Dual core Scaling: Yonah and X2

[IntelUser2000]

Here's what I calculated and concluded.

X2's(4800+, 2.4GHz 1MB L2) performance advantage vs 64(4000+, 2.4GHz 1MB L2), normalized to same clock speed and cache(1st score is X2 and second is 64, third is performance advantage), link is here: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2410&p=1

Business/General Use Performance
Business Winstone 2004- 23.4/23.6= -0.85%
Sysmark 2004 Communication-143/150= -4.67%
Document Creation- 247/195= 26.7%
Data Analysis- 159/146= 8.9%
Microsoft Office XP SP-2- 507s/521s= 2.76%
Mozilla 1.4- 315s/279s= -11.4%
ACD Systems ACDSee PowerPack 5.0- 497s/485s= -2.4%
Ahead Software Nero Express 6.0.0.3- 507s/492s= -2.96%
Winzip- 391s/398s= 1.79%

Business/General Use Avg. performance increase= 1.99%

Multitasking Content Creation
Multimedia Content Creation Winstone 2004- 41.6/36.1= 15.2%
3D Content Creation- 291/205= 42%
2D Content Creation- 365/260= 40.3%
Web Publication- 289/187= 54.5%
Mozilla+Media Encoder- 370s/550s= 48.6%
Multitasking Content Creation Avg. performance increase= 40.1%

Video Creation/Photo Editing
Adobe Photoshop 7.0.1- 311s/314s= 0.96%
Adobe Premier 6.5- 360s/368s= 2.22%
Roxio VideoWave Movie Creator 1.5- 246s/306s= 24.4%
Video Creation/Photo Editing Avg. performance increase= 9.19%

Audio/Video Encoding
MusicMatch Jukebox 7.10- 447s/458s= 2.46%
Windows Media Encoder 9- 263s/364s= 38.4%
Audio/Video Encoding Avg. performance increase= 20.4%

3D Rendering
Discreet 3dsmax 5.1-DirectX- 237s/244s= 2.95%
OpenGL- 297s/300s= 1%
Discreet 3dsmax 6(OpenGL)-SPECapc Rendering Composite- 2.78/1.53= 81.7%
Discreet 3dsmax 6(OpenGL)-3dsmax5.rays- 12.985s/20.968s= 61.5%
CBALLS2- 18.547s/34.766s= 87.4%
SinglePipe2- 86.125s/169.562s= 96.7%
Underwater- 126.813s/232.516s= 83.4%
3D Rendering Avg. performance increase= 59.2%

Overall Performance increase of X2 vs 64= 27%

Yonah's(Intel Core Duo, 2.0GHz, 2MB L2) performance advantage vs Dothan(Pentium M 760, 2.0GHz, 2MB L2), using the score from: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2648
The benchmarks used are same as the one used from X2 vs 64 in order to compare what CPUs benefit more going to dual core. (Yonah/Dothan=%)

Business/General Use Performance
Business Winstone 2004- 21.6/22.8= -5.56%
Sysmark 2004 Communication- 146/140= 4.29%
Document Creation- 215/203= 5.91%
Data Analysis- 138/117= 17.9%
Microsoft Office XP SP-2- 531s/505s= -4.9%
Mozilla 1.4- 315s/279s= 338s/310s= -8.28%
ACD Systems ACDSee PowerPack 5.0- 544s/545s= 0.2%
Ahead Software Nero Express 6.0.0.3- 512s/510s= -0.4%
Winzip- 396s/399s= 0.8%

Business/General Use Avg performance increase: 1.65%

Multitasking Content Creation
Multimedia Content Creation Winstone 2004- 34.7/28.3= 22.6%
3D Content Creation- 264/181= 45.6%
2D Content Creation- 323/238= 35.7%
Web Publication- 236/162= 45.7%
Mozilla+Media Encoder- 439s/634s= 44.4%
Multitasking Content Creation Avg performance increase= 38.8%

Video Creation/Photo Editing
Adobe Photoshop 7.0.1- 319s/325s= 1.9%
Adobe Premier 6.5- 405s/404s= -0.25%
Roxio VideoWave Movie Creator 1.5- 267s/400s= 49.8%
Video Creation/Photo Editing Avg performance increase= 17.1%

