I looked at your bencmarking setup and would like to point out a few things. A) Driver differences count. Your using the Window V3 drivers, and 3DFx made a lot of progress in the OpenGL support of the V3. (IE. V3 miniGL is a lot faster than V2 miniGL) B) The V3 has nearly double the clock speed, and a more optimized pipeline. C) You ran with stuff like sound off. The BeNews guys ran it with everything at default. Add all these factors together with the 17MHz difference, and the results are quite plausible. Also, you seem to have a histroy of conspiracy theories, saying that Tom's Hardware is biased. This probably dates back to the fact that Tom hand a nVidia approved logo on his site during the TNT vs. V3 wars. That was simply an acknowledgement of the fact that nVidia thought that Tom's methods were sound. Also, V3 was never as fast as TNT2. That fact is acknowledged by every major game rag, so I don't really get your viewpoint.
I am not blindly pro Be. I acknoledge their weaknesses, like there lack of a deep UI and the fact that networking sucks. The simple fact is that these benchmarks are not skewed. Think about it, Voodoo so far is about the best Linux 3D driver. 3DFx has poured an enormous amount of energy into it. 3DFx is also the fastest BeOS driver at this time. The use of "outdated" hardware, is a no brainer in this case, because it is the only driver that works well in this beta! release. This is not a guantlet to Linux users everywhere that Be is king. It is simply a status test. As such, it is not designed to be the perfect test, merely the best that can be achieved under the circumstances.
Sorry, it's been a long time since I had a look at Tom's Hardware. However, 3D Voodoo5 is in the pipeline for a patch to the R5 release. I assume it should be out before the fall. (Look on benews about articles about the new OpenGL implementation. I seem to remember something about Voodoo5 being there.) Also, 3DFx has already pleged BeOS support, so I'd wouldn't be suprised if the driver was available with the OpenGL release.
Actually, I have version 1, and it doesn't contain GIMP. I know, though, that it uses the GIMP filters. The GIMP project on BeOS is still alpha (as is the GTK+ port) but is progressing nicely.
Have you ever considered that it is just your hardware? Video works great for loads of other people, and even for me (who does graphics programming and BDirectWindow stuff, apps which tend to crash 50,000 times during their development) R5 hasn't crashed yet.
Read the whole thing. It says that turning off the procs yeilds a.6fps loss. However, the fact that the OpenGL on BeOS is multithreaded should mean big gains for apps that use OpenGL more extensively (other than just a simple rendering engine like Quake does), as they could leverage multiple procs automatically. Be's OpenGl also takes good advantage of SIMD, so stuff that uses the OpenGL geometry pipe should perform even better than Quake!
Just so you know, the development tools are a free download from Be's website. They did'nt put them in the main download because it's about a 20 meg zip file.
It has a big effect on stuff like audo composition, especially midi effects. If your system is fast enough to respond to sound events quickly, (especially MIDI sound events, since they aren't really streams of data) then you can mix more tracks of audio while still staying real time. Also, you can add more advanced filters and stuff and know that the system won't hold up the data and keep the tracks from playing in real time. Ff you've ever seen BeOS play two dozen MP3s, you'll know why audio latency matters.
I haven't seen the Be logo in so long I forgot what it looked like. Why the hell was this posted under graphics? Any OpenGL news about Linux is always posted under the penguin, especially the nVidia driver stuff, since that is about the same case as this. Or was the author just aware of the fact that anything having Be as the topic is automatically forwared to/dev/null?
I'm still right. I'm saying that they didn't use it in the conclusion. That means a lot, because the "smokes" part is probably where BeOS outperformed Linux and Windows by 50% on Crusher. And what's to get wrong? The fact is, that for Quake II on Voodoo2, BeOS smokes Linux and even Windows. That's the right idea!
Flawed reasoning! The system, first of all, was dual P3. Since you didn't catch that, you probably didn't catch the fact that they DID test without SMP, and the numbers were only.6fps lower, since QuakeII is not SMP.
Did you even read the freaking article? They used a freaking Voodoo2 running on Celerons! The sheer fact that their numbers are even near half yours is a miracle! And BeNews is not a huge operation. They used an SMP Celeron because that's what they had, and they used a Voodoo2 because that's what ran well.
