GeforceFX=surprisingly slow

ATI's drivers for gaming are OK... their drivers for application 3D are pathetic...and their drivers for their multimedia cards are pathetic...I wont buy one until they start to really make improvements.... I did just buy a wildcat 4110 for 150 off ebay though ahh the joys of a market economy.

[KrasH]

I'm sure that the GeForce FX will get a little bit faster with driver tweeks (even my ol' GeForce Ti 4400 got a 5%-10% speed increase with the latest detonators)

But currently I still think that ATI are in front in terms power + value, although I haven't seen how much a GeForce FX costs. Maybe when CPU's actually get faster, then we'll see what the true potential of each card is, because I feel that even a 3.06Ghz P4 is the bottle neck of the ATI Radeon 9700 pro and possibly even the GeForce FX.

I remember some time last year that a graphics card company, I think Matrox, cheated with their card drivers. They made their drivers lower their graphics detail in Quake III so that it looked like the card was fast. Was a marketing ploy, because Quake III is used as a benchmark for alot of Magazines that review cards.

[Crumb]

I remember some time last year that a graphics card company, I think Matrox, cheated with their card drivers. They made their drivers lower their graphics detail in Quake III so that it looked like the card was fast. Was a marketing ploy, because Quake III is used as a benchmark for alot of Magazines that review cards.

ummm that was ATI with their first radeons

But now with the 9700 and its new core they have done a good job and nVidia will have to work hard to beat them :-)

have you listened both cards?

GeForce FX Booting
GeForce FX running 3dMark

and now the 9700...
Radeon 9700 Booting

ummm now i've found something better in my Voodoo3, it doesn't have a fan so it's completely quiet.

[Rogue]

mips_proc wrote:
and with Cg its features will be easily implimented... its raw polygon push is much more appealing from what I've read... and I'm sure its drivers will be much better.

With DX9's HLSL, Cg will become pretty obsolete, more so since its support for backends is limited - only two cards I know of implement ARB_fragment_program at all.

Between OGL 2.0 HLSL and the DX9 one, I don't expect Cg to be around for long, more so since Microsoft will really push for their version.

If you judge a GPU by its FPS and not by its features....then your missing the point entirely... we're enter a time where FPS matter less and features matter more... and stability/drivers/etc matter more... the race for FPS was over at the Gforce4 ...now its time to see some features and higher poly counts at the same FPS...

That is, unless you recognize that the featureset of the FX is only unique for now. The others will follow. And, you don't care about that feature set unless it is really used by games, which is absolutely not the case for now.

To me personally... ATI will never be in the lead until their able to get decent windows drivers...

I am sure glad that right now there isn't really any leader at all. It keeps changing back and forth, and new players are also appearing. Right now, quality graphics cards cover a much broader range of price segments, starting from cheap cards like SiS Xabre and Radeon 9000 up to the expensive beasts like the 9700 PRO and GeForceFX. However, this means that games will try to appeal to all of them, and that more or less settles the feature set to the smallest common set.

Fortunately, HLSL's (regardless of which language) are starting to really kick off. I am curious to see what a P-10 can do.

[Herewegoagain]

I see you don't have a clue about SIS's Xabre and NV18. Both of these products are 8X AGP capable.

These were cross posts without a refresh, but I'm sure you knew this.  

That is not the issue since those cited motherboards worked with other 8X AGP capable cards.

Ok, let's take a look and see who is at fault, ATI or the reviewers of these cards (note the Xaber update in the Gigabyte bios).  These are the bios versions that were available at the time you claim these cards were reviewed.

[color=CC0000]MSI 648 MAX
BIOS Type  AMI® BIOS  File Size  491KB  
Version  1.1  Update date  2002-9-27  
Update Description -Fixed system unstable when using ATI 9700
-Fixed error message sometimes appear after S3 resume
-Fixed S3 resume failed when USB legacy set as enabled  
Special note  ,  
Download  6585v11.exe [/color]

** Soltek I could only locate the SL-85ERV2 on their site, and they list no bios updates that I could see.


[color=CC0000]Asus P4S8X BIOS 1003A
1003A  2002/08/27
Serial ATA ROM updated 376.bin(016).
Patch ATI AGP 8X.
Support Intel Pentium 4 2.8Ghz CPU.
[/color]

[color=0000CC]Gigabyte GA-7vAXP last 3 bios updates
7vaxp_f7.zip F7
(Oct. 25, 2002) 1. Disable CPU fast command when FSB166Mhz CPU plugged.  
2. Modify top performance setting frequency.
3. Fixed SIS Xabre AGP card STR fail on win98\ME.  
 
