Sign in with
Sign up | Sign in
Your question

WTF: 720QM Defeats 620M in Single-Threaded App?

Last response: in CPUs
Share
September 11, 2010 12:12:29 AM

I'm looking to buy an i7 720QM or i7 620M.

I was looking at the benchmarks on the site:
http://www.notebookcheck.net/Mobile-Processors-Benchmar...

And I was shocked.
If you take a look at the Cinabench Single-Threaded benchmark, the 2.66Ghz 620M gets about the same score as the 1.73Ghz 820QM.

I mean even in Multi-Threaded apps, 5.32Ghz should not be so far behind 6.92Ghz. But in Single-Thread?

WTF!

The reason I'm thinking of buying the 620M is because overclocking it to 3.06Ghz x 2 = 6.12Ghz, where as the i7 720QM for around the same price would gain me just around 7.02Ghz, and battery life would be abysmal.

I need the performance because I'm a photographer and I render together 20-80 Gigapixel images.
But at the same time, I'm going to be using the laptop for gaming while on the road, where 3.06Ghz will definitely defeat 1.8Ghz, since most games including Crysis and any emulator will only utilize the first two cores. So basically, you literally can't run a PS2 or DS emulator on the i7 720QM. So much for performance.

So is anyone willing to clear this up for me?
a b à CPUs
September 11, 2010 1:10:27 AM

Both have Turbo (rising the frequency when running heavy applications). So it's basically 3333 vs 2800. The 720QM consumes less power when running non demanding applications.

But most of all the 720QM is a quad with hyper-threading, the other one is a dual core with hyper-threading.

http://ark.intel.com/Product.aspx?id=43122

http://ark.intel.com/Product.aspx?id=43560

I wouldn't OC a laptop CPU, those two are fast enough.
m
0
l
September 11, 2010 3:44:13 AM

mosox said:
Both have Turbo (rising the frequency when running heavy applications). So it's basically 3333 vs 2800. The 720QM consumes less power when running non demanding applications.

But most of all the 720QM is a quad with hyper-threading, the other one is a dual core with hyper-threading.

http://ark.intel.com/Product.aspx?id=43122

http://ark.intel.com/Product.aspx?id=43560

I wouldn't OC a laptop CPU, those two are fast enough.


Oh, so you're suggesting that the Quads were Turbo-boosted to their max single-core preference for those tests?
Can you prove it?
Can you find me tests between the i7-720QM @ 1.6Ghz vs i7-620M @ 2.66Ghz?

P.S. Why are you a Veteran when you've been here since earlier this year and I have been here for 40 years?
m
0
l
a b à CPUs
September 11, 2010 3:49:49 AM

Rank is based off of points, which you got for posts (4 pts) and best answers. However old posts (from a couple years ago) were only given some fractional point value.
m
0
l
a c 201 à CPUs
September 11, 2010 3:57:48 AM

You fell into the classic trap of multi core processing. 2 cores at 3.06GHz NEVER equals one core at 6.12GHz. For a single threaded task a dual core at 3.06GHz and a single core at 3.06GHz with the same architecture will perform identically because there is nothing for that second core to do.

The 720QM is a true quad while the 620M is a dual with hyper threading, but turbo boost plays a big factor in the mobile i7s, a 620M has a turbo of 3/5 so in single threaded applications its clock speed increases by 5 bins or 666MHz to 3.72GHz, a 720QM has a turbo of 1/1/6/9 so in a single threaded application its speed increases by 1.2GHz to 2.8GHz, the 820QM has even better turbo at 2/2/8/10 giving it a boost of 1.33GHz during single threaded applications bringing its speed up to 3.06GHz.

If you are really only running single threaded tasks then you wont notice the difference between the i7 820QM and the i7 620M as their clock speeds will be close enough, but most image editing programs are multi threaded so the 820QM or the 720QM would help you get your work done faster.
m
0
l
September 11, 2010 3:59:16 AM

EXT64 said:
Rank is based off of points, which you got for posts (4 pts) and best answers. However old posts (from a couple years ago) were only given some fractional point value.


Yeah, but THG isn't worth it anymore. No one is left and this place has changed. God I miss the 7900GTX VS X1900XT days..
Still would have been nice to have received points for best answers back when.

I guess that means THG doesn't care about its actual veterans, huh?
Assholes.
m
0
l
a b à CPUs
September 11, 2010 4:13:24 AM

Yep, I think they wanted to get everyone on an equal playing field in the hope of making "competition" - as some of the long term posters already had 10K-20k posts.
m
0
l
September 11, 2010 4:18:24 AM

Haha. I fell into a trap? That's adorable.
Now do you see the problem of people signed up in 2009 being dubbed as veterans?

Son, I was writing columns on the veritable performance of Single, Dual and Quad-threaded apps since Quad-Core CPU's were invented.

