Barcelona's non_rate SPEC benchies

[yomamafor1]

http://realworldtech.com/forums/index.cfm?action=detail&id=83478&threadid=83478&roomid=2

SPECint2006 (base/peak)
AMD 1.9GHz - 9.97/11.3 - Note 2 socket system
Intel 3GHz - 18.9/20.8 - Note 2 socket system
Intel 2GHz - 14.2/15.6 - Note 2 socket system

SPECfp2006 (base/peak)
AMD 1.9GHz - 10.7/11.2 - Note 2 socket system
Intel 3GHz - 18.4/21.4 - Note 2 socket system
Intel 2GHz - 14.5/16.9 - Note 2 socket system

EDIT: for tone.

Seems like Barcelona falls short of its projected IPC. Now AMD's last chance is to pump up the clockspeed...but might be a little too late, as Intel's Penryn is just a few months away.

 

[virtualban]

Where did you get these? I know, the link at the top of the post, but that's something everybody could have posted!

 

[jaguarskx]

However there is a disclaimer for the seemingly higher Intel SPECint2006 benchmarks:

There are higher SPECint2006 scores for Intel, but they use some sort of OpenMP flags which makes me think they might have been doing something 'odd' and unreflective of the benchmark.

 

[yomamafor1]

This message was posted by David Kanter, one of the respected tech journalists out there.

I'll dig a little deeper into what "OpenMP" does, but as for AMD's main advantage, FP, also falls a little short of its projection. Seems like AMD's only advantage now is scalability....

 

[easyg]

These results are for IBM System x 3455 (AMD Opteron 2347). You can find more detailed info at SPEC at the links below:

http://www.spec.org/cpu2006/results/res2007q3/cpu2006-20070903-01948.html

http://www.spec.org/cpu2006/results/res2007q3/cpu2006-20070903-01949.html

No, these results aren't very good at all.

 

[dobby]

i have siad to may a amd fnboy that Bench marks need proving, same applies visa verca when disgracing chip. you best evidence for AMDs new phenom chip to lag behind intel is that their have been no chips to have real 3rd party bench marks, odd...

 

[dobby]

ironically on my browser (IE7 - no comments plaese in my OPINION its better than FF2 - but i have given it a fair try and i will do the same for FF3) there is AMD advertising all over ther given webpage that the link points to

 

[BaronMatrix]

http://realworldtech.com/forums/index.cfm?action=detail&id=83478&threadid=83478&roomid=2

SPECint2006 (base/peak)
AMD 1.9GHz - 9.97/11.3 - Note 2 socket system
Intel 3GHz - 18.9/20.8 - Note 2 socket system
Intel 2GHz - 14.2/15.6 - Note 2 socket system

SPECfp2006 (base/peak)
AMD 1.9GHz - 10.7/11.2 - Note 2 socket system
Intel 3GHz - 18.4/21.4 - Note 2 socket system
Intel 2GHz - 14.5/16.9 - Note 2 socket system

EDIT: for tone.

Seems like Barcelona falls short of its projected IPC. Now AMD's last chance is to pump up the clockspeed...but might be a little too late, as Intel's Penryn is just a few months away.

If you were to review all of the SPEC scores you would find wild differences in the same chips in a different machine. HP usually has the highest Opteron scores for some reason. Let's see what happens when more vendors post their scores.

And please don't say it's an excuse. Review the scores on SPEC.

 

[elbert]

http://realworldtech.com/forums/index.cfm?action=detail&id=83478&threadid=83478&roomid=2

SPECint2006 (base/peak)
AMD 1.9GHz - 9.97/11.3 - Note 2 socket system
Intel 3GHz - 18.9/20.8 - Note 2 socket system
Intel 2GHz - 14.2/15.6 - Note 2 socket system

SPECfp2006 (base/peak)
AMD 1.9GHz - 10.7/11.2 - Note 2 socket system
Intel 3GHz - 18.4/21.4 - Note 2 socket system
Intel 2GHz - 14.5/16.9 - Note 2 socket system

EDIT: for tone.

Seems like Barcelona falls short of its projected IPC. Now AMD's last chance is to pump up the clockspeed...but might be a little too late, as Intel's Penryn is just a few months away.

I dont think it really matters if a single core of a Intel can beat a single core of an amd. The single threaded performance benchmarks have lost all meaning just a short time after dual cores arrived. The article states this fact. The point here is amd's quad will see less loss in each cores performance than Intel's due to Intel's quad is a 2Xdual which will take a hit once the 3 and 4 core comes into play. Note the quote from the same link.

The good news is that single threaded performance is less likely to be a key purchasing concern for servers, so this isn't necessarily a substantial problem for AMD.

 

[rodney_ws]

ironically on my browser (IE7 - no comments plaese in my OPINION its better than FF2 - but i have given it a fair try and i will do the same for FF3) there is AMD advertising all over ther given webpage that the link points to

That's funny... I use FF2 and AdBlock+ and I see no ads... anywhere

 

[MikePHD]

The intel benchmarks make use of a -parallel flag [for compile time] which allows the compiler to optimize the code for multi-core.

The amd tests were run without it. These numbers are comparing apples to oranges, since it really is a compiler issue, and does not reflect the capability of the hardware.

Now if you want to argue that intel provides better support for compiling optimizations, you have proof for that right there : ).

 

[yomamafor1]

I dont think it really matters if a single core of a Intel can beat a single core of an amd. The single threaded performance benchmarks have lost all meaning just a short time after dual cores arrived. The article states this fact. The point here is amd's quad will see less loss in each cores performance than Intel's due to Intel's quad is a 2Xdual which will take a hit once the 3 and 4 core comes into play. Note the quote from the same link.

The good news is that single threaded performance is less likely to be a key purchasing concern for servers, so this isn't necessarily a substantial problem for AMD.

True. Due to native quad core design, AMD's still has the last advantage of scalability.

Proof:
In SPEC_Int_rate, which test the system as a whole, here are the scores.
AMD base rate: 77.3
AMD peak rate: 88.3

Intel base rate: 58.8
Intel peak rate: N/A

http://www.spec.org/cpu2006/results/res2007q4/cpu2006-20070908-01993.html
http://www.spec.org/cpu2006/results/res2007q3/cpu2006-20070820-01863.html

All of the machines tested are 2 socket systems Compare this with the data I posted above, its pretty clear that AMD can outperform when 2 CPUs are used. But for single threaded performance, for IPC, K10 simply cannot hold its ground against Core 2

I know that this may seem a little iffy because AMD submitted both tests, so there is no way to tell how it was ran. However, for simple illustration, we'll use it.

As for Baron, you said that result is all over the place. However I only see one SPEC_int result, one SPEC_fp result, and several other SPEC_int rate and SPEC_fp rate result. What is the "jump all over the place" you talked about? You know I'm a lazy person, so would you please post a link for us?

 

[yomamafor1]

The intel benchmarks make use of a -parallel flag [for compile time] which allows the compiler to optimize the code for multi-core.

The amd tests were run without it. These numbers are comparing apples to oranges, since it really is a compiler issue, and does not reflect the capability of the hardware.

Now if you want to argue that intel provides better support for compiling optimizations, you have proof for that right there : ).

SPEC int is a single threaded benchmark, so having an multi-threaded optimization may not skew the results too far off. However, since I'm not a CS major like scientia, I don't really know how that optimizer will affect the total score.

Assuming that it does affect the score significantly, we can simply toss it out, and focus on SPEC_fp, which is supposedly AMD's main selling point.