How did the Optimus 2x perform so well in benchmarking scores?

Cory Streater

Well-known member
Sep 21, 2009
9,493
3,428
0
As reported in the blog the Optimus 2X benchmarking scores were off the charts. That's awesome. What I don't understand, is that I keep hearing that the dual cores in all these upcoming phones cannot take advantage of the second core, until there's an OS release that supports it. Can anyone confirm whether this is true or not, and if it is how the scores can be so high?
 
My guess is that even though Android can't utilize two cores yet, both are still Cortex A9 processors which allow out of order processing unlike the Cortex A8 which are currently on the market which use in order processing. That may account for some of the performance boosts.
 
Egginator is correct, the A9s are at a minimum 20% faster per clock than the A8s. But thats only one part of it. The score in Quadrant, for example, looks at much more than that. AFAIK everything that was released last year is still using DDR(1) Ram where the new Dual Core chips will be using DDR2, that hugely impacts that portion of the score. Also the graphics chip has been shown to be about 20-30% faster than the SGX540 which is already a super fast chip so both the 3D and 2D scores will improve. The last thing I can think of is IO scores and for that it could be that the faster flash chips are cheaper now so more manufacturers have opted to use them. This isn't really because of the new SoCs but it also adds to the improved scores.

http://www.anandtech.com/show/4054/first-look-viewsonic-gtablet-and-tegra-2-performance-preview/2
 
Linpack scores were probably due to the better architecure as above. Another thing that probably helped is, despite the fact that the OS cannot take advantage on multithreading, the kernel can, to some extent. You don't have a core sitting completely idle I can almost guarantee. That means if you're running linpack in a single core chip, it might only be using 80% of the core, and the other 20% are going to other system tasks. On a dual core chip, it might be using 90-100% of the core, as the other core shares or completely inherits the system tasks.

As for the robot benchmark (name escapes me), that's probably limited by the GPU, as it was designed to be. If we take the score that the LG got, it was 29% more frames than my Galaxy S. The Tegra 2 GPU was rumored to be ~30% faster in benchmarks like that. Where the T2 really shines is the extra features of the GPU.

Quadrant was probably a mix of the above.

AFAIK everything that was released last year is still using DDR(1) Ram where the new Dual Core chips will be using DDR2, that hugely impacts that portion of the score.

You sure? I where did you hear that? I think that the Atrix and the Bionic are the only ones to use DDR2. That is one of the reasons (I thought) that motorola was making such a big deal out of it. Nvidia probably has controllers for both types of ram in their processor, which is not an uncommon thing.
 
You sure? I where did you hear that? I think that the Atrix and the Bionic are the only ones to use DDR2. That is one of the reasons (I thought) that motorola was making such a big deal out of it. Nvidia probably has controllers for both types of ram in their processor, which is not an uncommon thing.

I'm pretty sure. I know that anything using a Tegra 2 will have to use DDR2 Ram since it only has one mem controller for that. You could be thinking of DDR2 vs lpDDR2. Motorolla made a big deal to speecifically mention Desktop DDR2 (as opposed to the lpDDR2). Both are DDR2 architecturally but the lp one is a low power, lower clocked version. I would assume that all new dual core chips would have at a minimum a DDR2 controller and they wouldn't want to waste chip real estate on a second mem controller to support the older very unlikely used DDR(1). But besides Tegra 2 that is just conjecture.

NVIDIA Tegra 2
 
Yeah they are using DDR2 but only reason they are able to pull it was due to higher bandwidth with multi thread processing, as CL is still high which is equal to DDR.

Now until we folks at XDA benchmark this I don't trust any of it, then again if you understand basic you will know benchmark does not emulate real world load. Example take NS atm has the highest score they are surpassing 4000+ yet they can't play decent video without lagging lol.

