Softmoding X1800XT to FireGL Counterpart

octopus41092

Golden Member
Feb 23, 2008
1,841
0
76
So, I do a heavy 3D modeling. I'm currently up to about 6 million + polygons and in 3D Studio Max my HD4850 doesn't handle it. My question is would softmodding the X1800XT into the FireGL counterpart by editing the BIOS be better than the HD4850? Or would the HD4850 still be better hands down just because its that much faster?
 

QuixoticOne

Golden Member
Nov 4, 2005
1,855
0
0
There's no way an X1800XT or anything similar is going to be faster at *anything* than a 4850.

IIRC the ways the pro cards were different than the consumer cards were that certain features like
line-AA or stereo output or whatever were artificially disabled in the driver software for the consumer cards
whereas these were allowed to be used for the industrial model cards.

I believe in many cases the hardware is very similar or effectively identical (maybe with some differences in
MHz speeds and RAM quantity and of course the built in model number) for a industrial and a consumer card.

I believe that a lot of CAD/CAM/graphics programs that formerly supported the industrial model driver driver/API features may additionally or alternatively support the use of general purpose shader algorithms to garner the same effects via GPU acceleration without using the industrial APIs to do it. At least with DX9 / PS3 level cards that have general purpose shaders this should be very possible, and I'd see little reason that it wouldn't be commonly done just to permit users to use their software effectively with the consumer model cards.

So I'd expect the old generation FireGL cards to be vastly inferior to a 4850 in performance, and even if you could mod the X1800XT into a FireGL successfully I doubt it'd unlock any feature that would make a relevant difference for you, especially if your software either doesn't use those features, or has an alternative shader based implementation of those features that works with the 4850. In most cases the features are more about aesthetics and quality as opposed to performance anyway.

If your performance / capacity is lacking, I'd try to figure out if it is RAM limit, in which case going to a GPU
card with 1GB RAM may help and indeed that may be the major possible improvement. I thought I had heard that some 1GB 4870 models would be made; I'm not sure about their availability at the moment.
I believe some of the higher end NVIDIA cards have 1GB now too, but, again, I don't recall the details.

Also if you run under Vista 64 you might get some benefit from the 64 bit platform as well as the DX10 /Vista style virtualization of video RAM into system RAM, though if you're dealing with huge models that will be viewed all at once, it would be much better if your GPU had the RAM to hold it all on-board.

 

qliveur

Diamond Member
Mar 25, 2007
4,087
69
91
Wouldn't a 4850 be around 3x (ballpark estimate) as fast as an X1800XT?
 

JPB

Diamond Member
Jul 4, 2005
4,064
89
91
X1900XT faster than X1800XT
HD 3870 = twice as fast as X1900XT. Probably more.
HD 4850 = sometimes faster than 3870 Crossfire *or* 3870X2

I would say, it would take about 4-5 X1800's to = a HD4850
 

unr3al

Senior member
Jun 10, 2008
214
1
81
www.link-up.co.za
Even the HD3850 beats the pants off an X1800XT... In fact the HD3850 beats an X1950XT so I can't see how it could be possible by any remote chance that a flashed X1800XT could beat an HD4850. The FireGL BIOS doesn't make the card faster, it simply enables features that the Radeon doesn't have. If it can be flashed with said BIOS, that is.
 

PingSpike

Lifer
Feb 25, 2004
21,729
559
126
Originally posted by: octopus41092
Ok, well what are those features that makes the FireGL's so great?

Probably nothing really...I was under the impression that the pro versions of cards are pretty much just the regular version with a higher price tag and better support, maybe some software tuned to different applications of the hardware?
 

myocardia

Diamond Member
Jun 21, 2003
9,291
30
91
Originally posted by: qliveur
Wouldn't a 4850 be around 3x (ballpark estimate) as fast as an X1800XT?

