Notes (18May98): I'll slowly get the site converted to the new format. Until the Reporting System is official, use the Results link above.

The RivaTnT may be out as early as this Summer.

And 3Dfx should release the open beta 2 of OpenGL for Voodoo, Rush and Voodoo2 today.

Notes (15May98): Wohoo! I finished the FSBench Reporting System earlier than I expected. Head on over and test it out. I will delete the files before I open it up for use, so have fun, and let me know if you'd like to see any changes, and if there are any changes that are necessary.

Today it is official, and in six months it will be irrevocable: Texas A&M Univeristy had it's commencement today, so I am now Dr. Bruce Wilson, Ph.D., and I can prove it.  Wohoo!  The reporting system AND a Ph.D. the same day. I'll remove all this pretenciousness tomorrow; I'm just having a little fun!

From VE: OGR has passed along the following bit of info: The 3D patch for Red Baron II will be available in July.

GameSpot has posted a review of 8 new 3D accelerators.

Hold off folks on running any more Benchmarks on FS98.   I figure in 2 weeks we'll be ready for Reports again.

Notes (14May98): Well, don't I feel like a regular doofus.  Right after I say I'm going to make daily updates a regular thing I go and find an exception.  Here's the deal: I want to make the site appeal to users of all flight sims, so I will organize benchmarks for all the most popular flight sims.  Since I don't have all of them (and all the time I'd need to make 6 or 9 benchmarks), I need your help! If you want to put together an easy to use benchmark, I'll be very happy to look it over, and if I think it meets the following requirements, I'll use your Benchmarks and you'll be famous forever!  A good benchmark has very clear instructions.  It will remove as much randomness as possible, which means you pretty much have to define all the visual and some of the sound settings.  It needs to be fairly quick to run, and be made so it's easy for the user to get a single figure to report (Actually, what I'm after is three reported numbers, one for normal spot-plane flight, one for cockpit approach to a large airport, and one with the lowest frame-rate you can find; these numbers can be averaged for reporting, but it may be best to have all three numbers reported as well).  I'm aiming to have all benchmarks run at 1024x768 when possible, but the more resolutions the better.  For sims like FS98 and FUII, getting a consistent frame rate will be tough, and it might involve having the user pause the scene and recording the frame rate that way.  Ah, some of you noticed that I said FS98!   Yep, I'm redoing that, too.  It seems that there are some flaws in the current benchmarks (mostly a displayed FPS that jumps all over the place, and situations with frame rates that change with time.  If you want to try the FS98 benchmarks, try to model them on the current Benchmarks, and fix the problems.  So go to it, make a benchmark and submit it.  I'll review them all, get with the authors of the best ones to fine tune them.  I plan to have all this done in two weeks, so let's put a one week deadline on these, okay?

Now, the exception to daily posting is because I want to automate the Benchmark reporting process.  I'm going to add some CGI scripts to the site to make reporting and searching the reports much easier for both you and me.  The plan if for me to spend as much time as I can starting tomorrow modifying a flat database manager system to suit our purposes.  This will include all the important parameters for the Benchmarks, and will have a limit of reporting three benchmark numbers and an average.  It will be a bit of work for me, so I don't plan on spending much time doing page updates.  There are some problems I have to overcome (mostly the lack of flock() in Perl for Win95, and wondering how fast it will be when it gets a thousand entries), but I hope to work them out.  And the more perceptive of you will already know, this site will get a re-design when The Big Switch happens.

Those of you who have volentered your time, in addition to making some Benchmarks, what I want help doing is to scour this site and the 3D forum for advice, news, and requests for help (and answers, of course) into the Repository.  I'm heading over there now to post some instructions in the Welcome topic.  Thanks everyone!  I'm grateful you want to help out!

The Banshee chip is being sampled now, so we may hear something by next week about its pekrformance.   Should sell at $150 for the 8 MB version, with higher-than-Voodoo2 performance (awaits to be seen).

