PDA

View Full Version : Athlon XP 2400+ & GeForce FX - 17,000 3DMarks


Pages : [1] 2

SurfMonkey
01-15-03, 03:30 AM
The GF FX has been put through its paces at the CES in Sweden. It managed to make it to 17k 3DMarks. This was probably running at default settings.

You can get the story here (http://www.*********.com/index.php#1031) or go straight to source (if you can read swedish!) here (http://www.nordichardware.se/nyheter/index.php?category=10&id=4919&PHPSESSID=0a076dd4e0e69b967ef5a5e98d37ebdd)

Good, bad, or indifferent. It's definately faster than my Ti4400, but is it fast enough? At default settings in 3DMark the FX would definately be CPU bound. Since most overclocked R300s are hitting 16k on P4s, we need to see some of the visual gumbo in action to get a real score!

marqmajere
01-15-03, 03:38 AM
About what I thought. I didn't expect it to blow the 9700 out of the water. The next incarnation of the FX and 9700 will really be a showdown I'm betting. :)

-=DVS=-
01-15-03, 03:46 AM
Review/rumor says it hits around 17,000 with defoult setings (no AA or AF) very good score on AthlonXp 2400+

Radeon9700Pro goes around 14,000 , but then again Geforce4 scores close to Radeon 9700Pro on defoult speed, its not a suprise NV30 have high score it have high core/mem 500 afterall,

Real test would be to see what can it do with 4xAA and 8xAF

Considering how low the Nature scene was given from MaximumPC in their beta preview of the GeForce FX it seems that the results are not that far off :(

If score is low as it was in early previews then its disapointing,
Card will be fast without AA but not much faster then R300 in AA games witch everyone plays :rolleyes:

stupid Nvidia why they didn't use 256bus , we would have a stellar card by now :confused:

StealthHawk
01-15-03, 05:30 AM
3dmark is really CPU dependant, so the result isn't really all that surprising. what we really care about is AF+FSAA scores, as others have stated. that will tell the true story.

scott123
01-15-03, 07:42 AM
The only good news is with benchmarks like that we won't be seeing a overprice FX. If we do, then Nvidia can expect very poor sales, and ATI will reap the benifits.

Scott

zakelwe
01-15-03, 08:40 AM
I cannot make my mind up about how the GF FX will do with AA and AF.

With the physical specs of the DDR2 memory it uses and it's bandwidth you'd expect it to be about the same as the R300 or slightly worse, but then again the software side of the coin may make things a lot better. We don't know how good nvidia's claims of compression etc are. Certainly they have a lot of software engineers and in the past they have released drivers that have boosted framerates especially when AA and AF are present.

I don't know much about this subject but I would have a idiots guess that with AA and AF you have more room on the software side to improve matters than when running au natural. So it's a bit early to guess I think how it will do with AA and AF. Maybe that's the reason for the delay? If it comes out with Detonator drivers starting off 60.32 then we'll know that the entire driver team is about to have a nice long holiday before they expire :)

AF and AA might not be relevant anyway before R400 and nv40 as Doom will knock everyone back to 1024 x 768 without any ( only half joking ..)

Regards

Andy

savyj
01-17-03, 12:26 AM
The only good news is with benchmarks like that we won't be seeing a overprice FX. If we do, then Nvidia can expect very poor sales, and ATI will reap the benifits.

Isn't that just as much a difference in 3DMark scores as the ATI 9700 Pro to the NVIDIA GeForce 4 Ti4600? Didn't the 9700 Pro originally ship at $400?

Here's the way I see it. Performancewise, the ATI 9700 Pro to the NVIDIA GeForce 4 Ti4600 will be about the same as the GeForce FX is to the 9700 Pro. Both ATI and NVIDIA will have charged $400 for their latest generation cards (well, if you believe/disbelieve certains rumors). If this theory bears fruit, I think you're statement's a bit skewed.

gokickrocks
01-17-03, 12:57 AM
the msrp was at 400, at the release, the card could be had for as low as 320

savyj
01-17-03, 11:32 AM
There are still places selling the 9700 Pro for over $350. Are you sure it was available at launch for $320? Shopping around at that time I couldn't find it for anything less than about $380 (but I didn't look that hard, just at Best Buy, CompUSA, and newegg.com). Needless to say, I don't see how that's any different for the GeForce FX. It'll still be a $400 MRSP card and the price will surely start to decline slowly after launch.

Bigus Dickus
01-17-03, 11:57 AM
Originally posted by savyj
Performancewise, the ATI 9700 Pro to the NVIDIA GeForce 4 Ti4600 will be about the same as the GeForce FX is to the 9700 Pro.
Do you actually believe that? The 9700Pro is about as CPU limited as the Ti4600 in many cases. But, when it's not, such as in high res with AA/AF, it is 2X, 3X, or 4X faster than the Ti4600.