Audio/Video Encoding
MusicMatch Jukebox 7.10- 511s/529s= 3.52%
Windows Media Encoder 9- 295s/415s= 40.7%
Audio/Video Encoding Avg performance increase= 22.1%

3D Rendering
Discreet 3dsmax 5.1-DirectX- 256s/256s= 0%
OpenGL- 320s/322s= 0.63%
Discreet 3dsmax 6(OpenGL)-SPECapc Rendering Composite- 2.4/1.28= 87.5%
Discreet 3dsmax 6(OpenGL)-3dsmax5.rays- 14.297s/25.246s= 76.6%
CBALLS2- 21.287s/42.201s= 98.2%
SinglePipe2- 107.359s/203.492s= 89.5%
Underwater- 169.188s/316.055s= 86.8%

3D Rendering average performance increase= 62.7%


Overall performance increase of Yonah vs. Dothan= 28.4%

Overall Yonah gains more from Dothan than X2 does over 64. The difference overall is minute, with 3D rendering, Audio/Video Encoding, Yonah gains more, and Business/General Use, Multitasking Content Creation, Video Creation/Photo Editing, X2 gains more. It looks as though Yonah has more consistent performance increase, and X2 has little more fluctuations.

 

[Screech]

Dothan and Core Duo are not the same core.....as opposed to x2 and athlon 64. Hence, that slight increase likely results from the improvements to the yonah core over the dothan.

A more apples-apples comparison would be Core Duo/Core solo (assuming i got the names right) vs X2/64.

 

[IntelUser2000]

True, but I hate that comment about apples to apples to comparison. I hated when people say that Yonah should have an IMC while X2 should have 65nm, and DD2 and all that bullcrap.

The truth is that Intel's procs aren't going IMC anytime soon. AMD isn't going DDR2 until couple months after Yonah, while Yonah will be available in two weeks, specifically Jan 5th with CES. AMD's 65nm won't be out until Q4 of next year.

You can never equalize.

Though it will be interesting how Dothan would scale by going to dual core, in reality, people who has enough money to buy those kind of high end CPUs aren't gonna go for single core.

Look at the prices:
http://www.digitimes.com/mobos/a20050609A2006.html

The will be only available single core, the 1.66GHz is only $40 less than dual core of the same clock speed. Who the hell in the right mind would get a very low clocked single core when its only $40 less than the dual core of the same clock speed??

Only equal comparison is against Dothan, since it will be priced similarly, to Yonah with approximately one clock speed step less.

 

[Screech]

I agree with what you said--especially on the $40 for the extra core--and the 'apples to apples' thing is rediculously overused . However, the point stands for the purpose of dual core performance (without taking power consumption into account) scaling comparison, if you simply want to see how well the CPU will do with the added core.

 

[IntelUser2000]

Yea ok but in practical terms it doesn't matter. As long as we are theoretical, we can even speculate 5 years from now.

 

[IntelUser2000]

Let's look this way:
Yonah uses shared cache rather than independent cache and arbitration logic(or SRQ for AMD, which are really arbitration logic in a marketing term). I thought shared cache procs need arbitration logic, but the shared cache nature negates the need for that.

So they are not even using same kind of dual core architectures. For single core, Intel may be at disadvantage if it weren't for all the other enhancements not including dual core. Per core Intel has half the cache of the previous gen, while X2 has same cache, which decreases Yonah's performance advantage.

Core Duo, has 2MB L2 over 2 cores
Core Solo, has 2MB L2 over 1 core

X2, has 2MB L2 over 2 cores
64, has 1MB L2 over 1 core

Obviously, with dual core alone, X2 is gonna scale better, if dual core Dothan had an IMC.

Plus, being only 11 million more transistors than Dothan, die size of Yonah is pretty big. Yonah shouldn't have been more than 70mm2 when it has been shrunk fully from 90nm, but the 65nm Yonah isn't half the size of Dothan, its more likely 75%. I assume its to keep power/density similar or better. Prescott uses more power with smaller die size, which equals HOT HOT HOT.

 

[IntelUser2000]

Obviously, with dual core alone, X2 is gonna scale better, if dual core Dothan had an IMC.

Sorry, I meant to say SRQ rather than IMC, my bad.