Hmm, it occurs to me that they had better tested against XFree86 4.0 since that's a big step towards hardware acceleration support. >>>>>>>. They used a voodoo 2. Those perform better under 3.3.6. If they'd used 4.0, then you COULD have rightfully ripped them to shreds for not portraying Linux in the best light.
For the right graphics boards the Linux numbers should have been a lot more like the W98 numbers (extrapolating from Q3A >>>>> Voodoo2 is a good graphics board for both BeOS and Linux, and that's why it was used. Also, Q3A doesn't run on BeOS.
Okay, so now we can say that Linux nVidia drivers don't really perform as well as Windows because the test was flawed since it was run by a Linux website?
Yes, the article is a simple test given to the BeOS community about the results of the OpenGL beta testing. It was not meant to say BeOS smokes Linux, the conclusion doesn't even use the word "smokes!" they don't even use a superlative, simply saying that 'it is clear BeOS has made a very good OpenGL implementation' (paraphrased.) It was never meant to be a declartion of superiority by the BeOS community. It was simply a status report to BeOS people. If Slashdot escalated it such, then it's/.'s fault, not BeNew's.
Did you guys even read the whole article? How, exactly, is this benchmarketing? The guys put up a test between 3 OSs running an industry standard 3D test, namely Quake. Everyone and their mother has used Quake to test 3D cards and OpenGL. Hell, 3DLinux used Quake to test the nVidia Linux drivers. The tests showed that BeOS was faster in most cases, especially windowed operations. Is it benchmarketing simply because Linux didn't win? BeOS has shown this kind of amazing performance in other fields too, like audio latency around 2 or 3 millseconds opposed to around 50ms in Windows. Is there something wrong with the testing method? True, the machine used was SMP, but if you read the whole article, the BeNews guys said that SMP only lowered the score by about.6fps. Okay, so now it is 72.6 to 77.8 instead of 78.4! It isn't like this is a synthetic benchmark, it was Quake. If Be did something to cheat Quake benchmarks, then good for them. That means that any game that is Quake-like in terms of 3D use (ie. all of them) will run faster. If you're complaining about the narrow scope of the test, remember that Voodoo2's are the most stable cards available on BeOS, and that QuakeII is the only major 3D application. So what's the beef? A lot of people have tossed around the term benchmarketing, but if anyone can give me a specific example, then I'd be glad to listen.
Are you kidding? BeOS's GL output looks slightly better than Win98's GL output! And how is this bench marketing? Just because it's by BeNews? I got news for you, nVidia's OpenGL drivers were tested by 3DLinux. How's that for bench marketing? The tests were similar except one ran QuakeII, and the Linux guys ran Quake III.
I do think, however, that you're analogy is a bit wrong. Be has most of the cool UNIX stuff you like in Linux, but its hidden away under the / directory. (For you non Be users, the GUI has/boot as the reference so stuff like/bin,/lib, and/etc are never seen by the user.) It's more like a TV with an easily removed cover, wheras Win98 is like a TV with the cover welded shut. (For good reason, people would puke if they ever say what was in there!)
The main reason is that Quake isn't SMP enabled. If you'd read the article, they say that disabling SMP only lowered the score by.6fps, and the SMP machine was the only one available at the time.
True, but what did 3D Linux use to test nVidia support? Hmm, I don't know, Quake? The truth is that there are no professional calibar 3D programs available on BeOS (and Linux for that matter.) This should change with the introduction of Maxon'x 3D modeler and the ports of Maya to Linux. Also, 3D apps don't really use OpenGL for more than previewing so it doesn't matter. What I want to see, though, is some viewperf scores.
The BeOS filesystem is a hideosly fast journaling implementation called bfs. It IS faster than anything available on Linux. Messaging speed is also top notch. I don't know how you benchmark mem. management, but BeOS has good memory protection and uses about half the memory of Linux/GNOME on the same machine.
Yes, Be doesn't smoke in this area. However, as the other guy said, BeNews IS running RedHat Linux, so I wouldn't be talking. Be is coming up with a new networking architecture called BONE which should remedy most of that. Try it out at www.betips.com
Slashdot Mirror
I looked at your bencmarking setup and would like to point out a few things.