The GFFX will also be CPU limited in low IQ settings, just as the 9700Pro and Ti4600 are. In those cases, it might have a 10% or 20% advantage due to driver optimizations, but do you really expect it to be 100%, 200% or 300% faster than the 9700Pro when it counts?

Surely not...

jbirney
01-17-03, 02:16 PM
Originally posted by savyj
There are still places selling the 9700 Pro for over $350. Are you sure it was available at launch for $320? Shopping around at that time I couldn't find it for anything less than about $380 (but I didn't look that hard, just at Best Buy, CompUSA, and newegg.com). Needless to say, I don't see how that's any different for the GeForce FX. It'll still be a $400 MRSP card and the price will surely start to decline slowly after launch.

Four days after the 9700pro started to show up at on-line stores I found my for $330 with free shipping and I had to pay no sales tax. Thus I got my card about 7 days after it was avaible. Not too bad but I doubt you will see that with the FX.

Lars
01-17-03, 03:21 PM
I'm very interested in the AA + AF performance of the Geforce FX as well. As an avid benchmarker, I can assure you that even an overclocked R9700PRO is very cpu limited indeed. Even with my Vapochilled P4 ~3.5 Ghz.

The Feforce FX getting as much as 17k with a 2400+ with PC2700 and standard settings in general is most impressive IMO. Not that it really matters, as 1280x1024x32 with 8-16xAF + 4-6xFSAA is what I'm looking for.

I'm quite confident most potential Geforce FX buyers will be looking for high res. and high IQ results primarily.

BTW, there's a 3DMark03 teaser out over at Futuremark. I'm thinking maybe the default test resolution will be 1280x1024x32 with this release?

Arrgh, enough of this waiting game, just give us the blo*** cards.... NOW :D

savyj
01-18-03, 03:30 PM
Do you actually believe that? The 9700Pro is about as CPU limited as the Ti4600 in many cases. But, when it's not, such as in high res with AA/AF, it is 2X, 3X, or 4X faster than the Ti4600.

CPU limitation is a fact that all GPU's have to deal with. Ignoring anything that is CPU limited ignores a majority of what people do with their computers, which is just 'spin' in my opinion. You have to leverage the sweet spot benchmarks of the 9700 Pro against the Ti 4600 to get anywhere near the exagerated differences of 2X, 3X and 4X. Overall, and I mean OVERALL, the Radeon 9700 is just a rung up the ladder compared to the Ti 4600. Given the ever increasing focus on AA and IQ it should have been a given the 9700 Pro would have grown quantum leaps in those areas. Will the FX be 2X, 3X or 4X faster than 9700 Pro? Doubtful, nor would I expect it to be. Will it be 2X faster in a few sweet spot benchmarks? Probably.

I, for one, don't care. I'm more interested in the big picture. I can say this for sure, the FX will be 2X, 3X, 4X or 5X faster than the Ti4600 in those same 9700 Pro sweet spots and that's good enough for me. Do I think the FX will be the same one rung up on the 9700 Pro that the 9700 Pro was up on the Ti 4600? For me... Yep...for those sporting tunnel vision goggles, I doubt it.

savyj
01-18-03, 03:55 PM
Four days after the 9700pro started to show up at on-line stores I found my for $330 with free shipping and I had to pay no sales tax. Thus I got my card about 7 days after it was avaible. Not too bad but I doubt you will see that with the FX.

I'd have to say you're right on that. With the limited supply of FX's out of the gate, prices will stay near MRSP. Still, MRSP's for both cards will have been $400 ;)

PreservedSwine
01-18-03, 04:49 PM
I, for one, don't care. I'm more interested in the big picture. I can say this for sure, the FX will be 2X, 3X, 4X or 5X faster than the Ti4600 in those same 9700 Pro sweet spots and that's good enough for me. Do I think the FX will be the same one rung up on the 9700 Pro that the 9700 Pro was up on the Ti 4600? For me... Yep...for those sporting tunnel vision goggles, I doubt it.

Ohhh, the irony:(

PreservedSwine
01-18-03, 05:19 PM
I think it's still largely up for debate just what the "sweet spots" as you call them, will be on the FX benches.

Since it only has 16GB of raw bandwidth, it will be quite a challenge, even with it 500mhzDDRIIRAM and 500mhzcore, to keep pace with the R9700 in heavy FSAA and heavy AF applications.

I think you might find the FX's "sweet spot" a little less satisfying than you hope. It will have fps to spare in low FSAA ap's, but then again, so does every other graphics card ocer $100 bucks.

