By reading the review at: http://www.presence-pc.com/articlev3.php?article=95, you can conclude that Doom3 will not be a revolution in terms of DirectX or OpenGL usage. Still this is nearly unbeliveble, since ID is supposed to be cooperating with NVidia, and thus they have no benefit in designing a game that would require DirectX7 / OpenGL 1.2...

Also, given the fact that they advertised somehow to the magazines that they will use Pixel Shader's and Bump Mappings, dynamic shadows... Hmmm, you can't do that in DirectX 7, so I wonder why such an information can be found, and more, why it has been released...

Right here http://my.fit.edu/~mmichael/Smileys/deal.gif

Once translated that does provide good info -- basically put -- the 3Dfx hardware does not have the capabilities to run it -- the hardware is too outta date -- get an ATI Radeon (or GeForce if you are daring).

This would be DOG SLOW if it could be forced to run on a 3Dfx card -- not only does it not have the new hardware capabilities but it also doesn't have the clock/memory speeds to keep up - nor does it support bus-mastering, AGP 4X, GART or other important functions.

I'd get a radeon and build a new system using the Pentium 4 extreme w/hyperthreading.

The main problem these days is fill-rate. This is where the GPU-VRAM speeds come into the game. But if you have a V5 and the game still runs, go to 800x600 resolution and voila: 60fps or more in most of the games as oposed to the 15fps in 1280x960.

Get the resolution down a bit, change to 16 bit color depth, and the V5 can stand a chance of running faster than 60 fps, if the CPU is fast enough (2Ghz in this year).

Bus mastering, AGP 4x, are mostly less important than you would think. For example, a test with Unreal Tournament 2003 on Banshee hit a fps of 15. On a S3 Savage4 board (has AGP and Bus Mastering), the result was the same. AGP and Bus mastering give you some more speed, if the board is built for that, and IF THE GPU IS NOT ALREADY RUNNING ON IT'S LIMITS. If it runs at it's geometrical (polygons) or raster speed (textures)maximum, even an AGP 8x can't solve anything...

If you are thinking that "Hyper Treading" will give you any significative plus, you are wrong. HT will give you an advantage if the APP is build for that. And remember, even these days, things like 3DNow! and SSE (even SSE2), are not used in most of the games, and even the fewer that have this optimization it would at most bring you 5fps more. That's all that you can get.

Another surprise: Pentium 4 are worse for games and apps in general than the Pentium 3. Some magazines tested same speed P4 and P3 (1.4 Ghz), and in most of the cases, the P3 (especially in heavy working as in games), rocketed by 5 or more fps in front of the P4. Also the marketing maneuver with the P3 at 1.4 Ghz been priced so high is a proof of the will of excluding such unpleasant reviews from the public, and don't cast a shadow on the P4.

P4 is more than P3, much more a success of the marketing: even if the RIMMS were not good, were huge costly, and bring heating problems and high latency as oposed to DDR, still there were thousands of RIMMS sold. Heavy Advertising maked that possible.

Also, think this way:  if the XP 3200+ is a little behind the P4 at the same frequency (3.2 Ghz), and it succeed in certain cases to overpass it (in games mostly), even if it runs at 2 Ghz in real frequency, who's the useless, and the poorly designed: P4. And also costs a dozen more than XP 3200...

In regards to the hyperthreading -- things are changing -- not only is Intel coming out with the P4 Hyperthreaded -- extreme version for gaming but game developers are talking about implementing code designed for a P4-HT -- so if you got it you can smoke it!

So HT will give you benefits if you have WinXP - and they game developers design the code to make use of that. Time will tell on this one - we'll have to wait and see what happens.

As far as the other stuff you were devling into -  I already know that stuff. Though I do appreciate your wealth of knowledge that you show.  ;)

Still, things are not going to be that fast: you have to remember that the processor is always accesing the same memory, no matter how fast or in how many slices the bus is divided. Even with DDR, a multi-threading process still have to rely on the I/O commands of the Bios.

More, it has been demonstrated that if you access nearly the same thing in RAM, you gain no advantage. Let's say that 2 commands are issued: sound on one thread, video on another. What you get: bottleneck on the audio thread, since it always have to wait for the green signal to proceed to the PCI bus, after the video. And if the sound thread needs resources from memory to generate the necessary mixing, this thread is put on a waitstate, and performance will not increase.

