Anand's retail Radeon 9500 Pro review - much faster!!!....

[Oats]

Are you serious?????

Yes, I am I am seing games soon to be released or that have just been pushing a Radeon9700pro to its limits. A game like Asheron's call 2 comes to mind. It seems that software industry is catching up. There is little room in performance for AF/AA in these games. In such cases the 10-20% speed increase is not that all interesting to me, performance still remains average.

Maybe the value I place in these cards is out-of-sync with the prices we are seing these days. For a card that will offer good performance for about year, I am not willing to pay more than 150$. I see the radeon9500pro fall into that category.

 

[GraphixViolence]

Funny how when the Ti4200 started becoming available around 6 months ago for $199, it was hailed as the ultimate price/performance product. Now when the 9500 pro comes out at the same price with up to double the performance and DX9, some people think "it's too expensive".

Even better is when people start comparing the 9500 pro full retail price to the Ti4200's online discounted price. They seem to forget that once there's 10 different brands of 9500 pro on the market, you'll be able to find it online for around $150 or less. Even the 9700 pro is already available for ~75% of it's full list price.

 

[Doomtrooper]

Well said..(the usual suspects are the ones claiming its too expensive)...which will never change even if Nvidia went under they would paint the the big NV on their non NV Graphic card PCB.

 

[Oats]

Funny how when the Ti4200 started becoming available around 6 months ago for $199, it was hailed as the ultimate price/performance product. Now when the 9500 pro comes out at the same price with up to double the performance and DX9, some people think "it's too expensive".

Maybe the people finding the 9500Pro a bit expensive, found the 4200 a bit expensive then to? Maybe those people are actually not fanboys? Hard to beleive for some, but a lot of people don't need to value themselves through the success of others. A hint, slandering is not arguing.

For argument's sake, 6 months makes difference, 18 months over 12 of good gaming.

 

[Hellbinder]

Hyp-x

It has more than 50% advantage to R300, and given that R300 is rarely bandwidth limited, it can turn out to be a big advantage.

No, the Nv30 does not have more than a 50% advantage. Especially since there are no *shipping* Nv30 out yet. Even the trumped up Nvidia (from inside their own lab) numbers dont show an accross the board even 50% advantage. Doom-III is less than 30%, And They Do not have the as yet unrealeased ATI drivers that Enable Doubble sided Stencil support and other Doom-III enhancements. Easily attibuting around a 20% gain in doom-III performance.

There are some tough Questions that Nvidia and their supporters need to Think about. Its pretty damn clear that the Nv30 is flat out Slower than the R300 Clock-Clock. With a 200mhz core and ram advantage it is an average of 30% faster than the 9700pro based on pre-release info. I think many many people are forgetting that the 30% average gain is only being achieved by the Ti 5800 *Ultra*. And its 100$ more than the 9700pro. ITs pretty Questionable wether the *non ultra* version will even be faster than the 9700pro.

Bottom line Prediction? A crippled Nv30 with a lower clock is going to get its a$$ kicked. Or at best fall in between the 9500pro and 9700. How are they going to competatively price it?

 

[Doomtrooper]

Ahhh a price war..exactly what I need for my new Hammer set up after Xmas...bring it on baby...

BTW I think ATI will release a answer for the NV30 Ultra possibly a 400 mhz DDRII powered 9700...who knows. They are there (as seen on TechTV) and ATI has had 6 +months to prepare.

 

[Nagorak]

Isn't UT2003 a CPU dependent game anyhow?

Regardless. Here is what I have found:

Net effect = Turn (FSAA and/OR AF) on & 1024 is unplayable (FOR ME). Turn it off = playable.

Since benchmarking (done by anands for example) with an identical card at identical resolutions but with a faster processor yields higher frames I thought a faster CPU would allow things like FSAA to be playable all else being equal.

Someone educate me!

As Arjan said, you might go from being sometimes CPU limited and sometimes GPU limited, perhaps depending on the level you're playing or what part of the level you're in.