More of The Plan: I forgot a couple things last night.  First, because cards are getting better, I will now ask for the Benchmarks to be done at 800x600 AND at 1024x768.  Yes, I know it's more work for you, but I'm afraid it is essential to evaluate these cards in the mode you will likely use them.  And second, as some of you have assumed, I'm dropping the "98" form the tilte of this site; I will begin assembling Benchmarks for most fo the civil flight sims available, and some of the best military ones.  I haven't figured out how to display all the results, but I suspect it will have to do with CGI forms and databases.

AOpen plans to release a 16 MB version of the Voodoo2, with THREE(!) texel and one Pixel processors. Unreal (first person shooter) may be the only game that can use the third one, though. What will they think of next!

Ritualistic has posted a review of the Canopus Pure3D II.

Ripped from Op3Dfx: VE sent a little file our way which adds a property page in your disply control panel for overclocking your Voodoo2. You can get the 90k file here..

The Voodoo2 Overclock Property Page is a FREEWARE product that allows you to adjust the Graphics Clock setting of both Glide and Direct3D for any Voodoo2 based Graphics Board. It does this by adding a Property Page to the Windows 95 Display Properties.

I'd quote this next one, but I'd feel a bit like a Matrox pimp: the G200 site has a nice list of the 3D tech terms used to advertize the new G100/G200 boards.  I'll probably yank it down to the Repository, though ;-) It looks like Matrox plans to release these boards as AGP cards only, buy will make PCI versions if they sence a demand.

Notes (13May): Planet Quake has an article on 3D Gaming with Linux if you are tired of Win95/98.

TeamApache has news that a playable demo of Team Apache is available at Op3Dfx here. From Op3Dfx:

A review of Plane Crazy, an arcade flight sim obstacle course, over at hyperactive.  Wear your sunglasses when you read this review.

Ka-Ching!  It's another 3D accelerator! Ripped from Billy Wilson (no relation) at VoodooExtreme, in toto:

Speaking of Next-Gen, they have also posted a bit of info. on Rendition's Conspiracy project.  Here's most of this morsel:

While it has been known that Fujitsu was working on a 3D chipset for quite some time, few people actually knew what form it would take and how it would be positioned. Next Generation Online has learned that Rendition's Conspiracy Project is indeed the very technology that Fujitsu has been working on.

As previously reported, the chipset is said to be geared towards offering better performance for lower end CPUs or CPUs that don't offer stellar floating point performance. As one would expect, the emphasis of such hardware will be in the realm of geometry setup, thereby removing (or at least reducing) the need of the CPU to be particularly fast or good at flops.

As such, Cyrix has been fingered by one source as another company closely involved with the project given the 686's marked deficiency in said areas.

And because my middle name is convenience, just in case you missed Fujitsu's press release on the Pinolite, here's a high in fiber link (although the press release is a bit old now, I believe all of the info. on Fujitsu's Geometry processor is the same) -- all of the specs and stuff.

Hasten to NextGen, for they have updated their most excellent 3D accelerator FAQ.

The official web site for Voodoo Magazine, if you are a 3Dfx fan.

PII/400 not fast enough for ya?  Look here.

The Plan: Okay, here are some of my thought about this site.  I really like flight sims, and I think 3D is totally cool.  Flight sims have been the perennial best-selling game category, and now that all but one support 3D (and the odd one out will in the next version), it seems to me that we simmers must be buying a good bunch of 3D cards.  So my long-term goal is to find out if we simmers are a market force that 3D makers will want to notice, and if so, make the flight sim community a market force the 3D makers need to notice.  

Notes (10May98): From Psychonews, via Op3Dfx:   MSK entertainment, a development company they introduced the other day, in regards to their 3D futuristic space flight simulator, Planet Of War, have decided to shorten the path to 3D graphics for programmers by putting some 3D-programming tutors online :

From VoodooExtreme: There are some new sweet new shots of Incoming running on Matrox's new baby, the G200 over on Fast Graphics.   Here's a fat-free link to the goods:

Notes (8May98): Techweb has an article about 3D taking Center Stage at the Game Developers Convention.

