So.. What will Nvidia Bring to Counter the R520?

Reverend said:
Hellbinder said:
Hmm.. I dont see how my speculation on the R520 is being to abitious.
I speculate that all these speculations are nothing but unimportant speculations.

Hmm, I speculate that late January is booooring except for the NFL and there is nothing better to do that Speculate about unimportant speculations.
 
DaveBaumann said:
Your thinking is a little off WRT R520.

Who me? <-------- ???

Basically the same as the R420 with improved Memory controller and SM3 with somewhat higher clock speeds and perhaps a few more pipelines is a little off?

unless its actually more than that i am a little confused as to how it could be less.
 
Hellbinder said:
DaveBaumann said:
Your thinking is a little off WRT R520.

Who me? <-------- ???

Basically the same as the R420 with improved Memory controller and SM3 with somewhat higher clock speeds and perhaps a few more pipelines is a little off?

unless its actually more than that i am a little confused as to how it could be less.

I'm thinking he was talking about Jawed's post.

Back to the topic though, I just get this feeling that NV won't have the goods to compete with the R520 come this spring/summer. I still speculate that their refresh will be a .11u based, slightly tweaked NV40 core weighing in at around 500MHz or so, which might be able to run with the X900 pro or whatever ATI decides to call the "entry level" high end part. Who knows though, the R520 might turn out to be pretty underwhelming anyway, at least compared to whats out now.....
 
Hellbinder said:
Reverend said:
Hellbinder said:
Hmm.. I dont see how my speculation on the R520 is being to abitious.
I speculate that all these speculations are nothing but unimportant speculations.

Hmm, I speculate that late January is booooring except for the NFL and there is nothing better to do that Speculate about unimportant speculations.


You ain't lying. Football is just about over and their is no hockey over here in the states...so its quite boaring. Nascar starts next month so its all good.
 
Hellbinder said:
DaveBaumann said:
Your thinking is a little off WRT R520.

Who me? <-------- ???

Basically the same as the R420 with improved Memory controller and SM3 with somewhat higher clock speeds and perhaps a few more pipelines is a little off?

unless its actually more than that i am a little confused as to how it could be less.

r420 with slightly tweaked mem controller, slightly higher clocks, same # of pipes, some shader tweaks to make dynamic branching faster.
 
BigDuke said:
Hellbinder said:
DaveBaumann said:
Your thinking is a little off WRT R520.

Who me? <-------- ???

Basically the same as the R420 with improved Memory controller and SM3 with somewhat higher clock speeds and perhaps a few more pipelines is a little off?

unless its actually more than that i am a little confused as to how it could be less.

I'm thinking he was talking about Jawed's post.

Probably, and more specifically the bit about "hybrid" VS functionality. The new internal memory topology might be the main enabling factor in bringing it up to meet (exceed?) SM3.0 requirements.
 
So, way back there at post #2 in this thread --are you saying 'never believe the Marketing guy when he promises new gee-whiz product right around the corner' or are you saying that the PS3 gpu chip is going to be not-much-of-a-tweak to NV40? Because it has to be one or the other.
 
Whats going to happen is this


r520 16x1 at whatever with sm 3.0


nv40 refresh 16x 1 at whatever sm 3.0 of course


Who ever is slower will release a new sku with 24pipe lines then the other will responed in kind
 
BigDuke said:
Humus said:
Hellbinder said:
What will Nvidia counter with technology wize?

Dustbuster 3.0b


Not too funny anymore seeing that ATI is sporting one of their own.
They do? Where.
It may be a dual slot cooler, but it isn't loud at all- have you read any reviews complaining about the noise levels?
 
First of all, doesn't anyone read anything posted here? 11u does nothing for speed, so those thinking that the NV40 will go to 11u AND have an upgrade in speed just don't know how to assimilate the facts. 11u is a money saving move only....... see speeds of X800XL......

Also, as has been stated, the dual slot cooler on the 850XT is very quiet, so there is no way it - or, for that matter any of the "other" dual slot cooling solutions by nVidia or even the 3 party developers can't be compared to that stillborn piece of crap known as the NV30........
 
martrox said:
First of all, doesn't anyone read anything posted here? 11u does nothing for speed, so those thinking that the NV40 will go to 11u AND have an upgrade in speed just don't know how to assimilate the facts. 11u is a money saving move only....... see speeds of X800XL......

Thats a poor example.

The Nv40 is currently on a straight 0.13 process. 0.11 should be pretty much nothing but an increase over that, which is implied by the 6600GTs good headroom and stock speed.