Also, remember all the time, that even if you have Multi-threading, a serious part of the activity still is produced by the kernel itself, which even if it is occupying only one thread, it can still trouble the second thread speed, since the first thing that an app is doing is calling the kernel.

Changing the arhitecture of Windows, and develop a means to avoid using the slow Windows kernel (as in the Windows 3.1 era was done with the GDI to provide speed in graphics), is mostly the true answer to the performance problems, in a much more convincing and certain way than HT.

A well designed and efficient system, wouldn't require HT to go fast, as an Airplane didn't require a full passengers numbers not a half, to be able to have the passengers to feel comfortable.

You bring up some very good points.

To really see the true performance of all the bottlenecks in the core of the operating system as well as the hardware will have to be overcome in order to get what the hardware vendors promise (be it Intel, VIA, Nvidia, ATI, etc.).

Quote:

Changing the arhitecture of Windows, and develop a means to avoid using the slow Windows kernel (as in the Windows 3.1 era was done with the GDI to provide speed in graphics), is mostly the true answer to the performance problems, in a much more convincing and certain way than HT.

Quote:

You bring up some very good points. To really see the true performance of all the bottlenecks in the core of the operating system as well as the hardware will have to be overcome in order to get what the hardware vendors promise (be it Intel, VIA, Nvidia, ATI, etc.).

Wouldn't it be great if microsoft released a windows strictly for gamers, designed with a faster and more efficient kernel that could efficiently handle all these features, without all the networking and advanced GUI and all that crap? It could be a little less "user-friendly", but I would pay good money for that. Heck, didn't they do something along those lines a few years ago? I think it was called "linux"  :P

EDIT: fixed code

Believe it or not, not even Linux can enter the cathegory "Designed for games". Look at the "linux tux" game in Linux  (Red Hat 9) games package, try to play it with a K6-2 450 Mhz and a Voodoo Banshee--> around 5-8 fps, no matter what is displayed.

And in Linux tux, there are nowhere as much textures and geometric details as in a NFS5 game or a Quake3. And these games run much faster and stress a lot more the PC than Linux tux. However, Linux tux runs at 5 fps...

The OS for games has trully not been born until now. A hint on what is mostly close to this way is DOS. Dos is symply the most direct software/hardware approach there ever was, and the most efficient. When you play a game in DOS it uses 8 Mb of Ram for example, in WI\in or Linux it would require at least twice, and the CPU usage is also nearly twice as much...

However noone wants to build Dos games, because "They're obsolete". Why? It's easier to build 2 games on windows rather that 1 game for Dos at the same lenght of time of work.

wrote on 10.10.03 at 09:39:25:

The OS for games has trully not been born until now. A hint on what is mostly close to this way is DOS. Dos is symply the most direct software/hardware approach there ever was, and the most efficient. When you play a game in DOS it uses 8 Mb of Ram for example, in WI\in or Linux it would require at least twice, and the CPU usage is also nearly twice as much... However noone wants to build Dos games, because "They're obsolete". Why? It's easier to build 2 games on windows rather that 1 game for Dos at the same lenght of time of work.

Indeed... That is so very true...

Having said that, You will really be "LUCKY" if you will play Dx9 games on your Voodoo5.

Still it is questionable whenever there was any app or game to show the full maximum speed and quality for a Video card.

Again, it is a problematic issue of whenever any of the newer cards ATI, NVIDIA are trully showing their true potential... The same applies for CPU.

Thinking how good would a Carma game look for example in glide, and how low system you would need (not the Voodoo which was costly) to run it in all it's beauty, and now we have 10 times faster Video Cards and CPU as oposed to that time, but still results are not as amazing as we would expect...

A 2Ghz proccessor, and a 1Gig pixels/minute card should be an amazing power, and make some increddible effects. Unfortunatly this is true only on paper. The CPU gets quickl into waitstates for other devices, the video card is out of sync in transfers, the programs are having their own procedure slowdowns, Memory is slower to access as you get over the 64Mb boundary, you have memory fragementation, left over parts from closed programs in memory, monitoring apps...

With such a picture, it is not surprisingly that you will soon discover that your new "state-of-the-art" system is already slowing down compared to your expectations, after one year...