7vaxp_f6.zip F6
(Oct. 07, 2002) 1. Fixed system hang under win98 when Creative PCI sound card exist
2. Improve CPU to AGP performance for ATI AGP card.
 
7vaxp_f5.zip F5
(Sep. 05, 2002) 1. Mass production release.
[/color]

[color=CC0000] Via P4PB400
Version (Date) 1.13 (01/03/2003)
File Name (Size) P4PB4113.bin  (257KB)    
Update Description Support high capacity HDD

Version (Date) v1.10 (10/16/2002)
File Name (Size) P4PB4110.bin  (257KB)    
Update Description add LAN Boot and Enhance other functions

Version (Date) 1.05 (08/29/2002 (08/29/2002)
File Name (Size) P4PB4105.bin  (257KB)    
Update Description Enhance Some bootable device

Version (Date) 1.00 (07/22/2002)
File Name (Size) P4PB4100.bin  (257KB)    
Update Description First Release [/color]

So 3 of the 5 boards you listed already had bios updates several months before this site tried to review the card.  Soltek doesn't list ANY type of bios updates, and Via is extremely vague about what the fixed in the update.   Anyone with half a clue about PC products knows to check for bios updates and chipset updates when an entirely new product like this hits before they try to use it.

[Minion]

Hammer wrote:
As with NV25 and R300 examples, the driver support has to mature. Refer to serious OpenGL driver battle royal for the indications of NV30's potential. At the moment, official release drivers doesn’t even support NV30+ family (only in leaked beta form).

I know, but thats not the point.  This is the best Nvidia can come up with.  The processor has more transistors than the R300, yet it's slower clock for clock than the R300.  I suspect this will be the same for DX9+ games aswell, driver optimisations or not.

Anyway, nVidia has branched to chipset and integrated audio markets btw…

Likewise with “Pentium 4” product (a core change), it takes time mature.

PS; I recall, NVidia hasn't made the shift to pure 256bit bus unlike ATI, Matrox, 3DLabs.  Geforce FX is one crippled GPU, even with 1Ghz DDR technology (due to DDR overheads).

I dont think that the 256 bit bus on the R300 has as much latency as the DDR-II bus on the NV30, but the difference would probably only make less than a 5% performance difference.

[Minion]

Ilwrath wrote:
I'd hardly call the GeForceFX pre-release boards slow...  It does trample the current fastest graphics card on most tests, and comes very close on the few it loses...  
 

How does 30% increase represent a trampling when the GFX is ahead, but if the ATI card is 30% ahead, its very close?

[Minion]

mips_proc wrote:
The GeforceFX may be slower in terms of FPS on 'current' games... but its feature packed...and with Cg its features will be easily implimented... its raw polygon push is much more appealing from what I've read... and I'm sure its drivers will be much better.

Its lock will go up 300mhz from what I've read....

The GPU ran at 500MHz in the review.
This FPS talk represents an about face!  In a previous thread about upgrading, you were all for 130+ FPS!  Make your mind up.  (or did you agree with my points there?)

[Minion]

Hammer wrote:

Ilwrath wrote:

 Personally, I'm a bit worried that through all their other 'market segments' such as mediocre chipsets

NForce 2 is not a mediocre chipset relative to VIA "crap sets". It maybe true with the first nForce 1 release but not true on the second release. This pattern is similar to the original release of Geforce 256. It took the second release(e.g. Geforce 2 series)  to make this a real success(relative to 3DFX).

Deffo agree with that  - those Nforce2 chipsets are the tits!  Most of the computers in my house use Via "crapsets", and the Nforce2 is worth about another 200+ rating on them.