So, what you might avoid are the dips in framrates caused by the CPU (= higher overall framerates in a benchmark) but you will still experiance the dips caused by the GPU (which you must have since you're saying that enabling FSAA and aniso makes the game unplayable).

Also, by turning on FSAA and especially anisotropic you could easily be making yourself GPU limited. Anisotropic on the 4200 (the whole GF4 line in fact) offers crap for performance.

 

[Nagorak]

Are you serious?????

Yes, I am I am seing games soon to be released or that have just been pushing a Radeon9700pro to its limits. A game like Asheron's call 2 comes to mind. It seems that software industry is catching up. There is little room in performance for AF/AA in these games. In such cases the 10-20% speed increase is not that all interesting to me, performance still remains average.

Maybe the value I place in these cards is out-of-sync with the prices we are seing these days. For a card that will offer good performance for about year, I am not willing to pay more than 150$. I see the radeon9500pro fall into that category.

My point was ATi is aiming for the Ti4200's performance category, not the Ti4200's current price. The Ti4200 started out at $200, the only reason it's $120 now is because it has been discounted. Furthermore, at this time I wouldn't even consider buying a Ti4200. The performance just isn't there and at most it would last me a year.

It seems to me what you're saying doesn't make any sense. You complain about cards not lasting long enough, yet you'd prefer to buy one that doesn't last as long... A mainstream graphics card's decent performance life is approximately 18 months from its original release. If you buy a Ti4200 now it will last a year because it's already been out 6 months. If you spend $50 more you can get a R9500 Pro that will last you 18 months.

So you can pay $150 more every 12 months which equals out to $12.50 a month. Or you can spend $200 every 18 months which equals out to $11.11 per month. You have to think of more than just price when trying to figure out value. I don't think buying the Ti4200 would be the better deal at this point at all. Now if you bought it back in April, that'd be a different story. But, hey, you want an ATi card in the Ti4200s price range? I think there are still a few R8500s available which are only slightly slower...

As far as developers making titles that are graphics card limited on the R9700 Pro. How would that affect the R9500 Pro more than the Ti4200? Even assuming that the decent performance life of the 9500 Pro was cut down from 18 months to 12, that would mean the remaining life of the 4200 would also be cut down from 12 months to 6. But I don't believe that will be the case. For one thing you can always turn down the eye candy, and anything that will limit a high end card is going to be nothing but eye candy because the game still has to run well and look pretty good on lower end cards. And furthermore a game like AC2 hardly needs high frame rates to begin with (or a brain for that matter ).

(An interesting side note: if you actually only buy a $300 card every 2 years you actually pay the same amount as you would for a $150 every 12 months so that's actually the smarter buy if you can stand having only mediocre performance toward the end. Most people who buy high end, however, can't handle anything less than high end.)

 

[Hyp-X]

]Hyp-x

I don't know if it's good ide to reply to a flat out ATI-fan, who takes my words out of context and doesn't even understand what I wrote...

But since I'm being personnally attacked, and accused of something I'm definately not... here we go.

(Without such "arguments" the forum would be too boring isn't it ?

It has more than 50% advantage to R300, and given that R300 is rarely bandwidth limited, it can turn out to be a big advantage.

I was talking about clockspeed. (Which is evident from the previous sentence, not quoted by this guy)
Now let's see: 500 / 325 = 1,538
It's 153% or in other words it's 53% more.

Maybe there are people with different calculculators out there, which print different results of the same calculation...

No, the Nv30 does not have more than a 50% advantage.

Here he - after not quoting correctly - plays the stupid that doesn't understand what I wrote - or maybe he isn't?

Especially since there are no *shipping* Nv30 out yet.

Now, there's a point. We don't know what clock speed it will have. They may not be able to reach 500MHz. But somehow I don't think they would dared to emphasize on that number that much during their presentation, if they can't hit it.
So I assumed (like most people) that it will have 500MHz clock.

Even the trumped up Nvidia (from inside their own lab) numbers dont show an accross the board even 50% advantage. Doom-III is less than 30%,

Since when does clockspeed advantage translate to real word speed?
I actually doubt that 30% number, but to me that number is simply meaningless. The best way to handle it is to ignore it.
Because even if there is a level, a demo, a resolution, a detail setting, a build where this holds true, how it is relavant to the whole?

