Guys....
First off, I am not going to start another [H] thing .... but I wanted to show you something
here's an old thread from Gibbo at amdmb on the "speed cheat" and how to detect fake scores...
http://www.amdforums.com/showthread....ht=speed+cheat
here's a quote
Obviously there will always be a spot of 'experimental error' as no 2 runs give exactly the same result, but it can be used as a general guide.
So, the easy way to do this for someone's posted results is divide their score by the score for a 1.0 CPU (in this case TBird) given by Sandra...
To take his ALU Dhrystone score from the CPU bench:
4862/2805 = 1.733
So, ******* claim:
CPU Benchmark
ALU = 4862 Dhrystones ~1.733GHz
FPU = 2763 Whetstones ~2.009GHz
CPU Multimedia Benchmark
Integer MMX Enh = 9648 it/s ~1.766GHz
Floating Point 3DNow! = 11171 it/s ~1.633GHz
Gibbo's reality:
CPU Benchmark
ALU = 5092 Dhrystones ~1.815GHz
FPU = 2516 Whetstones ~1.829GHz
Hope this helps everyone
Now ...
Those who read [H]'s fantastic article by Steven Lynch, may have noticed the PCM score is only 9500 +/-
This puts his 4.44 GHz system right on par with ... a 3.7 GHz P4 ....
Kickass...
Of course no compare link to see the actual speed it was running
I forgot to mention that this beast is cooled by a Vapochill PE - no kidding!
4.44 GHz@1.85V .... poor vapochill.... must see 15-20C load temps on the core... but ok thats not the point.
Also the other benchmark has an error written in it but I dont know the significance of that...
Basically, the only thing which I consider real is the 4.44 GHz POST ... which is not that important since I could POST at more than 4.5 GHz with my 3.06 but it didnt run any higher than 3150 for 3D still.... so a POST isnt important
Then, what we have left is the SiSoft sandra screenshots. We know that the "speed cheat" will deform sandra scores slightly since it simply multiplies them linearly...
remember from Gibbo's earlier proof that in the CPU bench, the FPU will be too high compared to the Int ... and the other way around for the multimedia test.
look at this... the math here is...
we take the 2.80 SMT system as reference for the SMT score at 4.44 GHz. Since SiSoft counts pretty damn predictable, (you can try that on your own comp) it will always be right within some margin.... the CPU speed that is. The FPU and ALU should still come down to roughly the same clock speed
The CPU speed in the calculation came out too high, but not that high that its very obvious... the margin of error still acceptable here... but the difference between the ALU and FPU is definately a speed hack signature...
can it be more obvious..? 4.90 GHz is getting more than 4.44 by quite a margin now...
to sum things up.... the Sandra screenies clearly show the speed cheat being applied... without this program the values would never be so far apart....
The only thing which is more or less un-fakable is PCmark, and we see an amazingly LOW score there, on par with a 3.7 GHz Pentium 4... thats about 750 MHz short of 4.44 though.
I'll let everyone make up their own mind about this .... but using speed cheat when you have a CPU actually running 4.44 would seem unnecessary
THIS IS NOT A 100% SURE WAY to see if its actually been cheated .... just something to think about, nothing more. Its not like we care so lets not waste too much energy on it thats my suggestion...
Reply With Quote
Bookmarks