Radeon8500 "high poly count" bug?

It will be interesting to what extent this upcoming update affects real apps. Also, I'd expect Carmack to update his .plan if this ATI has actually created a workaround for the problem encountered by the DOOM3 engine.

Entropy
 
On 2002-02-24 21:05, tb wrote:
Sorry, but I will not risk my releations with ati, because of leaking drivers. I hope ati will release them next week in their beta area.

Thomas


Heheh, No prob Thomas just a little humour, I'd never ask anyone to break a NDA. :smile:
 
Yes, I would never want anyone to break NDA or get into any sort of trouble. I'm under several NDA's myself. Shame that it's nothing related to the pc-entertainment industry.

--|BRiT|
 
What's the situation like now?

Has problem with D3 been solved? Are the drivers which solved the current problems (with Serious Sam etc) real? Most importantly, is the 8500LE really somewhere in between the Ti200 & Ti500 on average as many have claimed now that this issue has been fixed? I originally though so but in some reviews I saw the 8500LE performing worse on average then the Ti200.

For those who remember, I was considering a graphics card for my comp. I recently found a Leadtek TDH GeForce3-Ti200 64MB Ti200 for around $447 and a Gigabyte AP64D Radeon 8500LE (at the standard LE 250/250 but with only 64mb RAM) for $495 so I was seriously considering getting the 8500LE instead of the MX440 but after I saw the disappointing 8500LE results I wasn't so sure..... I may get the Ti200 instead or just stick with a MX440 (for $290) or maybe even a Gigabyte AV64S-H Radeon 7500 Pro 64MB for $235.

So is the Radeon 8500LE really that great with the new drivers or is it still at or below Ti200 level?

Oh and once again please don't tell me NZ$50 is only US$20 and that's the price of a movie with the wife (according to spam anyway) so just get the... Or NZ$120 is the price of a game so just get the...... Money issues aren't so simple I'm afraid.......

My old post http://216.12.218.25/domain/www.beyond3d.com/forum/viewtopic.php?t=266&highlight=%2A8500le%2A
 
Well the lastest drivers do fix some OpenGL issues. i.e. the GLExcess benchmark shows a massive improvement, i.e. where it should be, not where it was (outperfromed by much older cards). I dont know whether Tribes 2/MOHAA/RTCW show improved performance as I dont play those games.

Some of the latest reviews with more up to date drivers show the 8500 well ahead of the Ti200, even in such nVidia friendly benches as Giants.

The Ti200 would be a 'safer' bet, but I'm happy with my 8500LE. You get other benefits with the LE - faster anisotropic filtering (though not as good as a Geforce3/4), better AA, but with a bigger FPS hit. Dual-head output and better DVD playback.

Get the Gf3Ti200 or 8500LE over the Gf4MX 4400 anyday.
 
Faster than a Ti200 ??

http://www.ocworkbench.com/2002/gigabyte/radeon8500deluxe/radeon8500p0.htm


Test System:

Intel Pentium 4 1.5G Engineering Sample
ECS L4S5A SiS645 mainboard running 100/166 async mode
IBM 15G 7200RPM ATA100 HD
AOPEN 40X CDROM
MSI Geforce4 Ti5700 Graphics card w/Detonator driver 27.20
ATi 8500 w/4.13.9017
Nanya DDR333 256MB DDR
Channelwell 375W AMD Approved PSU

Serious Sam:SE
se.gif


Picture_308_copy.jpg
 
The card looks super nifty with that red PCB and silver heatsink. If I can say one good thing about that review, it's that they take some nice looking pics.
 
I said AP64D 8500le

Note: I said AP64D 8500le (250/250) not AP64D-H 8500(275/275)...... It's significantly more expensive for me compared to the Ti200 so this isn't just a sinple Ti200 vs. 8500le either. The 8500le has to be sufficiently better to warrant the increased price.

Since I'm only going to have a oldish 15inch monitor and ATI's display quality legacy is based on them making their own cards, I don't know how significant this factor is either.

I don't have a DVD drive and even if I did I probably would have many DVDs. More significantly, I'm unlikely to ever want to do other stuff with my comp (a Tbird 1700+ probably oveclocked to 1900+) while watching DVDs so the better MPEG2 hardware decoding features aren't so significant since a software decoder should be sufficient. Dual Head is interesting but not really important either. Also, I'm not to sure if the Gigabyte card ships with a DVI to analog monitor plug so I will only be able to use DualHead with tv out and my analog monitor (unless I get a DVI monitor of coz). The LeadTek Ti200 also has a TVout and DVI (altho the DVI isn't so good) and it has just as good a heatsink IMHO. And personally I prefer the old Green PCB (in any case, I will never see it so it's irrelevant).