A) Driver differences count. Your using the Window V3 drivers, and 3DFx made a lot of progress in the OpenGL support of the V3. (IE. V3 miniGL is a lot faster than V2 miniGL)
B) The V3 has nearly double the clock speed, and a more optimized pipeline.
C) You ran with stuff like sound off. The BeNews guys ran it with everything at default.
Add all these factors together with the 17MHz difference, and the results are quite plausible. Also, you seem to have a histroy of conspiracy theories, saying that Tom's Hardware is biased. This probably dates back to the fact that Tom hand a nVidia approved logo on his site during the TNT vs. V3 wars. That was simply an acknowledgement of the fact that nVidia thought that Tom's methods were sound. Also, V3 was never as fast as TNT2. That fact is acknowledged by every major game rag, so I don't really get your viewpoint.
I am not blindly pro Be. I acknoledge their weaknesses, like there lack of a deep UI and the fact that networking sucks. The simple fact is that these benchmarks are not skewed. Think about it, Voodoo so far is about the best Linux 3D driver. 3DFx has poured an enormous amount of energy into it. 3DFx is also the fastest BeOS driver at this time. The use of "outdated" hardware, is a no brainer in this case, because it is the only driver that works well in this beta! release. This is not a guantlet to Linux users everywhere that Be is king. It is simply a status test. As such, it is not designed to be the perfect test, merely the best that can be achieved under the circumstances.
Sorry, it's been a long time since I had a look at Tom's Hardware. However, 3D Voodoo5 is in the pipeline for a patch to the R5 release. I assume it should be out before the fall. (Look on benews about articles about the new OpenGL implementation. I seem to remember something about Voodoo5 being there.) Also, 3DFx has already pleged BeOS support, so I'd wouldn't be suprised if the driver was available with the OpenGL release.
Actually, I have version 1, and it doesn't contain GIMP. I know, though, that it uses the GIMP filters. The GIMP project on BeOS is still alpha (as is the GTK+ port) but is progressing nicely.
Have you ever considered that it is just your hardware? Video works great for loads of other people, and even for me (who does graphics programming and BDirectWindow stuff, apps which tend to crash 50,000 times during their development) R5 hasn't crashed yet.
Read the whole thing. It says that turning off the procs yeilds a .6fps loss. However, the fact that the OpenGL on BeOS is multithreaded should mean big gains for apps that use OpenGL more extensively (other than just a simple rendering engine like Quake does), as they could leverage multiple procs automatically. Be's OpenGl also takes good advantage of SIMD, so stuff that uses the OpenGL geometry pipe should perform even better than Quake!
Sorry, wrong again. The QuakeII port has been around since the 4.x days, when OpenGL on BeOS wasn't really that good.
Just so you know, the development tools are a free download from Be's website. They did'nt put them in the main download because it's about a 20 meg zip file.
It has a big effect on stuff like audo composition, especially midi effects. If your system is fast enough to respond to sound events quickly, (especially MIDI sound events, since they aren't really streams of data) then you can mix more tracks of audio while still staying real time. Also, you can add more advanced filters and stuff and know that the system won't hold up the data and keep the tracks from playing in real time. Ff you've ever seen BeOS play two dozen MP3s, you'll know why audio latency matters.
I haven't seen the Be logo in so long I forgot what it looked like. Why the hell was this posted under graphics? Any OpenGL news about Linux is always posted under the penguin, especially the nVidia driver stuff, since that is about the same case as this. Or was the author just aware of the fact that anything having Be as the topic is automatically forwared to /dev/null?
I'm still right. I'm saying that they didn't use it in the conclusion. That means a lot, because the "smokes" part is probably where BeOS outperformed Linux and Windows by 50% on Crusher. And what's to get wrong? The fact is, that for Quake II on Voodoo2, BeOS smokes Linux and even Windows. That's the right idea!
Flawed reasoning! The system, first of all, was dual P3. Since you didn't catch that, you probably didn't catch the fact that they DID test without SMP, and the numbers were only .6fps lower, since QuakeII is not SMP.
Did you even read the freaking article? They used a freaking Voodoo2 running on Celerons! The sheer fact that their numbers are even near half yours is a miracle! And BeNews is not a huge operation. They used an SMP Celeron because that's what they had, and they used a Voodoo2 because that's what ran well.