BTW Auzo where are you getting these info from? From the vendor? They can claim ULP is faster than GSX540/AMD Z460 but fact is its slightly faster and again they pulled this off in real world speed translation due to dual core. You will "not" notice it but GLBenchmark 2.0 seem to be decent in this case.
 
Yeah they are using DDR2 but only reason they are able to pull it was due to higher bandwidth with multi thread processing, as CL is still high which is equal to DDR.

Now until we folks at XDA benchmark this I don't trust any of it, then again if you understand basic you will know benchmark does not emulate real world load. Example take NS atm has the highest score they are surpassing 4000+ yet they can't play decent video without lagging lol.

BTW Auzo where are you getting these info from? From the vendor? They can claim ULP is faster than GSX540/AMD Z460 but fact is its slightly faster and again they pulled this off in real world speed translation due to dual core. You will "not" notice it but GLBenchmark 2.0 seem to be decent in this case.

I was merely trying to give some plausible reasons as to why the scores were higher even though the benchmarks were not optimized for dual cores. I know that synthetic bencharmarks don't always translate into real world performance and they can be cheated, but they do give you some idea.

I do agree the latency of the DDR2 Ram will be higher, but if you were bandwidth limited before the extra bandwidth can more than make up for it.

And in my first post here I linked an article from Anandtech which shows some 3D benchmarks of the Tegra2 vs Hummingbird based SoC with the SGX540. Scroll down to the "12/20/2010 - Updated Graphics Benchmarks" section of the 2nd page to see the results. But it shows a 30% advantage in GLBenchmark 2.0 pro for the Tegra2. The other two tests show an advantage for the Tegra2 but they are less than 30% and vary. I know that the Hummingbird is a single core processor, but these 3D benches aren't heavily utilizing the CPU so i figure the results shouldn't be skewed.

But you're right, I do make a lot of assumption. I wasn't trying to give a comprehensive reason as to why the scores were higher, more just some possible reasons.
 
Simple answer...anything will take advantage of dual-core on some level. Look at when dual-core was first released for regular computers. They were faster than single-core and got even faster when windows became optimzed.

The same thing will happen on phones. They will be faster then when Android gets optimized (Honeycomb on phones) the spread will be even greater than it is now.

Also keep in mind that the dual-core snapdragon is expected to meet or exceed the performance of Tegra 2. But then nvidia will announce Tegra 3. :p
 
I do agree the latency of the DDR2 Ram will be higher, but if you were bandwidth limited before the extra bandwidth can more than make up for it.

Now keep in mind I know if anyone here understand how memory is allocated you will bring up case latency issue. B what you have to understand is yes ddr2 does have higher CL but since ddr2 runs at double external data bus clock rate as ddr it can provide twice the bandwidth at the same latency thus improvement over ddr.

And in my first post here I linked an article from Anandtech which shows some 3D benchmarks of the Tegra2 vs Hummingbird based SoC with the SGX540. Scroll down to the "12/20/2010 - Updated Graphics Benchmarks" section of the 2nd page to see the results. But it shows a 30% advantage in GLBenchmark 2.0 pro for the Tegra2. The other two tests show an advantage for the Tegra2 but they are less than 30% and vary. I know that the Hummingbird is a single core processor, but these 3D benches aren't heavily utilizing the CPU so i figure the results shouldn't be skewed.

Far as Terga2 goes when it comes out the UPL GPU will still be faster than AMD Z460/Adreno205 inside 2GS SoC's. Only Hummingbird fanboy seem to think they for some silly reason has the upper hand lol even when the truth is that its still faster than SGX540 by small margin same as SGX540 faster than Adreno205 in small margin. But do to having dual core that's 2x threads each scaled at 1ghz will win.

iamlilysdad you know its serious Biz when HTC is on board lol as they are the stand alone leader in Smartphone, regardless of what other tells you Moto/Samsung you wish.
 

Trending Posts

Forum statistics

Threads
956,662
Messages
6,969,398
Members
3,163,595
Latest member
aaronr