PDA

View Full Version : Discuss BFG 5700 Ultra Review


Pages : [1] 2

Clay
10-23-03, 09:59 AM
Here is where you can discuss my latest review here at nV News, ask questions, offer suggestions, etc, etc.

Link to review:
http://www.nvnews.net/reviews/bfg_geforce_fx_5700_ultra

panzaman
10-23-03, 10:03 AM
...yep, where's the review?

The Baron
10-23-03, 10:11 AM
Except it works now. So this is a MOOT POINT OF DOOM!

The Baron
10-23-03, 10:16 AM
My SWAG on these results is that the BFG 5700 Ultra held its own for the most part regarding OpenGL performance. The large discrepancy in the Direct3D demos is a more difficult call and one I'll wait on making until I know more. At any rate, these are not representative of game performance and hopefully Humus (or others) can shed some light on things in the near future.

Well... to be nitpicky:

1. The first demo didn't work because of the render targets bit. GFFX cards do not support it, as far as I know.

2. The performance delta is because of PS2.0 speed, basically, and the fact that a 9800 Pro is 8x1 while the 5700 Ultra is... uhhh... 4x2? 2x2?

Clay
10-23-03, 10:21 AM
Thanks Tim, I'll add your clarifications to the review ASAP...got that link to page 9 fixed...that's what I get for not sleeping last night! :D

mikechai
10-23-03, 10:29 AM
Reading it now ...
++ lots of nice pics.
++ Nice layout.

Clay
10-23-03, 10:30 AM
Originally posted by mikechai
Reading it now ...
++ lots of nice pics.
++ Nice layout.

Thanks! I wish I'd had more time...I will be adding more games to this review if at all possible.

MUYA
10-23-03, 10:33 AM
Nice stuff Clay!!! I will post front page asap....not on my own computer, long stiry made short, i made a booboo while flashing my mobo, got it fixed at the local ABIt distributors. !! But lets not let that detract from a nice review mate!!

I wonder how it fare against the 9600XT head to head.

Cheers
MUYA

The Baron
10-23-03, 10:40 AM
Very tasty review. Great job on such a short time span.

...and Sweet Mother of God, you quoted me. Scary. If you'd like, I could shoot you 9600 Pro numbers from Icetomb (I could do some crazy underclocking stuff, try to get my machine as close to yours as possible).

Clay
10-23-03, 10:41 AM
Originally posted by MUYA
Nice stuff Clay!!! I will post front page asap....not on my own computer, long stiry made short, i made a booboo while flashing my mobo, got it fixed at the local ABIt distributors. !! But lets not let that detract from a nice review mate!!

I wonder how it fare against the 9600XT head to head.

Cheers
MUYA
Thanks MUYA! I already posted, Mike asked that I do that...took me forever...couldn't find my login info! I so sleepie...must...get...shower...and...coffee... :D

I heard about your mobo woes, glad you're back online!! Can you send out e-mail notifications? Just wondering if Mike asked you to or not, he didn't say anything to me about that and I know you're the one with all the connections. :D

Later!

Clay
10-23-03, 10:44 AM
Originally posted by The Baron
Very tasty review. Great job on such a short time span.

...and Sweet Mother of God, you quoted me. Scary. If you'd like, I could shoot you 9600 Pro numbers from Icetomb (I could do some crazy underclocking stuff, try to get my machine as close to yours as possible).
Well hey, thanks Tim. :D Yeah you're quoteworthy from time-to-time!

Sure, I'd love to get some 9600Pro numbers from you...now now...my puter ain't all THAT slow. Hhahaah...I need a freakin' upgrade! I'm outta here for a while...whew! :)

darkmiasma
10-23-03, 10:48 AM
i liked the layout and thought it was a very well organized review. i am disappointed that you did not benchmark the card at stock speeds in addition to the overclocked speeds.

overall i thought you did a good job, i just would have preferred that extra bit of info be included.

- mike

panzaman
10-23-03, 10:54 AM
..uhmm, since this is supposed to be a DX9 card I would have preferred some real world Shader intensive benchmarks like Splinter Cell, Tomb Raider, Halo etc.....

