and at around 54 seconds the speed dropped briefly (1-2 seconds) to 2.8GHz before settling in at 2.7GHz for the remainder of the test. At the highest point in the test, power draw for the notebook peaked at 89W; once the speed settled at 2.7GHz (which it appears the notebook could sustain indefinitely in our 70F testing environment)

As for Single core and Dual core you will probably see higher speeds than 2.7 GHz within TDP because the CPU have a lot more termal room with less cores active. Meaning it could stay forever at higher speeds than 2.7 GHz with not so good cooling system.

You can see it here:
YouTube - Sandy Bridge Turbo Mode - IDF San Francisco 2010

And i don`t really think that IGP under stress would affect the 2.7 GHz limit (Quad) for the 2820QM, but will result in shorter rounds of speeds above 2.7 - 3.4 GHz. How much the quality of the cooling system affect the speeds above TDP when the CPU/IGP is stressed, i don`t know.

Will wait for more reviews to update this thread about power consumption and GHz numbers. The above give you atleast some indication

 

But it can't.

Intel's Sandy Bridge Architecture Exposed - AnandTech :: Your Source for Hardware Analysis and News

Real World Technologies - Intel's Sandy Bridge Microarchitecture

 

Yes but it will draw less wattage with less cores active and therefor still staying with TDP with higher speeds than 2.7 GHz.
Look at the video i linked to. You can see the green coloumns (which is within TDP) of Dual core and Single core going higher than the 4 (Quad) green AND white coloumns which is when the CPU goes above TDP.

I could be wrong though

 

I know this is mostly a desktop-only question but, anyone know if it is possible to overclock only the turboboost frequency for the cpu? ie, idle at the same nice, low power envelop, but then burst into 4Ghz territory ? for that matter, is the base frequency also the idle frequency?

 

Only the 2920XM (and any XM coming later) will be overclockable.

No. It can go as low as 800MHz idle, I think.

 

2820QM downclocks to a minimum 800 MHz when idling, so no
EDIT: too slow

 

Sorry, to add a bulletpoint to what I said.

Applies mostly for quad core Turbo. Dual will obviously reach higher and single even higher.

 

I am a ninja.

If I wanted one of these... I'd say that the 2720QM offers the best value.

 

agreed about the 2720QM but i'm probably waiting for ivy bridge refresh in 2013-2014.. my laptop is good enough for now..

 

really? if that is so then that is good news indeed. you got a link?
so laptops with discrete gpus will have it automatically switch on during cpu intensive tasks?
and laptops with no discrete cards? will be held back from max clocks by the igp then?

 

2720QM is the best deal and has great performance. if your an enthusiast and are overclocking the XM of course.

 

agree. then just wait for the xm qs/es available over ebay

 

Real World Technologies - Intel's Sandy Bridge Microarchitecture

The bolded part doesn't make sense. Why would the integrated chip activate when you are using a discrete GPU?

There's no difference between the Sandy Bridge GPU doing nothing and being disabled. It might not be 0W, but it'll be very low and has its own power plane so it can power down independently of the CPU.

In theory yes, but I doubt you'll see more than few % impact at the most. Most of the time, the CPU will be held back by the GPU for games and the little time it requires the extra performance, it'll be cool enough.

There's 3 types of clocks on Sandy Bridge
-Base clock
-Sustained clock
-Above TDP clock

1 and 2 can be sustained indefinitely(unless say your laptop is burning ) but the 3 has time limit that expires regardless of cooling.

 

thanks for the article
actually what i meant by that was the snb laptop would switch to (turn on) the discrete gpu during cpu intensive tasks. for example, i do video editing and i would like to use the chips full tdp and performance and let the discrete gpu handle the graphical duties of the o/s while doing this. it should allow the cpu to operate above base clocks for extended periods, especially when using a laptop with excellent cooling like a clevo/sager.
afaik it still takes a fair amount of wattage to keep the graphical o/s running (not 0watts), and id rather shunt that overhead to the cpu when on a/c or performance mode.
however id like the igp to kickin when on low power/surfing mode mode.
although it should automatically transition between igp ond gpu on optimus, it would be better to have manual igp/gpu switching for more control and for ati users.
although the article does point out that the igp has a seperate power plane, i believe ive read somewhere that with igp enabled (just running o/s), the max o/c above the base clock (on all threads) is limited, leading to lower benches when igp is in use. again, this would be a moot point if we had manual control of igp/gpu switching. i guess we'll just have to wait and see.

edit:
just saw one of your post stating that 25secs is the max the snb m chip can go above tdp. so i guess igp and notebook cooling effects could be negligble is it can only go max for 25secs
it would be interesting to see however if these affect sustained clocks above base.

 

Okay, if you are running a program, you aren't using the Aero features. It'll merely be refreshing the screen. Even in games, when there's no action and you are just standing there, your performance is high. So the GPU can clock down. Turbo isn't active if the demand of the program is very low. It's a stretch to call Aero "3D".

The GPU also has a seperate 2D engine, with a fixed ~300MHz clock or so. They were able to integrate 2D functions into graphics chips back in the TNT2 days, so the power it takes is negligible.

 

interesting.
lets wait for cloudfire to post benches w/ and w/o igp for a defenite answer to this.