I just got out of a meeting with nVidia. They aren't showing the RivaTNT on the floor, so instead I setup an appointment and viewed it in their suite. There's a reason for this; the silicon they're showing is the first rev, and only about 6 days old now. They've barely had time to get basic VGA, Win95 and NT support up. Not exactly the kind of thing you want to show game developers for their first impression. MORE HERE

Notes (6May98): http://fsbench.bw.ml.org is now the primary FS98 Benchmark site, updated often.  The site at AVSIM is updated weekly or whenever.

Tom's Hardware: a Graphics Chip Review with side by side comparisons
Editorial on Multi-Monitor Support.
Cool Tech site: AnandTech.com
For you multi-carders: DirectControl version 1.0.34.  CombatSim has part 1 of their 3d hardware survey.
DirectX6.0beta.
Texture memory: Check it out

Notes (4May98): It finally happened: S3 had announced another 3D accelerator, one it hopes will be competitive with the next crop of 3D chips.  The Savage3D is AGP, 32-bit color, 125 MPixels/sec, one-pass trilinear filtering, texture compression, specular and diffuse shading, alpha blending, multiple textures, palletized textures, edge anti-aliasing, vertex and table fog and 16/24 bit Z-buffering, bump mapping, anisotrophic filtering, reflection mapping, shadows, texture morphing and procedural textures. The regular features for the next generation. By the way, the K6-3D is now known as the K6-2 which supports the 3DNow! FP instruction extensions.

Notes (3May98): Looks like the G200 will come out in 2 flavors, the Matrox Millenium G200 and the Matrox Mystique G200.  The Millenium has a 250 MHz RAMDAC vs. 230 MHz, and the Mystique has built-in TV out.  What's the functional difference between them?  None.   None...functional difference.  I suspect the Mystique will have a better set of pack-in games.  By the way, get the 16 MB version of the card, over the 8 MB version.  It's more expensive, but you'll need the extra memory when we start running 3D in 32-bit color mode.

Yesterday I started reading my first novel in 4 years. I take it as a sign that I'm finally unwinding from doing a Ph.D.

I guess it's time to work up the page graphics for the i740, G100/G200, Riva TnT, PVRSG, MPact, and whatever else I can think of that's come out recently.  Maybe I'll fiddle with the graphics a bit.  Simplified.  Check out the Home page for info on the best way to display this page.

Check out the fly-in pages I made.  Pretty cool, eh?

Notes (27Apr98): There is a cool article on Bootnet on 3D visual quality. And this Comparison at Tom's Hardware is required reading.

Notes (26Apr98): Here is a list of the new 3D chipsets coming out this year:

nVidia Riva128ZX. An update of the Riva128, with 30% faster 2D acceleration and 8 MB frame buffer. 3D performance same as the Riva128.

Matrox MGA-G100 and MGA-G200. These 2D/3D chipsets will be slightly slower than the Voodoo2 in 3D (the G100 will be slower and cheaper than the G200), but they will have a better image. They render 32-bit color in 3D mode, and will have what we have come to expect of Matrox: excellent 2D performance. Due out in April for $100 to $200 ESP, with 8 to 16 MB memory. With 32-bit rendering, it will need a lot of 2D-card RAM for the frame buffer, 8 MB minimum, probably 16 MB to run comfortably.

Notes (19Apr98): Between finalizing my Dissertation, writing a bunch of web pages for my virtual airline hub fly-in, and writing a big article for a well-known flight sim magazine about 3D hardware, I just haven't had time to update before this. But here it finally is! I still don't know what the deal is with the Voodoo2 not performing as well in Direct3D as it does in Glide and OpenGL. Since I have orderd a board (from BuyComp, still waiting for it) I keep hoping that the next driver set will bring FS98 performance up. I keep waiting. There is a lot of discussion on the AVSIM 3D Video forum. If you have any question about 3D hardware, head on over there and ask. I'm the moderator, and I've made it clear that it is a forum for both the newbies who haven't discovered Alt-Enter, and the seasoned 3D-pro trying to get another 0.5 fps out of his setup.

