A World Without Limits

[Typedef Enum]

In case you haven't seen it yet...ATI has updated their frontpage with a new Flash presentation. R300 is coming...!

The Inquirer is now reporting that ATI is set to launch a .15u R300 soon, and then bring a .13u part later this year to compete with NV30...

What I'm wondering is...what do you think nVidia will do to counter this product?

A. Bring a higher clocked, higher BW NV25 to the table...
B. Launch Det. 5 drivers the same day that Anantech/etc. do a R300 preview.
C. Announce NV30 details the same day that ATI announces R300.
D. Both A and B
E. Both A, B, and C.

 

[Doomtrooper]

F: Nvidia corporation releases free sweat bands to all employees..

Seriously I think it will be some PR release about NV30, maybe even a demo running on a .15 Micron Nv30.

 

[John Reynolds]

I don't think R300 will force Nvidia to pre-announce NV30. I hope the company's management learned its lesson back in '98 with the original TnT.

I'd go for option B. A Ti4600 with a new set of drivers could very well outperform R300 with immature drivers, at least in benchmarks like Q3 and 3DMark.

 

[Ailuros]

Hmmm I agree with John.

For me there's in a far fetched corner another option:

Z. Waiting for both products to mature.

 

[Brent]

perhaps info on nv28 with r300 launch?

who knows

will be interesting to watch, i'm sure nvidia will do something though, they always do

i am actually looking forward to Det. 5's, I want to see how much they can improve anisotropic performance with AA

 

[Randell]

Hmmm I agree with John.

For me there's in a far fetched corner another option:

Z. Waiting for both products to mature.

Ail that's your response as an informed buyer

How about announce a higher clocked 4600 with an NV30 hype launch?

 

[Ailuros]

That would fall into the NV28 possibility mentioned above, Randell. I doubt too though that NV will announce NV30 that soon.

Informed buyer? No idea. It never hurts to play safe though

What I'm waiting to see from either next generation product is if and to what degree an upgrade would be a necessity if someone has currently a R200 or NV25. Expectations might be high, but they were high for Parhelia too, although a really bad example in that case *snicker*.

 

[Randell]

I was goingto say NV28, but always get my NVidia dominations confused, glad you're here to correct me.

Looking at Parhelia - 1024x768 with 16xFAA, it is so close to being 'really' playable, but its still too pricey for me.

It looks like my 8500/XP@1500mhz will be enough on lower detail settings without AA at 1024x768. I will start saving now for a replacement at Christmas to give me higher detail/AA and Aniso (though I may have to wait for Spring '03).

 

[Edward]

I don't think R300 will force Nvidia to pre-announce NV30. I hope the company's management learned its lesson back in '98 with the original TnT.

I'd go for option B. A Ti4600 with a new set of drivers could very well outperform R300 with immature drivers, at least in benchmarks like Q3 and 3DMark.

I don't think a 4 piped card will outperform an 8 piped card, on the same clockspeeds.

 

[Gollum]

I have my doubts a .15 micron R300 @ 107 million transistors can get close to the same clockspeed a ~63 million transistor NV25 can. I believe the R300 will certainly be faster than any NV25 due to its raw power, its just that it won't achieve their clockspeeds, it will take a .13 micron part to reach/exceed those IMHO ...

 

[LittlePenny]

I don't see how nVidia can get more speed out of NV25 without "optimizing" for specific games. They are almost already at the theoretical fill-rate for that card. I think that is something special because I do not recall that feat being done with any other IMR.

 

[jb]

Well unlike past PR announcements we have had some hints of what nVs plans are. This time around nothing. Either nVidia has done an impressive job at stopping leaks or does not have anything as earth shattering as they have had in the past. But as time gets closer we may here a few tid bits.

I kind of agree with JR. And with JohnyC's comments on the E3 show we dont know if his GF4 was a special one, or just a new speed bin part so that may lend some creditablity to that. However some web sites have been under an NDA with nV30 info so I can see how taht might be a answer. We all know that there is this nV28 part and Anandtech has hinted about those before so maybe thats thier answer. Finally they did not do anything on the Parhelia launch. Maybe the knew it would not have as much Thunder as the R300?

But If I were to bet, I would probably go with B) Just wanted to give you something to think about...

 

[Hellbinder]

I'd go for option B. A Ti4600 with a new set of drivers could very well outperform R300 with immature drivers, at least in benchmarks like Q3 and 3DMark.

You cant be serious.... Have you been paying attention? Anand says the R300 puts everything else to *shame* that does not imply a slight lead...
All the information surrounding the next detonator release says that it enhnances Anisotropic filtering and thats it.....

R300 8 pipes at least 16 textures per pass... You actually think that the GF4 ti 4600 is going to be faster with a driver?????

Not only that but ATi's R300 Drivers are a far cry from beta. The card has been out there circulating around for over 2 months now.

Those comments made by JC were from an alpha card running at a lower clock with initial drivers. Clearly based on Anands recent comments things have changed a lot!!!.

 

[Ailuros]

I don't think a 4 piped card will outperform an 8 piped card, on the same clockspeeds.

Hmmm for the case you intend that comment that could be true. But since it's a wee bit overgeneralized if you place at same clockspeeds and equal DDR ram/raw bandwidth a 4 piped Tiler against an 8 piped IMR, I'm not so sure that the above comment remains accurate.

Of course both cards will be IMR's.

 

[Ailuros]

Hellbinder,

John probably based his comment on NV's past track record. Within that reasoning it's not entirely impossible.

Not that I see any sense in the amount of textures per pass at the moment since most current games are still using quad-texturing at best.

I'm not saying that they definitely will, but since NV has pulled some sudden 20+% performance improvements out of the blue more than once in the past, it's not unlikely at all.

 

[Hellbinder]

I suppose that the R300 might get clocked at 250mhz but all the leaked information that comes out week after week keeps pointing to 350mhz. Everyone keeps saying its impossible... but we'll see....

So far everything is pointing to a 350mhz chip with 350mhz memory... Regardless of how *impossible* its *supposed* to be.....

Could be the biggest heat sink on earth. or they could have smart-chip technology where it powers down areas of the chip not being used to save power and heat.

 

[pascal]

The only two comments from id about R300 are:
- JC said it is faster than GF4 in ALL benchmarks
- During E3 id wanted to use it with non compressed textures and anisotropic filtering Remenber that JC had to use compressed textures and turn off aniso to get some good framerates with GF3.

Also it has the ideal feature set for Doom3 (JC words) then IMHO the GF4Ti4600 is no competitor for R300.

 

[Ailuros]

Hb,

Rumours can get confusing and a lot of mistakes can be made from source to source. If the Inquirer is in anyway close to being right that they also plan a 13um respin at the end of this year or early 2003, then yes a 350+mhz clockspeed makes perfectly sense and puts ATI in the right position against the competition.

If it's as complex as predicted to be and has a 256bit bus then 250mhz should be the highest possible yield at 15um.

 

[Bambers]

The 8500 chip is far cooler than gf3s and 4s, thats why they can be overclocked too beyond 300 just with the crappy little black HS they sell with. Putting a blueorb (lapped) on mine barely improved the overclock at all.

 

[John Reynolds]

Off the top of my head, I could quote several examples of next-gen. hardware that was considerably more powerful and yet initially ran somewhat slower than its predecessors in certain applications. Why? Drivers.

And Carmack's comments all center around one thing: Doom 3. I wouldn't be surprised to see R300 lose to a Ti4600 in Q3 while yet outperform it in Doom 3.

In other words, lighten up. My comments implied that R300 might run slower in certain benchmarks, not that it absolutely would.