Clay
10-23-03, 10:59 AM
Originally posted by panzaman
..uhmm, since this is supposed to be a DX9 card I would have preferred some real world Shader intensive benchmarks like Splinter Cell, Tomb Raider, Halo etc.....
The actual relevance of TRAOD as an accurate DX9 benchmark is becoming more and more debateable. I only had Halo demo (and it stinks IMO, very dissapointed in it) and I only had Splinter Cell demo. I admit, I need to increase my game stockpile but this was the best I could do on a very short timeframe. I appreciate your feedback though and your point is well taken. If at all possible I will update this review with a DX9 game in the near future. Thanks! :)

Clay
10-23-03, 11:01 AM
Originally posted by darkmiasma
i liked the layout and thought it was a very well organized review. i am disappointed that you did not benchmark the card at stock speeds in addition to the overclocked speeds.

overall i thought you did a good job, i just would have preferred that extra bit of info be included.

- mike

Thanks Mike. Yes I chose not to include stock speeds to save time. I may come back to this with those in the near future. Thanks for the feedback!

Paul
10-23-03, 11:02 AM
The fact you're not taking screenshots from exactly the same place makes the image quality comparisons almost pointless. Angle variance can make a world of difference with the kind of small differences you're looking for in IQ comparisons.

No mention of ATi being able to full trillinear (when the settings are correct), nor commentary on the nVidia drivers, was disappointing also. Also, shouldn't you have been testing with stock clock speeds INSTEAD of the o/c speeds? Much better to establish a baseline result then build upon that at a later date (if you were that pushed for time), than to get the maximum speeds straight away - Many readers/purchasers may not be able (or want) to get those same o/c clock values, making the scores invalid for them.

But thanks for taking the time to do the review. Much appreciated.

Clay
10-23-03, 11:47 AM
Originally posted by Paul
The fact you're not taking screenshots from exactly the same place makes the image quality comparisons almost pointless. Angle variance can make a world of difference with the kind of small differences you're looking for in IQ comparisons.

No mention of ATi being able to full trillinear (when the settings are correct), nor commentary on the nVidia drivers, was disappointing also. Also, shouldn't you have been testing with stock clock speeds INSTEAD of the o/c speeds? Much better to establish a baseline result then build upon that at a later date (if you were that pushed for time), than to get the maximum speeds straight away - Many readers/purchasers may not be able (or want) to get those same o/c clock values, making the scores invalid for them.

But thanks for taking the time to do the review. Much appreciated.
Wow, everyone is providing great feedback. I can't really disagree with you on many of your points. I will say the following though:
This was not a driver review. With all that is involved with drivers lately they really merit a review in and of themselves. I would like to have included a bit more, however, the timeframe limited me.

Regarding clock speeds, it's a toss up really. People would complain if I tested only default clocks. The best of both worlds is ideal but I chose to OC because it helps to prove the cards mettle since it could obviuosly go slower. Again, gotta play the time card unfortunately.


Thanks for showing appreciation. Many people don't realize what kind of time and effort goes into a review. I will definitely keep you comments in mind on future reviews. (or updates to this review)

Paul
10-23-03, 12:09 PM
Originally posted by maxpower
This was not a driver review. With all that is involved with drivers lately they really merit a review in and of themselves. I would like to have included a bit more, however, the timeframe limited me.

I'm not really looking for a driver review, merely just some kind of acknowledgment that cheating has gone on in the past, and that those same cheats may still be present in the drivers used in the review - along with the trillinear "optimisation" already mentioned above. Coupled with a quick overview of any new features in the drivers, and that would be more than enough.

I know that time constraints prevent anything in-depth, but I think it would be nice to have a paragraph or two in there just explaining the situation, and perhaps putting any speed increases into perspective.

jbirney
10-23-03, 12:44 PM
I think lack of Tri-linear should be counted as a negitivie as trilinear was interoduced when, 1998???

Also poor ps2.0 performance should have also been considered.

That being said I think you did a good job and the 5700 is a decent card.

Clay
10-23-03, 12:45 PM
Originally posted by Paul
I'm not really looking for a driver review, merely just some kind of acknowledgment that cheating has gone on in the past, and that those same cheats may still be present in the drivers used in the review - along with the trillinear "optimisation" already mentioned above. Coupled with a quick overview of any new features in the drivers, and that would be more than enough.

I know that time constraints prevent anything in-depth, but I think it would be nice to have a paragraph or two in there just explaining the situation, and perhaps putting any speed increases into perspective.
I understand where you're coming from and I think I will add a paragraph or two on the conclusion page addressing all of this. Honestly, it's such a can of worms that I think too much attention to it can actually devalue a review. Everyone in this business cheats/optimizes/etc to stay ahead. There's even some new info on ATI cheats out today:
http://www20.tomshardware.com/graphic/20031023/nvidia-nv38-nv36-19.html

At any rate, I agree with you that some acknowledgement of this issue should be in the review so you can look for it to be there soon.