 

LOL, 1-2% differences on a new platform that delivers 30-50%. Well, if you wait a few weeks, my friend will get a laptop, and maybe little longer than that perhaps I will get it too, so we can all benchmark it.

 

lets wait for the benches before we LOL
but i do sure hope your right though

 

Agreed.
It offers a great performance/price ratio even if you are doing cpu intensive tasks (such as using a cpu intensive programs like 3dsMax in my case).
Couple it with a decent mid range gpu (of say ATI 6xxx generation) and you can get a nice modern power-house in a laptop form.

I wonder though how much money it will cost though here in UK.
Hopefully, Acer can offer such a configuration with decent cooling like they did in the Gemstone series in 2008 for a low price.

 

2720QM runs really close to the flagship 2920XM, looks like an even sweeter deal than 2820QM.

 

The lowest end CPUs are almost always offer the best performance per dollar, since marketing gives higher end CPUs an exponential mark up.

 

imo the only 2 cpus to ever buy is the low end of whatever is the max core model out at the time(quad core for now) or the XM model for overclocking fans. I would still role with the lowest model if multipliers weren't locked ^^

 

the real benefit of the xm is its overclocking abaility. if throttlestop can be made to work with it and if it overclocks anything like the 920xm, it should hit (or atat the v least) 4ghz on all cores.
thats where you will see the big difference.

 

as i said earlier is if you could play with multipliers in the lower end quad core there wouldn't be such a huge difference....hence why intel locks them so overclockers buy the XM lol marketing at it's finest ^^

 

true. however intel has purposely designed the core i series to be locked. snb is an even more locked system than its predecessors.
there was a big debate over at anandtech when the nehalems came out over the death of overclocking. many lamented that gone were the days when you could get a midrange cpu and overclock the sh*t out of it to approximate higher end models.
fortunately some modders found a solution for some models thru setfsb. would be interesting is they could actually unlock the snb qm models.

 

I am sure they will find a way but it is BS how they are not making it an easy go into bios and click click click 10% over clock like my P4 was :/

 

i think locking down non extreme cpu's is part of intel's internal roadmap. it really is too bad since o/c'ing is what drove the enthusiast market for many years.
however it leaves the doors open for amd. if they can give us good performance with off the shelf o/c ability then intel could have a run for their money.

 

That would be awesome, but also a miracle IMO. With the current XM's 3GHz under load for 24/7 operation is hard to pull off.

 

?
my friend has his sager8690 on 3,6ghz with hyperthreading on when he uses his comp. he doesnt have it on 24/7 though. i dont know anyone who does actually...

 

mine is lol but i don't have a XM :/

 

I think Intel should just give people a way unlocking any CPU at least at Core i3 branding and charge $30-$50 for it. Not the best solution for either side, but certainly the best compromise (especially for pirates ).

 

Probably easier to get people to pay more for more performance/features, or pay less and get less depending on which way you want to look at it. CPUs are binned on perfection so offering a performance upgrade might not be so simple for the less perfect chips. One guy gets 50% more performance for his $50 another guy gets only 10%.

I'm not 100% but I believe that time is changeable but don't get your hopes up at something like that being provided in the BIOS. Who knows ODMs might find they need to cut maximum power back and reduce it to 1 or 2 seconds.

 

He might have his ratios set to 3.6, I guarantee it's no where near that under prime. I have modified cooling and is still get into the 90's at 2.8 running 8 threads.

 

Like my G51j where they have a crappy cooling system and they put a 90watt thottle limit and underclocked the gpu by like 10%+ and i still hit 105C in 60s on furmark....stupid cheap asus. Also we will just have to see what uncle webb can wipe up for us ^^

 

interesting. ill have to ask him to do prime. he has the es and i know some es downlocks at 88*c so he may have it at 3.6 and it could downclock to to 2.8 when it reaches that temp.

 

I'm guessing Throttlestop shows a 20-21 multiplier after 5 min of prime. Nothing wrong with that tho, it's still blazing fast for a notebook. But ya, lemme know.

 

OP could also include some info on that drm present in snb tnx

 

I thought that was still in development and not yet present in any CPUs.

 

Does anyone know the performance difference relating to the cache in the 2720qm (6mb) vs 2820qm (8mb)?

Im also curious how the 840qm stacks up against the 2720qm in performance.

 

What are you planning to do with the notebook? For day-to-day tasks, there will be a 0%, while on the other end, for certain programs that can utilize the cache and ignoring clock speed differences (which is pretty minimal at 100MHz -> <5% difference anyways), the cache can make at most 10-15% performance benefit (this is mostly from extrapolating from previous generations and nothing concrete since I haven't tested the two). Though realistically, I'd say the overall benefit would not be significantly noticeable.

 

I usually go for the biggest processor i can to sort of future proof (a little!) - main power usage would be gaming.

I dont know what impact 6mb cache in the 2720qm has vs the 840qm with 8mb. This of course isnt going to be as straightforward as comparing the first i7s with different levels of cache - or is it?

 

It also depends on the game, but majority of games are GPU bound and hence you won't notice any performance difference between the 840QM and 2720QM. While there are other benefits of the 2720QM (such as battery life), upgrading to a faster CPU (2820QM or 2920XM) is not worth it for gaming and will yield pretty much no extra "future proofing" since it will not be able to do anything more than the 2720QM can do.