It's when the going gets tough that your'e digging into your wallet for...and it remains to be seen if the FX will be able to deliver with heavy FSAA, high resolutions, all with it's anemic 128bit bus....

Maybe, someday, they'll actually finish this long overdue GPU....and allow someone to put out a review....

savyj
01-18-03, 08:10 PM
I think it's still largely up for debate just what the "sweet spots" as you call them, will be on the FX benches.

It's ALL up for debate until (and even after, to a point) the darn reviews come out.

Since it only has 16GB of raw bandwidth, it will be quite a challenge, even with it 500mhzDDRIIRAM and 500mhzcore, to keep pace with the R9700 in heavy FSAA and heavy AF applications.

I'd have to disagree here. The 9700 Pro only has 23% more THEORETICAL bandwidth and as we all know, theoretical bandwidth is never reached due to various overheads inherent in every design. So simple math could be way off here. I for one think NVIDIA has always been better at wringing performance out of their hardware through effeciency goals both at the hardware and software level. I'm sure they'll do better in light FSAA (2x-4x) and light AF (4), I guess I'm less sure about the heavy ones (6x-8x FSAA and 8-32 AF) since the 9700 Pro seems to be just incredible at those.

No, I think if NVIDIA has done their homework right (and for goodness sakes, they've had quite a bit of time to do it) they could outperform the 9700 Pro even with their "anemic" 128bit bus. I think they've really gone for a completely new foundation with the NV30 and that it'll be easier for them to refresh this product line than past ones. Of course, this is all speculation, on either side. Given the amount of time they've had, the card specs and their history, I'm gonna give NVIDIA the benefit of the doubt and conclude that they will indeed reclaim the performance crown from ATI. Then it's all about the R350 and the next NVIDIA products (which seems to be confusing right now so I'll refrain from trying to layout the roadmap) and their shipping schedule. NVIDIA's gonna need to perform better as a company to keep from losing more market share, even if the FX is a hit.

Kudos to ATI for laying on the heat, here's to the next few weeks for some new editions to the GPU melee ;)

Phyre
01-18-03, 08:51 PM
Well, considering what has been rumored with the bandwitdth, 128-bit vs. 256-bit, etc., what do you guys think the possibility of the refresh having a 256-bit bus? Considering the way this area works, I would like to see it, but I don't think it is possible. However, it would be nice to see what the GeForce FX chip could do on a 256-bit bus if current rumors are accurate.

Phyre

Bigus Dickus
01-18-03, 10:27 PM
Originally posted by savyj
Do I think the FX will be the same one rung up on the 9700 Pro that the 9700 Pro was up on the Ti 4600? For me... Yep...for those sporting tunnel vision goggles, I doubt it.
Wow. Your logic is just astonishing.

Let me see if I have this straight: you are saying that for you what matters are settings where the cards are CPU limited (and therefore would be perfectly playable even on a Ti4200 or Radeon 8500), and not settings where you actually need the power of a 9700 or GFFX? And so long as the GFFX is as much faster in than the 9700 as the 9700 was over the Ti4600 in cases where all three are running at insane framerates, then the GFFX is as much better than the 9700 as it was relative to the Ti4600???

How's the food on your planet?

savyj
01-18-03, 11:06 PM
Let me see if I have this straight: you are saying that for you what matters are settings where the cards are CPU limited (and therefore would be perfectly playable even on a Ti4200 or Radeon 8500), and not settings where you actually need the power of a 9700 or GFFX? And so long as the GFFX is as much faster in than the 9700 as the 9700 was over the Ti4600 in cases where all three are running at insane framerates, then the GFFX is as much better than the 9700 as it was relative to the Ti4600???


No, let me clarify just special for you BD. All settings on all benchmarks are important to me, not just ones which let a particular card shine. Let me give you some examples so you don't get confused again:

GPU 1
-------
Benchmark A @ 1024x768 --- 17,000 points
Benchmark B @ 1024x768 --- 200 fps
Benchmark B @ 1600x1200, 4x FSAA, 8 AF -- 100 fps
Benchmark B @ 1600x1200, 8x FSAA, 32 AF -- 40 fps

GPU 2
-------
Benchmark A @ 1024x768 --- 14,000 points
Benchmark B @ 1024x768 --- 150 fps
Benchmark B @ 1600x1200, 4x FSAA, 8 AF -- 85 fps
Benchmark B @ 1600x1200, 8x FSAA, 32 AF -- 45 fps

I would rather have GPU 1 and not GPU 2 even though at the highest settings GPU 2 is it's over 10% faster than GPU 1. Why?
Because OVERALL GPU 1 is superior to GPU 2. Hopefully that clarifies it for ya.