Hmm, it occurs to me that they had better tested against XFree86 4.0 since that's a big step towards hardware acceleration support.
>>>>>>>.
They used a voodoo 2. Those perform better under 3.3.6. If they'd used 4.0, then you COULD have rightfully ripped them to shreds for not portraying Linux in the best light.
For the right graphics boards the Linux numbers should have been a lot more like the W98 numbers (extrapolating from Q3A
>>>>>
Voodoo2 is a good graphics board for both BeOS and Linux, and that's why it was used. Also, Q3A doesn't run on BeOS.
This isn't a rigorous evaluation. It isn't even a final word. It is just some preliminary benchmarks about the OpenGL beta meant as a status report.
Okay, so now we can say that Linux nVidia drivers don't really perform as well as Windows because the test was flawed since it was run by a Linux website?
Yes, the article is a simple test given to the BeOS community about the results of the OpenGL beta testing. It was not meant to say BeOS smokes Linux, the conclusion doesn't even use the word "smokes!" they don't even use a superlative, simply saying that 'it is clear BeOS has made a very good OpenGL implementation' (paraphrased.) It was never meant to be a declartion of superiority by the BeOS community. It was simply a status report to BeOS people. If Slashdot escalated it such, then it's /.'s fault, not BeNew's.
Did you guys even read the whole article? How, exactly, is this benchmarketing? The guys put up a test between 3 OSs running an industry standard 3D test, namely Quake. Everyone and their mother has used Quake to test 3D cards and OpenGL. Hell, 3DLinux used Quake to test the nVidia Linux drivers. The tests showed that BeOS was faster in most cases, especially windowed operations. Is it benchmarketing simply because Linux didn't win? BeOS has shown this kind of amazing performance in other fields too, like audio latency around 2 or 3 millseconds opposed to around 50ms in Windows. Is there something wrong with the testing method? True, the machine used was SMP, but if you read the whole article, the BeNews guys said that SMP only lowered the score by about .6fps. Okay, so now it is 72.6 to 77.8 instead of 78.4! It isn't like this is a synthetic benchmark, it was Quake. If Be did something to cheat Quake benchmarks, then good for them. That means that any game that is Quake-like in terms of 3D use (ie. all of them) will run faster. If you're complaining about the narrow scope of the test, remember that Voodoo2's are the most stable cards available on BeOS, and that QuakeII is the only major 3D application. So what's the beef? A lot of people have tossed around the term benchmarketing, but if anyone can give me a specific example, then I'd be glad to listen.
Sorry about that. I've been paying too much attention to GeForce 2.
Are you kidding? BeOS's GL output looks slightly better than Win98's GL output! And how is this bench marketing? Just because it's by BeNews? I got news for you, nVidia's OpenGL drivers were tested by 3DLinux. How's that for bench marketing? The tests were similar except one ran QuakeII, and the Linux guys ran Quake III.
I do think, however, that you're analogy is a bit wrong. Be has most of the cool UNIX stuff you like in Linux, but its hidden away under the / directory. (For you non Be users, the GUI has /boot as the reference so stuff like /bin, /lib, and /etc are never seen by the user.) It's more like a TV with an easily removed cover, wheras Win98 is like a TV with the cover welded shut. (For good reason, people would puke if they ever say what was in there!)
The main reason is that Quake isn't SMP enabled. If you'd read the article, they say that disabling SMP only lowered the score by .6fps, and the SMP machine was the only one available at the time.
True, but what did 3D Linux use to test nVidia support? Hmm, I don't know, Quake? The truth is that there are no professional calibar 3D programs available on BeOS (and Linux for that matter.) This should change with the introduction of Maxon'x 3D modeler and the ports of Maya to Linux. Also, 3D apps don't really use OpenGL for more than previewing so it doesn't matter. What I want to see, though, is some viewperf scores.
The BeOS filesystem is a hideosly fast journaling implementation called bfs. It IS faster than anything available on Linux. Messaging speed is also top notch. I don't know how you benchmark mem. management, but BeOS has good memory protection and uses about half the memory of Linux/GNOME on the same machine.
Yes, Be doesn't smoke in this area. However, as the other guy said, BeNews IS running RedHat Linux, so I wouldn't be talking. Be is coming up with a new networking architecture called BONE which should remedy most of that. Try it out at www.betips.com