And They Do not have the as yet unrealeased ATI drivers that Enable Doubble sided Stencil support and other Doom-III enhancements. Easily attibuting around a 20% gain in doom-III performance.

Doubble [sic!] sided Stencil alright?
Do you by any chance know what that feature is?
Do you know that shadow volumes are mostly eighter CPU or fillrate limited?
Do you know that double-sided stencil saves on neighter of those?

Now I'm open to discussion about the other enhancements.
Sure drivers are constantly being updated and getting faster.

I think that will backfire on nVidia similarly to the R8500 launch only in reverese.

There are some tough Questions that Nvidia and their supporters need to Think about.

I wonder what those questions are... as you follow up with (baseless) statements instead of those questions.

Its pretty damn clear that the Nv30 is flat out Slower than the R300 Clock-Clock.

The P4 is flat out slower than the P3 Clock-Clock.
People still buy P4's instead of P3's.
Do you know why is that?

With a 200mhz core and ram advantage it is an average of 30% faster than the 9700pro based on pre-release info.

Since when does the NV30 has a RAM advantage?
Are you swallowing that 48Gb/s bullshit marketing bait?

If you are not, then - what's your point again?

And where do you get that "average of 30% faster" information from?
I too liked to have a product that is 30% faster on CPU limited titles - but guess what - that's next to impossible.
Even my R9700Pro tend to be CPU limited in most cases...

I think many many people are forgetting that the 30% average gain is only being achieved by the Ti 5800 *Ultra*. And its 100$ more than the 9700pro.

Hear hear the guy with inside info (4x4 pipeline), has final product naming, clockspeed and pricing information.
"That must be true - I read it on the internet"

ITs pretty Questionable wether the *non ultra* version will even be faster than the 9700pro.

I don't think it matters much.
nVidia will hype their fastest cards anyway.
But if people will buy nVidia cards it will be NV31 not the NV30.
Whoever wanted the high-end bought R300 and it will likely last until R400 arrives.

Bottom line Prediction? A crippled Nv30 with a lower clock is going to get its a$$ kicked. Or at best fall in between the 9500pro and 9700. How are they going to competatively price it?

They have 4 options:
- price according to real-world performance
- try to sell higher using their brand name (a value which fading day-to-day)
- sell it lower to regain market share (their not in that position that they need that - at least not yet)
- sell at too high prices and ensure total faliure (Matrox style)

I somehow doubt they'd choose the fourth.

 

[Iceman]

Yes, I am I am seing games soon to be released or that have just been pushing a Radeon9700pro to its limits. A game like Asheron's call 2 comes to mind. It seems that software industry is catching up. There is little room in performance for AF/AA in these games. In such cases the 10-20% speed increase is not that all interesting to me, performance still remains average.

Well I played the AC2 Beta and it was very playable (didn't look for fps though) in 1024 with max Details + 4x AA + 16x QAF on my 9700Pro.
Exept Doom 3 I don't see any Games on the horizon wich really pushes the 9700 to its limits.

 

[WaltC]

"

Well but in the end thats bad for ATI. I don't see any possibility to earn money with a product like that for such a low price.
The GF4Ti4600 costs a lot less in production and that includes chip and board.
Such a low price indicates that until now ATIs market situation did not improve at all. Reports that they again lost market share indicated that already.
I am not that optimistic for them anymore. To much wasted time already. I don't think they will have that much time again.
Once you wake up the big guy it gets a lot more difficicult. Just look at AMD vs Intel.

Problem here though...long before nVidia became a "big guy" ATI was much bigger than nVidia in the graphics-card business. Who are you referring to? Looks like to me it's ATI which has awakened.