StealthHawk
01-18-03, 11:23 PM
Originally posted by savyj
I'm sure they'll do better in light FSAA (2x-4x) and light AF (4-8), I guess I'm less sure about the heavy ones (8x FSAA and 16-32 AF) since the 9700 Pro seems to be just incredible at those.

a few errors in your hypothetical.
1) r9700 maxes out at 6x FSAA. it is also questionable at this point whether gfFX's 8x FSAA will match r9700's 6x.

2) gfFX has a max of 8x AF.

savyj
01-18-03, 11:25 PM
Well, considering what has been rumored with the bandwitdth, 128-bit vs. 256-bit, etc., what do you guys think the possibility of the refresh having a 256-bit bus? Considering the way this area works, I would like to see it, but I don't think it is possible. However, it would be nice to see what the GeForce FX chip could do on a 256-bit bus if current rumors are accurate.

Oh, I'd definitely say that a 256-bit bus is in the works. It's just not necessary right now in NVIDIA's eyes and if it's not necessary they're not going to use it because a 256-bit bus costs much more to implement (or so the rumor goes). ATI chose the 256-bit bus because they needed an edge and that definitely paid off for them. NVIDIA decided to go with insanely high bandwidth RAM on a 128 bit bus in hopes that they could just start increasing bits and pieces of the GPU and release them as product refreshes (IMHO) so they could make decent margins on their products. I think it will all work out in the end for NVIDIA. Of course I could be wrong. Then again, if all of ATI's upper management goes to prison who's to say that things might not get real strange later in the year :D

I for one hope not 'cause it sucks for the consumer when there's only one player in an industry.

savyj
01-18-03, 11:26 PM
a few errors in your hypothetical.

Well, you got me there. I fixed the original post so as to not confuse further.

savyj
01-18-03, 11:38 PM
I'm not sure it would actually be that difficult for NVIDIA to go 256-bit bus. Don't they use a crossbar memory controller consisting of 4 32-bit busses now? If that's the case and it's a switched architecture, I don't think it would be difficult for them to add 4 more 32-bit busses into the mix. Just more expensive I suppose. Could be wrong...anyone have better ideas on this?

Bigus Dickus
01-18-03, 11:39 PM
Originally posted by savyj
GPU 1
-------
Benchmark A @ 1024x768 --- 17,000 points
Benchmark B @ 1024x768 --- 200 fps
Benchmark B @ 1600x1200, 4x FSAA, 8 AF -- 100 fps
Benchmark B @ 1600x1200, 8x FSAA, 32 AF -- 40 fps

GPU 2
-------
Benchmark A @ 1024x768 --- 14,000 points
Benchmark B @ 1024x768 --- 150 fps
Benchmark B @ 1600x1200, 4x FSAA, 8 AF -- 85 fps
Benchmark B @ 1600x1200, 8x FSAA, 32 AF -- 45 fps

I would rather have GPU 1 and not GPU 2 even though at the highest settings GPU 2 is it's over 10% faster than GPU 1. Why?
Because OVERALL GPU 1 is superior to GPU 2. Hopefully that clarifies it for ya, otherwise I'm gonna have to get out the crayons and butcherpaper and have my kids draw it out for ya.
I think you should start passing out the crayons! :D

First, let me point out that even the numbers you posted above illustrate my point perfectly. Look at them closely: Benchmark A is, well, a benchmark and not a game so it doesn't even matter. Loot at benchmark B's results: GPU2 is still maxing out your monitor's refresh rate when it is "losing" to GPU1. What does it matter if it gets 85 fps or 382 fps? If those results included minimum framerates, then perhaps we could have a discussion about that. But look at the last example: GPU2 is faster when you need it to be (i.e., when you're not maxing out your refresh rate).

Make sense? I guess it's a different definition of "overall." To me, scores at CPU dependent settings, or in settings where you're already getting 100+ fps don't mean a dang thing because all the cards are ultimately limited by your monitor. It's only in settings where you begin to drop below your refresh rate that concern me, which just happen to be high resolutions, high AA, and high AF settings. That's where the 9700 blows away the Ti4600 (again, when you need that extra muscle), and there's no chance that the GFFX will have that kind of margin over the 9700. Therefore, from where I'm standing, there's no chance that the GFFX will be anything like the kind of improvement over the 9700 that the 9700 was over the Ti4600. Even if it was 50% faster in every medium res no AA no AF benchmark. Why? That 50% doesn't do jack for what I see on my screen.

Now, having said that, I have to say that I don't think real benchmarks will necessarily be representative of the numbers you posted. I think the third line will show the two cards even closer than 85 vs. 100.