3.72Ghz? Woah, what have you been smoking?
That's 3.06Ghz, as in 2.66Ghz + 666Mhz, not 3.06Ghz + 666Mhz.

When I was talking about 6.12Ghz I was OBVIOUSLY talking about multi-threaded apps/multi-tasking.

Now this is where it gets interesting -
I compare 3.06Ghz on the 620M vs. 2.4Ghz on the 720QM because:
There is no point in using only one core. Two is the minimum for smooth multi-tasking.
So basically, we're talking about single-threaded apps in a two-core minimum package

The only point of this thread is the single-thread benchmarks.
Do you agree that the Quads were overclocked by using only a single core and are thus cheating? I do.

hunter315 said:
You fell into the classic trap of multi core processing. 2 cores at 3.06GHz NEVER equals one core at 6.12GHz. For a single threaded task a dual core at 3.06GHz and a single core at 3.06GHz with the same architecture will perform identically because there is nothing for that second core to do.

The 720QM is a true quad while the 620M is a dual with hyper threading, but turbo boost plays a big factor in the mobile i7s, a 620M has a turbo of 3/5 so in single threaded applications its clock speed increases by 5 bins or 666MHz to 3.72GHz, a 720QM has a turbo of 1/1/6/9 so in a single threaded application its speed increases by 1.2GHz to 2.8GHz, the 820QM has even better turbo at 2/2/8/10 giving it a boost of 1.33GHz during single threaded applications bringing its speed up to 3.06GHz.

If you are really only running single threaded tasks then you wont notice the difference between the i7 820QM and the i7 620M as their clock speeds will be close enough, but most image editing programs are multi threaded so the 820QM or the 720QM would help you get your work done faster.

m
0
l
September 11, 2010 4:19:43 AM

EXT64 said:
Yep, I think they wanted to get everyone on an equal playing field in the hope of making "competition" - as some of the long term posters already had 10K-20k posts.


Yeah, but if someone has 20,000 posts, they deserve to be top dog because there aren't many of them, so no one's gonna care.
m
0
l
a b à CPUs
September 11, 2010 4:22:14 AM

The system is strange. "Veteran" means old poster. Maybe they should use veteran and stuff for the longevity and put some stars or something for the points. For instance 5 star apprentice and 5 star veteran.
m
0
l
September 11, 2010 4:30:18 AM

I don't care as long as I get some sort of recognition.
I've been here since 2006 and know more about hardware than anyone's I've ever talked to.
But please let's not debate this.

Whatever THG SHOULD be doing they're not going to change.

mosox said:
The system is strange. "Veteran" means old poster. Maybe they should use veteran and stuff for the longevity and put some stars or something for the points. For instance 5 star apprentice and 5 star veteran.

m
0
l
a b à CPUs
September 11, 2010 4:31:41 AM

Yeah, I didn't like the change. Personally I think the new ranks are too simplistic, boring, and too few. I mean, almost anyone with over one post is a veteran or addict.

Oh well, guess we'd better get back on topic, sorry for causing a diversion.
m
0
l
September 11, 2010 4:42:00 AM

I just thought of something scary.

The whole TURBO thing..

Does it just change the cores/frequency by itself.. as in when it thinks that 1.6/4 is right, all the cores will be working and when it thinks that it only needs one core, it will just automatically shut off the rest of the cores..
Or can you choose in the BIOS?

I think the former is absolutely retarded. There is no time in which not all cores are useful, even if you only have one app running.. there's a ton of things in the background that could utilize the extra 3 cores. And the whole 2.8 vs 3.06 debate is meaningless if so.

Please tell me that it's a choice or that there's a hack to choose.
I mean in the single-threaded benchmark.. the CPU couldn't POSSIBLY have just known "oh, the guy's trying to benchmark a single thread, I'll shut off the other three".. no, it would been like "oh, you're benching? ah, but I'm also doing lots of other stuff with the rest of the cores so.. go into the BIOS and choose one core if you want that."
m
0
l
a b à CPUs
September 11, 2010 6:33:48 AM

Man, I haven't seen that avatar in a long time :) 
m
0
l
September 11, 2010 10:27:58 AM

WB Track
The app determines its needs mostly, so true real single thread causes turbo to activate at its highest state
m
0
l
a b à CPUs
September 11, 2010 1:30:22 PM

The main reason I think that turbo is effective is that it isn't a software implementation, the processor has a co-processor that decides extremely quickly based off of your usage. So, if you do have significant loads on the other cores, it won't turbo (or not as high). Whatever combination of cores, as long as it stays under the TDP, it is happy. If it can't due to loads on other cores, then it just doesn't turbo.
m
0
l
September 11, 2010 2:05:08 PM

Which may mean, down the road, a cmt implementation vs a smt one, may prove somewhat superior
m
0
l
a b à CPUs
September 11, 2010 2:28:50 PM

Yep, it will be very interesting to see just how successful that implementation is.
m
0
l
!