• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Van Smith's COSBI benchmark in action - UPDATE: Download COSBI!

Van is always good for an amusing chuckle. I like the fact that the 1.7GHz Celeron sweeps several of the benchmarks against the 2.53GHz Pentium 4. And I won't even touch on the Athlon results.

Did you see where you can get the source for this Open Source benchmark? Or at least the compiler and compiler flags he used for it?
 
I'm not sure if he has released the source code, but I know he is using Delphi to compile the benchmark. As I understand it's an old compiler and never targetted for optimimal code but rather optimal compile times... Which is fine if you're developing a small windows applet but hardly fine for a CPU benchmark.
 
Delphi? I remember Delphi - it's the Borland equivalent of "Visual Pascal". Easy to read code but not exactly a mainstream for performance applications. Thanks for the information, Andreas.

I don't think Van is an idiot. He clearly is fairly intelligent. He's just very, very fanatical about his choice of CPU vendor and, like some other fanatics that I've met, he probably views the universe through a reality distortion field. 🙂

His "benchmark" still needs a bit of work though.
 
Heh... nice "CPU test".

How could a similar cpu with 67% of the mhz of another (not to mention 1/4 the cache and 3/4 the fsb) score higher?
rolleye.gif
(Btw, I'm comparing the 1.7ghz Celeron to the 2.53ghz P4.)

If I saw results as messed up as that, I'd choose not to post them.
 
What's really rich................apparently a few people have already gone to his "forums" and started asking questions and ripping COSBI.........convieniently the posts disappear and the member is banned by the "man" himself............😉 LOL! What a guy..............
rolleye.gif
 
I looked around on his site and found what seems to be an early version of his COSBI benchmark. Here you can check it out for yourself 🙂

LINK
 
Originally posted by: Adul
van is an idiot
Click to expand...

Van definitely isnt an idiot.
He's extremely intelligent, and has a lot of experience.
I'd lay bets he's a hell of a lot more intelligent then the overwhelming majority of people on this board.

Having talked to him in the past, I've no doubt at all that he's quite knowledgeable indeed.

Granted, he has a fanaticaly hatred for Intel and will boast any vendor that competes with them.
Specifically he has an unusual fondness for VIA, AMD and Transmeta.

He's full of crap most of the time, but that's because he's decided to devote his time to misleading others in an attempt to paint Intel as satan.
Not because he's not intelligent.


The Celeron 1.7GHz results are quite amusing to say the least.... I can think of perhaps a 1 in a billion situations in which the Celeron 1.7GHz may under some extremely unusual instances very marginally outperform a Northwood 1.7/400.
But the requirements for such an situation to happen are obviously not even remotely realistic, unless one was to almost code specifically for that situation and absolutely ensure that the L2 cache played no part in the test results beyond the minor overhead for an L2 cache look-up.
I certainly cannot imagine that happening in ANY real-world application code.
 
How can you even make a Celeron 1.7 beat a 2.53Ghz p4 ?

I mean, the p4 is better in every conceivable way possible; it should always get a higher score unless you specifically write code to halve the scores of the p4 on output. I'd have to really stress my thinking to write something where a Celeron 1.7 could win significantly against a 2.53 p4 if that is indeed possible.
 
What makes the Celeron and Pentium 4 2.53GHz results even more interesting is that the 2.53GHz Pentium 4 (Precision 340) is a Dell Workstation with the i850E chipset and Rambus, and the 1.7GHz Celeron (HP 512c) uses the i845G chipset with PC1600 DDR. The HP 512c is literally a bargain basement retail computer and the Dell is part of Dell's workstation product line.

I'm starting to become very curious as to what could possibly have gone wrong here. I wonder if the Dell wasn't busy doing something else when they benchmarked it. It's puzzling as well because Dell usually leaves their computers relatively free of all of the user-friendly "junk" that slows down retail computers by filling up the taskbar while HP is famous for it.
 
The most obvious thing about this benchmark is the fact that it's heaviliy influenced by *platform*.

It seems as if the platform makes the biggest difference in this situation.
 
On his forums Van is blaming the strange results on 2D video card performance and differences in color depth (16bit vs 32bit). He also explains how the benchmark basically measures how fast your computer can word process and notes that
Word is derived from RichEdit so at least for some word processing tasks your Athlon XP will be much, much faster than the P4.
Click to expand...
and
the Pentium 4 is apparently very, very poor at this test when compared to the Athlon XP
Click to expand...
rolleye.gif
 
Ah....A benchmark from Van huh?