The X800s on the other hand were built on 0.13 low k thus in some respects 0.11 is worse (and it certainly is in the case of the X800XL)
 
Do you guys think that one single R520 will be faster than current 2x6800U in Sli? :?
If yes, why do you think that? :)
 
Apple740 said:
Do you guys think that one single R520 will be faster than current 2x6800U in Sli? :?
If yes, why do you think that? :)

r520 won't have the raw fillrate but i could imagine in heavy shader cases r520 being faster
 
Remember that NV can do it's next hi-end chip at TSMC as ATI will. So don't be so sure that 0.09 from NV won't come this year.
 
My point with 11u is that nVidia will get little if any added core speed out of it. Their product line is still very strong and competitive. While ATI does have higher performing models in the high end, nVidia can always counter that they, via SLI, have the performance crown. Not counting the hefty cost of intry, few can argue with this. There really isn't a good reason for nVidia to do anything. Anyone here think a 520 will blow away a SLI 6800GT/Ultra from a "best 15 benchmarks and games as presented by damn near every hardware site" point of view?
 
martrox said:
"best 15 benchmarks and games as presented by damn near every hardware site" point of view?

**Tangent Alert!**

I was thinking about this recently. Wouldn't it be nice --and not all that burdensome for the sites involved-- if the top twenty sites got together and agreed on what the "usual suspects" for benchmarking are (D3, HL2, etc). Then each was assigned a draw order, and in that order each site picked a title of their choice not in the "usual suspects" list and not previously chosen by one of the sites ahead of them in the draw order. This title would be benchmarked by them in all graphics card reviews for the next year. Next to the results in their review for that title would be a nice icon/logo representing the consortium of sites. When you click on that, it pops a new window for the page either on that site or they could as a group maintain a one page domain/site somewhere (pretty cheap) that would explain the consortium and list each site and the "usually overlooked" title they had chosen to benchmark for that year.

Once a year they get together and rejigger the list of titles/sites.

Oh yeah, and use objective criteria for which sites would be eligible to join --say any site that had published at least x (5? something reasonable) card reviews in the previous twelve months.
 
martrox said:
First of all, doesn't anyone read anything posted here? 11u does nothing for speed, so those thinking that the NV40 will go to 11u AND have an upgrade in speed just don't know how to assimilate the facts. 11u is a money saving move only....... see speeds of X800XL......

Well that's right but you have to keep in mind that NV40 does not have low-k.
So in that case 0.11 could still be a littlebit better concerning clock speeds.
If you look at the mainstream you can see that 0.11 works different for both IHVs.
ATI had big trouble with their X700XT. They couldn't reach the 500MHZ and still with lower clock speeds compared to the GF6600GT they had trouble.


I think what Nvidia might do is to shrink NV40 with a smaller process just to get the costs down. Maybe some small tweaking in the shader department. That could be enough to counter R520 because they have the option to put two NV40s on one board and still have a SLI connecter for a two card system.
Right now it's obvious that Nvidia and its partners are testing and trying a lot with SLI.
Could be some signs of what might happen in the future.
Asus for instance has a dual GPU GF6600GT card with SLI connectors on the board:
http://www.computerbase.de/news/hardware/grafikkarten/nvidia/2005/januar/nvidia_sli_4x_6600_gt/

It's a german article so you might need a translator.
 
Apple740 said:
Do you guys think that one single R520 will be faster than current 2x6800U in Sli? :?
If yes, why do you think that? :)

only in games that nvidia doesnt create a profile for, in games that nvidia does, no freakin way
 
geo said:
martrox said:
"best 15 benchmarks and games as presented by damn near every hardware site" point of view?

**Tangent Alert!**

I was thinking about this recently. Wouldn't it be nice --and not all that burdensome for the sites involved-- if the top twenty sites got together and agreed on what the "usual suspects" for benchmarking are (D3, HL2, etc). Then each was assigned a draw order, and in that order each site picked a title of their choice not in the "usual suspects" list and not previously chosen by one of the sites ahead of them in the draw order. This title would be benchmarked by them in all graphics card reviews for the next year. Next to the results in their review for that title would be a nice icon/logo representing the consortium of sites. When you click on that, it pops a new window for the page either on that site or they could as a group maintain a one page domain/site somewhere (pretty cheap) that would explain the consortium and list each site and the "usually overlooked" title they had chosen to benchmark for that year.

Once a year they get together and rejigger the list of titles/sites.

Oh yeah, and use objective criteria for which sites would be eligible to join --say any site that had published at least x (5? something reasonable) card reviews in the previous twelve months.
Actually that wouldn't be as hard to do as it sounds I bet, I think it could be arranged between at least 5-6 sites off the top of my head that are all on friendlies.

The trouble is, how do ya pick the games? A new thread perhaps? :|
 
Back
Top