Yeah, if not moreso, assuming we're talikng about gaming. The problem is that a FireGL card is about 2.5 or 3x faster than it's gaming equivalent (the gaming card with the same GPU and RAM). A long-time member here (Duvie) said that his 6800GT was ~2.5x as fast at rendering CAD, once he soft-modded it into a Quadro, as it was before it was soft-modded, and that was using the same CPU. RAM, etc.

My guess is that an X1800XT that has been flashed to it's FireGL equivalent would be roughly as fast as a 4850, although it would most likely give its owner better rendering image quality than the 4850. There's only one way to find out, though.:D
 

octopus41092

Golden Member
Feb 23, 2008
1,841
0
76
Hrmm.... So, is there a HD4850 FireGL counter part yet? If I could just simply flash the BIOS to switch and flash to go back. I could just switch between them depending on what I'm doing. Because according to what myocardia said, wouldn't that make my HD4850 be crazy fast?

Edit: I just realized something.... My CPU utilization is going up to 100% while my GPU is barely hitting 10%. Is there any way to offset CPU utilization and use the GPU instead? Or is my only solution to buy a faster processor (Currently have an X2 6000+)? The program I use is 3D Studio Max and Mudbox. Both are Autodesk programs.
 

QuixoticOne

Golden Member
Nov 4, 2005
1,855
0
0
The X2 may be a limiting factor. Check your RAM bandwidth as well as L2 cache bandwidth and total cache size. A P35 / Q6600 can push 4-5GBy / second RAM bandwidth to DDR2, and around 15GBy/sec to 20GBy/sec L2 cache bandwidth.

Your X2 is probably better than mine if it is an AM2 platform, so I don't know quite how it compares, but I was surprised to see how good the Intel chips really are compared to the S939 older X2 platforms anyway.

http://area.autodesk.com/index...rums/viewthread/15350/

Posted by blindartist on May 25th, 2008, 04:24 AM
3ds, maya and pretty much every modeling suite do not use gpu's for rendering, rendering is 100% cpu, they only use gpu for real time viewpoints, not rendering, there are a couple half baked plugins(RTsquare) to let 3ds render with a gpu but their extreamly weak and look awful with very few features
Quote
Posted by morbo on May 25th, 2008, 08:25 AM
Quote
Posted by Unixlord on May 25th, 2008, 01:17 PM
I take it you've heard of Nvidia Gelato. You're right about the currect implementation not being fully featured but it depends on what features you need (ray tracing is lacking for example) they're headed in the right direction. Well that's why they bought rayscale. To bring complete, fully featured hardware accelerated rendering on their quadros. Also using the viewpoints in real time is no joke. Unless you're working on a simple scene with few objects you'll need gpu acceleration.
Quote
Posted by blindartist on May 25th, 2008, 03:21 PM
the results that rtsquare or gelato yield right now are extreamly basic and useless for just about any real application, its a step in the right direction sure but for all intents and pruposes it is usless for any serious or profesional user, and yes veiwpoint performence bottoms out after you go over about 2 or 300k poly's unless youre sporting a quatro or firegl, mainstream video cards dont really cut it for this stuff i find, but i cant afford one of thoes, i can push my scenes up to about 500000 poly's before i get any real viewpoint lag

For now, under Windows XP (32/64) I can tell you for certain that Max is NOT limited or crippled in ?gaming? cards. This is for DirectX. As for OpenGL, its just outright outdated and I don?t think its fair to use it for comparison to anything. Poor Mac/ Linux users aren?t really my concern here.

In WinXP in D3D in 3DS MAX 8+, any current gaming card will run leaps and bounds over ANY quadro/fireGL card. I?m saying D3D and Max 8+ as these are the only things I?ve tried myself. Max might be working well in versions earlier than 8 and OpenGL executions might be better on gaming cards too, but as far as I?m concerned both Max sub8 and OGL are seriously outdated, so I don?t really care about their performance.

http://www.hardwareanalysis.com/content/topic/71764/

http://www.rtsquare.eu/index.p...view&id=288&Itemid=256

It sounds like the GPU based rendering is not quite ready for serious use even though it does exist in the marketplace as a new and somewhat unproven option. Either wait 9 months or whatever for that capacity to mature in future software / hardware generations, or give up on the GPU when looking for more performance and upgrade your CPU system / build a render cluster with additional PCs or whatever.