The other problem with your analysis is...if it hurts ATI to sell such a product at this price point, it's going to hurt nVidia as much when it starts shipping a DX9 product with similar performance at that price point. Ti4600 is not really a competitor because it lacks the DX9 hardware support the 9500Pro has--so in fact, Ti4600 will have to sell for less than 9500Pro in order to find its niche until nVidia can field a DX9 competitor. If vendors try and market the Ti4600 alonside the 9500Pro this Christmas at $199, the 9500Pro will murder it in sales--because, again, 4600 is not a DX9 part, and because 9.5KPro either outperforms (with AF/FSAA) the Ti4600, or else performs right alongside it, where Dx8.1 software is concerned. For a 3D-card buyer this Christmas who is even marginally in the know, but restrained to a $200 budget--the 9.5KPro is a real "no brainer," it seems to me.

 

[WaltC]

It took 3 Radeons to get it right, but it's nice to see they made it. I feel good for ATI. I bet they must be pretty darn happy inside. NVIDIA, time to get start kick'in!

*chuckle*--Took nVidia several iterations of GeForce to get it "right" as well... (Still don't know whether the last GeForce iteration will put it ahead or behind at this point, though.)

 

[Doomtrooper]

Texture compression is still broken on the Geforce 4 according to Daniel Vogel of Epic Games...thats four generations..

 

[KimB]

Texture compression is still broken on the Geforce 4 according to Daniel Vogel of Epic Games...thats four generations..

Two.

 

[WaltC]

I'm quite impressed with the performance of the 9500 PRO. With so much less bandwidth than the 9700, I thought performance would really suffer more (as it seemed in the initial 9500 PRO previews).

I'm also surprised at the cost ATI is able to produce them at. It's still a ~100M transistor chip, after all, and rightly so due to the DX9 support. I guess their board design is much cheaper than the Ti4600, not to mention the slower and cheaper Hynix memory.

The yields issue has already been commented on--the better the yield the lower the price.

Also, the thing that's often forgotten here is that at its introduction the Ti4600 was a $400 card (if not higher in some retail areas.) Additionally, for several months the 4600/4400/4200 (although not shipped at the same time) stood pretty much without peer in the 3D market. Without close competition nVidia and its OEMs were able to price the products on a basis other than strictly competitive. IE, the cost of the GF4's should not be inferred by the asking prices for them for most of the year.

Had the GF4's been released into the sort of competitive environment the GF5's will be released into I think you would have seen much more competitive pricing from nVidia--without a concurrent question of profitability. Yep, you can charge less and still make money... (You just don't make as much, of course.)

 

[KimB]

The other problem with your analysis is...if it hurts ATI to sell such a product at this price point, it's going to hurt nVidia as much when it starts shipping a DX9 product with similar performance at that price point.

Ah, but ATI is still selling a 100+ million transistor product on the .15 micron process. Once the problems in the .13 micron process are ironed out, nVidia's mainstream and performance solutions will be much more cost-effective for them.

At the same time, I don't think that nVidia's going to be putting much emphasis on the Ti 4600. The design is, quite simply, more expensive to manufacture than the Ti 4200's design.

Ti4600 is not really a competitor because it lacks the DX9 hardware support the 9500Pro has--so in fact, Ti4600 will have to sell for less than 9500Pro in order to find its niche until nVidia can field a DX9 competitor. If vendors try and market the Ti4600 alonside the 9500Pro this Christmas at $199, the 9500Pro will murder it in sales--because, again, 4600 is not a DX9 part, and because 9.5KPro either outperforms (with AF/FSAA) the Ti4600, or else performs right alongside it, where Dx8.1 software is concerned. For a 3D-card buyer this Christmas who is even marginally in the know, but restrained to a $200 budget--the 9.5KPro is a real "no brainer," it seems to me.

Now that I have a 9700 here, I can't disagree more. While the card has promise, it's just far, far too buggy. One of the current major bugs deals with the z-buffer. Quite simply, there are some games that have blatant and obvious z-buffer errors (Morrowind is one, and I think I saw some in the Tenebrae mod of Quake, too). Apparently it has to do with a simple bug that's as stupid as just not reporting that the card supports 24/32-bit z-buffer.