[Hammer]

 (note the Xaber update in the Gigabyte bios).
 

That's nothing to do with AGP 8X. "STR" has something to do with "Suspend-to-RAM" feature.

It doesn’t automatically destabilize the card IF one doesn’t use this feature. One could use "Suspend-to-Disk" instead.

Asus P4S8X BIOS 1003A
 

ASUS’s updates may have two update values for example.

Via ASUS auto update feature
1001E has update value of "01/13/2003" (For A7N8X)

But from the ASUS web site;
1001E has update value of  "2002/12/23"(For A7N8X)
1001G has update value of  "2002/12/24 "(For A7N8X)

BIOS 1002 has update value of 2003/01/30, but it doesn’t exist (broken link perhaps) in the FTP sever at this time (AUS 31/01/2003).

2. Improve CPU to AGP performance for ATI AGP card.
 

A performance tweak. This is a common practice.

MSI 648 MAX
 

Update on website and FTP site(or via MSI's auto update feature) may not be in sync btw. Note that I do own a late model MSI VIA KT class board.

[Hammer]

I know, but thats not the point. This is the best Nvidia can come up with

I recall, the official launch date was somewhere in late February.

The processor has more transistors than the R300

NV30 has support for 128bit floating-point color feature, which R300 doesn’t not support.  A feature may require more transistors.

I dont think that the 256 bit bus on the R300 has as much latency as the DDR-II bus on the NV30

Well, NV’s DDR-II solution only delivers 16Gb/s, while ATI’s 256bit solution delivers 19Gb/s.

Go figure that out....

[Hammer]

Deffo agree with that - those Nforce2 chipsets are the tits! Most of the computers in my house use Via "crapsets", and the Nforce2 is worth about another 200+ rating on them.

Where?

[Minion]

Hammer wrote:

I know, but thats not the point. This is the best Nvidia can come up with

I recall, the official launch date was somewhere in late February.

20 odd days aint goona get a huge performance increase in all likelyhood, is it?  (and yes I know its been done before)

The processor has more transistors than the R300

NV30 has support for 128bit floating-point color feature, which R300 doesn’t not support.  A feature may require more transistors.

The R300 does support 128 bit FP colour, as FP colour is a prerequisite for DX9.  Check Toms hardware and you'll see the 128 bit FP support in the R300

I dont think that the 256 bit bus on the R300 has as much latency as the DDR-II bus on the NV30

Well, NV’s DDR-II solution only delivers 16Gb/s, while ATI’s 256bit solution delivers 19Gb/s.

Go figure that out....

I was pointing out that the Nvidia solution was SLOWER.  i.e. less throughput and more latency
Go figure THAT out........

Where

I was agreeing with you that Nforce2 is good - vastly better than VIA's attempts at a chipset mainly due to the fact that Nvidia can design a memory interface.
No need to assume that because I think that the Geforce FX is sh*te that I think everything Nvidia is crap.
I have a Geforce 2 ultra in one of my computers and  Ti in another.  I dont give a rats ass who makes my hardware as long as its decent.

[Hammer]

Deffo agree with that

Is that relevant?

 - those Nforce2 chipsets are the tits!

Is that relevant?

Most of the computers in my house use Via "crapsets",

Is that relevant?

I do own MSI built VIA KT class chipset and have access to ASUS built to VIA KT class chipset test machines.

 and the Nforce2 is worth about another 200+ rating on them.

What do you  mean by "Nforce2 is worth about another 200+ rating on them"?

[Hammer]

 Check Toms hardware and you'll see the 128 bit FP support in the R300

Sorry, I should be referring to "Max pixel shader precision".

Refer to
http://tech-report.com/etc/2002q3/nextgen-gpus/index.x?pg=5

I was pointing out that the Nvidia solution was SLOWER. i.e. less throughput and more latency
Go figure THAT out........

Refer to my "Posted : 2003/1/29 11:45" for a similar statements.

[Hammer]

20 odd days aint goona get a huge performance increase in all likelyhood, is it? (and yes I know its been done before)

The drivers v42.70 was still in beta form...