From what people seem to be saying your current GPU card should be about as good as anything you can buy in the industrial grade series, certainly it isn't your performance limiting factor, so as long as the visualization works for you once the CPU rendering is done then it is doing its job.

If you have only one DIMM RAM then consider adding more to fully populate all your RAM slots so you can operate in dual channel or dual channel interleaved mode for a bit more RAM performance.

 

octopus41092

Golden Member
Feb 23, 2008
1,841
0
76
I do have an AM2 socket X2 6000+. How would I check the bandwith, using CPU-Z? Also if like you said viewport uses the GPU then why does my CPU usage jump up to 100% whereas my GPU doesn't even hit 10%. Does this mean that i'm severely CPU limited?
 

bunnyfubbles

Lifer
Sep 3, 2001
12,248
3
0
ugh, hate to see so many people chiming in when they don't appear to have true experience regarding the situation...
 

sgrinavi

Diamond Member
Jul 31, 2007
4,537
0
76
Originally posted by: bunnyfubbles
ugh, hate to see so many people chiming in when they don't appear to have true experience regarding the situation...

Right on.

There is a huge thread on this at GURU3D.COM which disucusses softmodding an ATI card into one of several graphics workstation cards. After you complete your softmod you will use the OpenGL graphics driver in max. It works very well, you will typically see performance in the same ballpark as the workstation card of the same gen.

You use RivaTuner and some scripts. The only problems is that it only works on XP and is hit or miss on XP64.


From what I gather the 2900 series are the easiest and most successful to softmod; HD2900 XT and FireGL 8600 use the same core.

The 38xx series can be modded to the FireGL 7700's, a less expensive option ( I have seen 3870's going for under $100 on ebay)
 

QuixoticOne

Golden Member
Nov 4, 2005
1,855
0
0
http://cpu.rightmark.org/products/rmma.shtml
http://cpu.rightmark.org/products/cpurm.shtml
http://www.memtest.org/

Those will tell you things like your L1/L2/RAM bandwidth and sizes and stuff.
If your RAM BW is better than 3000MBy/s then there's not a huge amount of benefit the newer CPU/memory platforms will give you since they top out around 5000MBy/s-6000MBy/s with typical fairly upper end modestly overclocked hardware.
I guess if less than 50% difference is killing you you could consider a change.

I don't know how important L2 size/BW is for your applications, check the benchmarks for a Q9450 vs an X2 AM2 to get an idea of the possible benefits of moving up.

I'm not sure what your CPU is doing when you're visualizing, it depends on whether you still have it doing job work in the background or if it is just dedicated to visualization of a fully static scene or what. If the CPU is still heavily used and the GPU isn't though it doesn't sound like a GPU is your bottleneck.
 

octopus41092

Golden Member
Feb 23, 2008
1,841
0
76
Erm... I don't really get how to use these but I ran the RMMT.exe which was the Multi-threaded memory test and I got 20000MB/s on the default settings. Could you explain how to use these things... lol.
 

Lord Banshee

Golden Member
Sep 8, 2004
1,495
0
0
use directX in 3dsmax and get the 4850. Most Pro features are only for opengl and 3dsmax is so much faster with directx there are no reasons "i know" to use opengl in 3dsmax.
 

sgrinavi

Diamond Member
Jul 31, 2007
4,537
0
76
Originally posted by: Lord Banshee
use directX in 3dsmax and get the 4850. Most Pro features are only for opengl and 3dsmax is so much faster with directx there are no reasons "i know" to use opengl in 3dsmax.

Yes, DirectX is faster than openGL if you are trying to run a gaming card. but for very large polygon models the openGL run on a workstation class card is more effective.

Softmodding the various ATI cards will yeild the desired results.