Another huge problem is performance. If you visit the Rage3D forums, you'll notice a few threads dedicated solely to stuttering games. With some investigation on my own using UT2003 framerate graphs, I've come to the simple conclusion that ATI's current drivers have absolutely horrid performance compared to what they should be able to do. Put simply, the framerates of the Radeon 9700 appear to fluctuate wildly over time.

Based on this, I would take a GeForce4 Ti any day over a Radeon 9500.

 

[T2k]

The other problem with your analysis is...if it hurts ATI to sell such a product at this price point, it's going to hurt nVidia as much when it starts shipping a DX9 product with similar performance at that price point.

Ah, but ATI is still selling a 100+ million transistor product on the .15 micron process. Once the problems in the .13 micron process are ironed out, nVidia's mainstream and performance solutions will be much more cost-effective for them.

Wrong. Later, not now - you talk about two different things: the best option for the future (when 'promblems are irnoed out") and the current yields.

At the same time, I don't think that nVidia's going to be putting much emphasis on the Ti 4600. The design is, quite simply, more expensive to manufacture than the Ti 4200's design.

But right now they have NOTHING to compete w/ 9500 Pro.

Ti4600 is not really a competitor because it lacks the DX9 hardware support the 9500Pro has--so in fact, Ti4600 will have to sell for less than 9500Pro in order to find its niche until nVidia can field a DX9 competitor. If vendors try and market the Ti4600 alonside the 9500Pro this Christmas at $199, the 9500Pro will murder it in sales--because, again, 4600 is not a DX9 part, and because 9.5KPro either outperforms (with AF/FSAA) the Ti4600, or else performs right alongside it, where Dx8.1 software is concerned. For a 3D-card buyer this Christmas who is even marginally in the know, but restrained to a $200 budget--the 9.5KPro is a real "no brainer," it seems to me.

Now that I have a 9700 here, I can't disagree more. While the card has promise, it's just far, far too buggy.

Facts, please. So far, I have no problems here.

One of the current major bugs deals with the z-buffer. Quite simply, there are some games that have blatant and obvious z-buffer errors (Morrowind is one, and I think I saw some in the Tenebrae mod of Quake, too). Apparently it has to do with a simple bug that's as stupid as just not reporting that the card supports 24/32-bit z-buffer.

IIRC, Morrowind uses specific extensions, right?

Another huge problem is performance. If you visit the Rage3D forums, you'll notice a few threads dedicated solely to stuttering games. With some investigation on my own using UT2003 framerate graphs, I've come to the simple conclusion that ATI's current drivers have absolutely horrid performance compared to what they should be able to do.

Funny - and how did you calculate the amount what they should be able to do w/ better drivers?

Performance is absolutely astonishing for me.

Put simply, the framerates of the Radeon 9700 appear to fluctuate wildly over time.
Based on this, I would take a GeForce4 Ti any day over a Radeon 9500.

Hehe, it's really funny.
Do you think an older generation, more expensive but at least slower card (it's fact!) would be better choice instead of a next generation, faster and cheaper card?
Your opinion sounds really weird...

EDIT: typos

 

[WaltC]

Ah, but ATI is still selling a 100+ million transistor product on the .15 micron process. Once the problems in the .13 micron process are ironed out, nVidia's mainstream and performance solutions will be much more cost-effective for them.

At the same time, I don't think that nVidia's going to be putting much emphasis on the Ti 4600. The design is, quite simply, more expensive to manufacture than the Ti 4200's design.

The key right now, of course, is the word "once".... As of right now, the .15 micron process is much more mature [than .13] and the problems have been ironed out and the circumstantial evidence seems to suggest that, 110,000,000 transistors or not, ATI is getting some good yields.

As far as the 4600 goes, nVidia's already announced an 8x AGP version of it (at the same time as the 8x version of the 4200.) Like it or not, until nVidia can ship some DX9 hardware, the GF4 family set up on 8x AGP is all nVidia has to emphasize

Now that I have a 9700 here, I can't disagree more. While the card has promise, it's just far, far too buggy. One of the current major bugs deals with the z-buffer. Quite simply, there are some games that have blatant and obvious z-buffer errors (Morrowind is one, and I think I saw some in the Tenebrae mod of Quake, too). Apparently it has to do with a simple bug that's as stupid as just not reporting that the card supports 24/32-bit z-buffer.