Thanks again for your intelligent feedback.

Clay
10-23-03, 12:51 PM
Originally posted by jbirney
I think lack of Tri-linear should be counted as a negitivie as trilinear was interoduced when, 1998???

Also poor ps2.0 performance should have also been considered.

That being said I think you did a good job and the 5700 is a decent card.
Good points along the same lines as Paul's. I will include the trilinear as one of the cons and update the conclusion page. Regarding PS2.0 performance...I'm still not fully convinced as to the validity of some DX9 game's PS2.0 relevance. So much still seems to be up for debate. I'm always willing to hear anyone, read more info that I haven't read before, etc. In short, I'm admitting that I don't personally have enough knowledge to on this topic to firmly plant my opinion on it in either side of the arguments. Thank you for your feedback, good stuff.

jbirney
10-23-03, 01:13 PM
Originally posted by maxpower
Regarding PS2.0 performance...I'm still not fully convinced as to the validity of some DX9 game's PS2.0 relevance. So much still seems to be up for debate. I'm always willing to hear anyone, read more info that I haven't read before, etc. In short, I'm admitting that I don't personally have enough knowledge to on this topic to firmly plant my opinion on it in either side of the arguments. Thank you for your feedback, good stuff.

Thank you for taking the time to consider it. If you need more back ground info head over to the techreport as they have a really nice graph:

http://www.tech-report.com/reviews/2003q4/geforcefx-5950ultra/index.x?pg=11

If you check out their 5700 you will see the same type of delta. This can be found in other reviews.

Now the billion dollar question, how will that relate to games? Not sure. If HL2 and TR:AOD are anything to go by....

darkmiasma
10-23-03, 01:27 PM
Originally posted by maxpower
Regarding clock speeds, it's a toss up really. People would complain if I tested only default clocks. The best of both worlds is ideal but I chose to OC because it helps to prove the cards mettle since it could obviuosly go slower. Again, gotta play the time card unfortunately.

The problem is, everyone's card will come @ stock speeds, and not everyone will obtain the same overclock. You have to establish the baseline, and in this case, you failed to do so.

Its nice to look at a card that has been overclocked and say, wow it made such a difference, but when you don't have that baseline to compare it to, it just doesn't look the same.

Hellbinder
10-23-03, 01:38 PM
Originally posted by maxpower
I understand where you're coming from and I think I will add a paragraph or two on the conclusion page addressing all of this. Honestly, it's such a can of worms that I think too much attention to it can actually devalue a review. Everyone in this business cheats/optimizes/etc to stay ahead. There's even some new info on ATI cheats out today:
http://www20.tomshardware.com/graphic/20031023/nvidia-nv38-nv36-19.html

At any rate, I agree with you that some acknowledgement of this issue should be in the review so you can look for it to be there soon.

Thanks again for your intelligent feedback.
Becuse Lars says they are Cheats?? where was his comparrison to Raster images for a baseline? simply a comparrison to the way nvidia did it. The UT2003 issue is a valid one though.

Lets not forget that Nvidia hand dug these out of the woodwork and gave them to various sites on a silver platter with their own comentary.

Cheat or not it needs to be established by correct methods and from outside sources exactly why there are differences. Nvidia was not guilty of aggressive optimizations or cheats based on comparrisons to Ati.. but comparrisons to the Basline Raster image.

Humus
10-23-03, 01:46 PM
Originally posted by The Baron
Well... to be nitpicky:

1. The first demo didn't work because of the render targets bit. GFFX cards do not support it, as far as I know.

2. The performance delta is because of PS2.0 speed, basically, and the fact that a 9800 Pro is 8x1 while the 5700 Ultra is... uhhh... 4x2? 2x2?

True. The render target is a floating point cubemap. The floating point texture support on the GFFX isn't that great. It only supports texture_rectangle, no standard 1D/2D/3D/Cube, and no mipmapping or wrap mode.
And that the last number of demos run slow are indeed as you say, the GFFX is slow at PS2.0 shaders. The OpenGL demos used are kinda old, so they use PS 1.x level shaders, and won't see as dramatic performance difference.