I presume that part of the code checks for the string "GenuineIntel", and then proceeds to run the benchmark with the following flags:

MMX: Off
SSE: Off
SSE2: Off
FastSave&Restore: Off
AGP: Off
L2 cache: Off

rolleye.gif
 
Yeah.........now go try to argue the point with him..........I sat there for about 15 minutes and watched posts appear, only to be gone when the thread was refreshed....................a thread was started also and was gone before the second post could be made! Damn.................if you've got enough guts to print BS results like those you should at least be man enough to take the heat when the results are questioned.............
 
Originally posted by: dexvx
How can you even make a Celeron 1.7 beat a 2.53Ghz p4 ?
Click to expand...

Well were you to say test the L2 cache look-up overhead the Celeron could quite conceivably do better, as it'll have less overhead due to the smaller L2 cache size. Even so, on 100% identical systems the sheer MHz of the Northwood 2.53 should prop it far above the Celeron.

One would have to almost pick out individual tiny examples of specific scenarios to put together a benchmark that would show the Celeron performing better.
AcesHardware had an article quite awhile ago on the old CPUMark99 (I believe it was that test) and why it would show higher results on the Coppermine-Celeron rather then the PIII Coppermine, it gives a number of viable examples of scenarios in which one inferior processor could consistently yield better results then another hypotehtically superior processor with as few differences as the Celeron/PIII.

In any case, none of that is applicable to real world application usage.
The mere fact that there were clear differences between the P4 platforms should invalidate the test IMHO, if indeed he is specifically benchmarking the processor itself as stated.
 
Well were you to say test the L2 cache look-up overhead the Celeron could quite conceivably do better, as it'll have less overhead due to the smaller L2 cache size.
Click to expand...
This would probably be true in other architectures but in the case of the Celeron it isn't due to the way that the cache circuitry is set up. A cache tag lookup on the newer Celerons have an identical latency compared with the tag lookup on the microarchitectural family that it is a member of (in this case the Pentium 4).

I've run the "benchmark" myself and it really seems to me to be testing the 2D output capability of the video card along with a smaller emphasis on integer operations. Operations like "plot random dots" and "random circles" seem to be far less of a test of CPU performance than the framebuffer operations of the video cards. Having run the benchmark myself, I'm wondering as well if this may not be a issue of refresh rate.

Why are we even questionning the fact that VAN's benchmark could be accurate when we know its nothing to do with real world performance like sisoft sandra... to an extent ?
Click to expand...
Sandra scales fairly well within a given family (for example, Pentium 4's) whereas Van's (supposed) benchmark doesn't seem to give consistent results within a family. I have never seen a test on Sandra that showed, for example, a 1.7GHz Celeron to be faster than a 2.53GHz Pentium 4, but Van's tests seem to do this. Either one assumes that his benchmark is intended to be deliberately deceptive (which, to be honest, I doubt), or there is a problem in the method that he is using to perform the benchmark such that it is measuring something other than CPU or system performance. In this case, I'm wondering if the results aren't dependent on 2D video card performance, or the video cards refresh rate.
 
/*code*/

/*code*/


if(!Celeron && CPU.Speed >= 2000) {

for (int i =0; i<2000; i++)
NOOP();

}


/*code*/

/*code*/


//That answer anybody's question?


 
I think they should redo the test with all systems using the same videocard.

Then we can see what's really cooking here..

A much better CPU test, would be to turn the hardware acceleration notch down to 0. Then run the benchmark.
 
Is this supposed to be a joke? Why did he test stock machines at Sam's club instead of using standardized test systems?
 
ROFLMAO.

Van Smith really gives a good laugh. Sometimes he has a decent article and I -almost- think he is coming around ... then he comes up with some bazooka like COSBI.
 
2D score? The Dell Precision is armed with a GeForce2 Quadro class video adapter and the other guys have built in video. How on earth can you say a Quadro class video card is lower than the crap built unto cheapo computers? You have to remember that the 2.53 Ghz is Dell's Precision line, using mostly top of the line OEM goods.

If this supposed benchmark is testing for CPU speed, then people needn't worrry about other components within the computer. A CPU benchmark like Sisoft would give consistent results for CPU speed even if the 2 systems are very differently configured. (2 Ghz p4, i845 128MB SDR, TNT vs 2 Ghz p4, i850, 512 RDRam, GeForce4).

Correct me if im wrong, but there are Office Suite benchmarks for Word and they show the Pentium 4 on par with an Athlon.

I've never heard of Van's hardware before, but I dont think i'll be opting to visit his site anytime soon. Next thing you know he'll be saying Intel Gigabit NICs can be beat by 3Com 10/100's.

Edit:
Apparently the link to the COSBI bench you gave us is only version .2. Van uses version .5, just like to point out that there are some benchmarks in .2 that van removed in .5 and of course there are some benchmarks in .5 not available in .2.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top