Yea, I have a 9700 Pro too, which easily bested my Ti4600 for the spot in my machine at home (the GF4 finding a spot in the wife's machine.) I have to laugh at your melodramatic expression of "far, far too buggy"...*chuckle* The blatant z-buffer errors I've seen affect exactly 1 out of the 35-40 3D games I've tested this card with since I bought it (Morrowind), and the overall affect on the gameplay is extremely minor and in no way prevents the playing of the game or garbles any visual information needed to play the game to its full extent. It's noticeable at times and annoying--but doesn't prevent game play--not even close. And actually, this is a *driver bug* rather than a "card" bug, and one of such small importance I expect it will be fixed very shortly. In the other 34 + 3D games I play it rousts the Ti4600 easily and literally blows it away when FSAA/AF is employed (not just in performance, but in the Image Quality--even when I exclusively used the 4600 I ran with FSAA turned off because it's just too ugly with the 4600. ) Basically, to me, the 4600 isn't comparable to the 9700 Pro--maybe for the 9500Pro for DX8.1 software the 4600 is competitive--but certainly not for DX9 and the current reviews also show even the 9500Pro trouncing the Ti4600 when FSAA is turned on. The generational gap between R300 and GF4 is definitely showing, IMO.

Another huge problem is performance. If you visit the Rage3D forums, you'll notice a few threads dedicated solely to stuttering games. With some investigation on my own using UT2003 framerate graphs, I've come to the simple conclusion that ATI's current drivers have absolutely horrid performance compared to what they should be able to do. Put simply, the framerates of the Radeon 9700 appear to fluctuate wildly over time.

Based on this, I would take a GeForce4 Ti any day over a Radeon 9500.

Yea, I visit the Rage3D forums almost every day, and almost every day the same people bring up "stuttering" again and again--granted, there do seem to be some original posts on it from time to time--but at any rate it doesn't affect me. The only "stuttering" I've ever had with either the Ti4600 or the 9700 Pro was vsync related and was remedied by turning off vsync in the couple of cases where I had it with both cards. My 9700 Pr0 doesn't "stutter" and in fact I sometimes have difficulty determining what the people who talk about it are actually talking about. One guy on R3D is a Troll who admits having returned his 9700 Pro a few days after buying it--and yet is unable to refrain from visiting the R3D sites and making repetitive "stuttering" posts...*chuckle* Poor sap....to be so bored....

Anyway, I could never recommend the long-in-the-tooth Ti4600 over the DX9 9700 Pro--no way.... (And when I owned the 4600 I was convinced that nobody would eclipse it in 2002. Was I ever wrong about that.) I might also add here that stability is so good I'm running with MS's DX9 RC0 installed and ATI's RC0 beta DX9 drivers installed--and don't crash or otherwise suffer inconvenience.

 

[Iceman]

I can confirm the stuttering in some D3D Games (Morrowind and NfS6 for example and some others) but I could fix them with the 3D Analyzer from www.tommti-systems.de

 

[WaltC]

I can confirm the stuttering in some D3D Games (Morrowind and NfS6 for example and some others) but I could fix them with the 3D Analyzer from www.tommti-systems.de

I realize it's a problem for some people with certain hardware configurations, but it's never afflicted me--even once--and I was one who never had a "DirectX problem" in 3D with the card, either.

For that matter, though, with the GF4 Ti4600 I could always find a forum where people reported problems I never had with it, either. Based on the fact that the 9700P works so well for me in my system at home (true of the 6166 drivers & up--even to the 6228 DX9 RC0 betas I'm using now--still no "stutter"), I'm not sold on the idea that this is related specifically to the 9700 Pro hardware or software, but rather is probably related to a person's support hardware and software (eg, motherboards and/or possible software conflicts elsewhere.)

All I can state unequivocally is that a "stutter" problem has never affected me (and I bought one of the earlier 9700 Pros.)