Truform seems interesting but not likely to add much value IMHO. GeForce3 AA quality isn't as good however with ani filtering it can be just as good and it's still faster (or at least in the reviews I've seen) so honestly, I consider the GeForce3 to be a better AA card. And Nvidia has a lot better driver+developer support then ATI

So really, it falls down to the performance of the 8500le vs. the Ti200.

P.S. If you're gonna start talking overclocking, you need to consider overclocking all cards, the Ti200, the 8500le, the MX440 etc etc etc. And I have no doubt that the Ti200 & 8500le are a lot better then MX440 but they are a lot more expensive too.
 
8500 vs. 8500le

How do these cards tend to compare? They have 10% diff memory and GPU speed so I would guess maybe 15% performance diff?
 
hmm maybe quincunx (or4x)+ 8x aniso on a Ti200 is faster than 4x + 8x aniso on Radeon (but in most recent games neither is fast enough). But thats due to SS v MS implementations of AA. I would have thought that 2xAA plus highest aniso on a Radeon is faster than 2x + aniso on a Ti200 due to the bigger aniso hit the Gf takes.

Remember there are differences with both AA+aniso implementations on both cards, which means say the performance of 2xAA + 8xaniso on a Radeon is not at the same 'overall' IQ level as 2xAA + 8xaniso on a Ti200, the Gf would need a higher level of aniso to compensate for MS lack of texture filtering, which is even more true for quincunx as it blurs a bit.

Gf AA is faster, not much at 2x but quite a bit at Quincun & 4x, but doesnt touch textures.
Radeon Ansio is faster, but doesnt do angled surfaces well (? at all?).

In general I would surmise with a good CPU at 1024x768, 2xQuality AA and highest aniso, the Radeon (even the LE) gives better IQ and is faster most of the time.

Again someone here could probably correct that for me, or expand that.
 
It's a well known fact that the R200 was never actually completed. ATI had to compete against Nvidia so they quickly tapered off developement even tho it wasn't really ready. Of coz, they removed the parts which didn't get completed (i.e. weren't working) to save on die space (this also helped them to cut down on production costs).

They're smart tho and whenever their talking about die size & features they talk about the R200 (the never complete GPU). The current GPU is actually the r200 or the Radeon 8500.

The trouble with this is it means the speed varies greatly and it's missing a few important parts (so it doesn't support a number of things in hardware). The poly issue JC encountered in one of many the Rad 8500 have.

Realising benchmarks are everything, they designed their drivers to dump about 25% of the less important info. Then when people began noticing quality of graphics of the Rad 8500 lost out to everything including old TNTs and Savage2000s, they silently fixed the 'bug' claiming it was a problem in their Z removal causing it to remove important stuff/stuff it shouldn't remove (which is true since they implemented this "bug" in the Z removal part).

Of coz, the Rad 8500 which used to be able to compete with the Ti500 now loses to a G4 MX440 and a G2 Ti200 (it's on par with these cards on average but it varies greatly depending on the game, features used etc) but everyone looks at old reviews anyway so their happy.

And they've managed to implement many features (or improve them) in software so they've solved many of the compability probs and other issues (the notorious fog issue and the many color and texture problems), at the expense of speed, related to the missing features.

Still the poly bug is proving hard to nail. People think the Rad 8500 is a good GPU let down by poor drivers. It's actually a poor/incomplete GPU greatly improved by ingenious drivers.

One good thing did come out of this tho. Many game devs are interested in their fixed 'bug' since it is very good at deciding on the importance info. It will be very good for GPU scaling in a game.

Also, ATI driver devs have managed to do many things in software and in general learn a hell of a lot in their efforts to ensure the Rad 8500 works as well as possible with everything.

So anyone who is even considering the Rad 8500 hasn't done research....
 
People wonder why graphics cards discussions turn into flame-fests. I dont think its the importance some normally sane people attribute to one card over another. It's sheer amazement and frustartion when you try to comabt crazy rumours & stupidity (block-headed fanboyism) expressed as fact and given out as a insider 'info' to help you 'make the right decision' that churns around the net.
 
Actually MouseAnony speaks the truth, it was a known fact that the R200 was designed in Dave Ortons back yard of at a barbecue party (b.y.o.b). The ATI engineers drew out the chip design on a hamburger wrapper (bad mistake) and there was a small ketchup stain (hence the High Poly Bug) that caused some issues. Rumors around ATI state that hamburger wrappers WILL not be used for the next chip design but they will move to a clean paper plate. :rolleyes:
 
Back
Top