I'll be moving from Texas to Utah next month, so if you plan to run the Benchmarks, do it in the next week or two. There will be a bit of a dry spell during the move. My email address won't change, though. The @Earthling.net address is an email forwarding address provided by iName, so it's mine for life.

Notes (28Mar98): Sorry about the long delay since the last update, but I really do have a good excuse. Yesterday morning I passed the last requirement for a Ph.D. in Chemistry! So now that's done with I'll have a little more time. Thanks everyone who have sent in their good wishes. It's been a great 24 hours!

In other news, the Voodoo2 wasn't too improved by the latest driver set, but the Riva128 was improved by the lastest drivers. For those of you still looking for a card, two makers have released info on cards coming out this summer: nVidia and NEC/VideoLogic. Go to their websites for more info on these little blazers and see if it's worth it to you to wait for them rather than enjoy 3D now. I'm still waiting for my cheap Voodoo2 boards to arrive from BuyComp.

Notes (8Mar98): Wanna see what trilinear filtering is? Look at the article at 3D Generation

Notes (4Mar98): A small request for those who have either a Voodoo2 or a Verite2200:

If you wouldn't mind running the FS98 Benchmarks at 800x600 and at 1024x768 to test the preliminary observation that these cards have frame-rates that do not depend on resolution, like everything else does.  If you have a Riva128 card, could you do the same thing?

Here's what I'm trying to find out:  If FS98 is running totally CPU-bound then changing the resolution should have no effect on frame rate.  If FS98 is somewhat fill-rate-limited then 800x600 will be somewhat faster than
1024x768.  It seems that the V2200 and Voodoo2 run totally CPU-bound (the frame rate does not differ between 800 and 1024), but Riva and Voodoo have some fill-rate dependence (800 is faster than 1024).  Now if you look at the benchmarks for the Riva, you'll see that the AGP configuration is clearly faster than the V2200 or Voodoo2.  So what gives?  If Riva is running fill-rate limited, and AGP removes that limitation, then the frame rates seen are the maximum frame rates the CPU can crank out.  But tests show that the Voodoo2 and V2200 are running totally CPU-bound, and they are slower than the Riva.

Notes (3Mar98): More Voodoo2 entries.   The entry in the P233MMX slot is for an overclocked CPU at 291 MHz (3.5x83MHz), so that's why the number should be 25% bigger than the P233MMX Riva and Voodoo entries. Actually, it's 40% bigger, a 15% bonus!

Notes (1Mar98): There is a lovely new trend in 3D hardware showing up these days: resolution-independent frame rates. With the Voodoo2 and V2x00 chipsets, which each do hardware triangle setup, the observed frame rate does not matter if you are 640x480 or 1024x768; it will always be about the same. If you think about it it's pretty obvious. If you are sending the card only the coordinates for the polygons, it doesn't matter if you scale them for 640 pixels or 1024 pixels, the data stream is the same. For other chipsets the CPU has to calculate where all the polygons are first then send them all to the 3D card, consuming CPU time and data transfer time. The long and short of this is that if you have either a Rendition V2x00 or 3Dfx Voodoo2, you can run 1024x768 just as fast as you can in 800x600 or 640x480. Very cool!
The Permedia2 column got moved to the left. It's looking better for FS98. It's still not a recommended card for other 3D games, though, like Quake2. Funny thing is, the PM2 is the only card good for editing Quake2 levels.
If you want to send in frame rates for Flight Unlimited 2 and Longbow 2, I'll be happy to throw them into a page. I don't plan on making it a big deal like I did with FS98, though.  Someone else can if they want to and integrate the tabulations into this site.  Let me know if you're interested.

Notes (25Feb98): I've begun reorganizing the columns of the Summary to reflect (somewhat) the speed of each chipset in FS98, fastest to the left, slower to the right.  I left the Matrox and 2D columns on the far right because they do not do full 3D accelleration.  If you have a V2200 or Voodoo2 board we'd love to have some more Reports. 

News (25Feb98): First Voodoo2 result is in, on a P2/300 at 1024x768x16, ave = 32.7.  800x600x16 results are about the same.  This seems strange to me, because the Voodoo2 does everything faster than the Riva (triangle rate, pixel fill rate, etc.) so why are the Riva numbers higher?   I suspect the new Voodoo2 drivers aren't tuned and tweaked yet, and that the Voodoo2 drivers are not using the hardware triangle setup.  At least I hope that's the cause, because if it isn't there is something important about 3D I don't understand.   Another thought, Riva is faster only in AGP cards--which might indicate a data-transfer-bound situation.  Unlikely.  FS98 runs CPU-bound, I've always thought.  Any way to test this?

News (22Feb98): "...And on behalf of bozo's everywhere, I proudly accept the Dunderhead of the Year Award."  This is what I had to say when I played FSBench3.vid after upgrading my computer, and discovered that the complaints about a loud tone after the OM were my fault, not yours.  I guess Patch 1 changed something, so whereas it didn't give the gear-not-down warning tone before, now it does.  So I made a new FSBench3.stn file (starting in the air this time) and put it in the new distribution file, FSBenchE.zip.  Would a couple of you re-run FSBench3 and check that your numbers haven't changed too much.  I noticed my Voodoo Rush numbers improved (by about 50%) when there wasn't a constant sound sucking up the data bus.  Other chipsets shouldn't be as affected.  Please let me know if your report changes.  Gad, what a cock-up if they all change!  Sorry about this, folks.

News (21Feb98): The Voodoo2 is shipping!   Creative Labs shipped the 12 MB 3D Blaster Voodoo2 on the 20th, and it should be in your local shop as early as Tuesday (24Feb).  But in reality we may have to wait a couple weeks, because the Quake guys had the first week's production preordered long ago.   The cheapest prices so far: www.necx.com ($260), Firstsource ($260), www.computability.com ($275), www.insight.com ($290).

NEC is announcing the PowerVR3.  See the writeup at Next-Gen.  It's supposed to be cheap, scalable to PC, console and arcade gaming, 2D/3D, do 120 Mpixels/s (Voodoo does ~45 MP/s), 1 Mtriangles/s (Voodoo does ~0.4), hardware texture compression (so 10-40 MB textures will fit on the card), and 3D Winbench scores above 1000 (current cards max at 300).  Well, I'll believe it when I see it.  The PowerVR2 was supposed to be faster than the Voodoo.  Look below for evidence of that.   Not there?  The VR2 pretty much fell short of every claim made. Will the VR3 fare better?

Speaking of the 3D Winbench scores, Tom (www.tomshardware.com) has a nice little writeup of something we've seen for a while: the trouble with synthetic benchmarks.  That's one of the biggest reasons I strated this page, to stop people using Winbench and Quake scores to assess which 3D accellerator to buy for FS98.

The Intel740 is also being reviewed (Tom takes a look at it) and the consensus so far: great if you run benchmarks, average if you play games.  Time, again, will tell.  I'll bet it will have great drivers, though.

Sorry this update took so long to get out.   I've been writing a paper (to be published in Biochemistry), and I defend my dissertation in a month, so updates will continure to be sparse.  I will try to get this news section updated more often, though.

I've looked at the Summary tabe both ways, and I think I'll keep all the entries.  I apologize to those who are using 640x480 who need to see both sides of the table, but some people wanted to see the Matrox and 2D data left in the table.  If I had the Virge and Rage2 data still in the table I'd include it too.  Columns to be added soon: i740, MPact2, Oak5 (if they are released), the Trident 3D chip (if anyone buys one).

Request for info: anyone have FS98 running on NT using the Direct3D yet?  Any interest in Benchmarking other 3D flight-sims?   I noticed there is an SPP frame rate page at Rev. Karin Conover-Lewis's Home Page.