geforce 6800 to be only 400MHz core, 1.1GHz ram

[Geforce4ti4200]

So, 220 million transistors, 16 Pipelines, 400MHz GPU clock, 550MHz memory clock, memory specifications, thermal power of circa 150-160W, thermal control designs, power requirements, retail product name, product briefing dates, launch dates, retail availability and 3DMark 2003 performance scores


http://www.theinquirer.net/?article=15265

[-=DVS=-]

:D lotsa double posts :eek: if ATI can manage 500mhz part with equvalent of 16 nvidia pipes then its gonna be sweet :drooling: why do we speculate when its only 3 or 2 days left :eek2: :screwy:

[Geforce4ti4200]

3 days left at least, but will leaks happen on the 13th?

[Demirug]

3 days left at least, but will leaks happen on the 13th?

4 days left for nVidia.
??? days left for ATI.

You should never trust Inquirer. Some things are right and some things are wrong.

[killermmn]

:drooling: wait :drooling:

[Fiver]

So, 220 million transistors, 16 Pipelines, 400MHz GPU clock, 550MHz memory clock, memory specifications, thermal power of circa 150-160W, thermal control designs, power requirements, retail product name, product briefing dates, launch dates, retail availability and 3DMark 2003 performance scores


http://www.theinquirer.net/?article=15265

Yes and according to theinq those are also ref boards with wires hanging off. I'm sure final clocks will be higher.

[Toaster]

Urg, so if the Ultra version is clocked at 400mhz, then at what speed will the regular one be clocked? 350mhz?
Heh clockrates are supposed to go up as time passes by, not down.

[saturnotaku]

Perhaps the tables are turning. NVIDIA has always been a company that does its graphics thing through brute force in terms of clock speeds while ATI is fast through finesse. Obviously the 9800XT is clocked much lower than the FX5950, but the two cards are comparable in terms of overall performance.

With talk of ATI going to a 500 MHz part and NVIDIA's being clocked at 475 or lower, maybe there will be a swap here, provided both cards offer similar performance.

[volkskrant]

Perhaps the tables are turning. NVIDIA has always been a company that does its graphics thing through brute force in terms of clock speeds while ATI is fast through finesse. Obviously the 9800XT is clocked much lower than the FX5950, but the two cards are comparable in terms of overall performance.

With talk of ATI going to a 500 MHz part and NVIDIA's being clocked at 475 or lower, maybe there will be a swap here, provided both cards offer similar performance.

What if Ati does both? Finesse and high clock speeds

[OWA]

What if Ati does both? Finesse and high clock speeds
Could be trouble... :)

[ragejg]

What if Ati does both? Finesse and high clock speeds

Well, I hear nVidia's doin Pert Plus and high clock speeds... or was it Pantene?
:retard: :rolleyes:

[Paul]

Perhaps the tables are turning. NVIDIA has always been a company that does its graphics thing through brute force in terms of clock speeds while ATI is fast through finesse. Obviously the 9800XT is clocked much lower than the FX5950, but the two cards are comparable in terms of overall performance.

With talk of ATI going to a 500 MHz part and NVIDIA's being clocked at 475 or lower, maybe there will be a swap here, provided both cards offer similar performance.

I can see where you're coming from, but NV40 is clocked lower because of the realities of manufacturing a 200~ million transistor part without Low-K, not because they're going for "finesse". I assure you, it's a miracle the thing even works, let alone working at high clock speeds. If they could get 500+ mhz without sacrificing the slim yields they'll already face, and without ramping up the cooling solution even more, then they would.

Depends how you look at it. You could say it's great engineering from NVIDIA to get all that working on a standard 0.13 process, or you could say ATi are better for doing a similar part with 50~ million less transistors ;)

(depending on how they count, obviously).

[Geforce4ti4200]

Urg, so if the Ultra version is clocked at 400mhz, then at what speed will the regular one be clocked? 350mhz?
Heh clockrates are supposed to go up as time passes by, not down.


anyone remember the geforce at 120MHz and the tnt2 ultra at 150+ MHz? :lol2: and do you hear those rumors how much power the nv40 takes? I think they do not want another dustblower fan this time, also maybe they realize 400MHz is fast enough to match ati and still not be insanely hot. we could see 475 on nv45 refreshes when they get their yields up. theyd need to find a way to lower the volts so 475 with less volts=less heat :retard:

[jAkUp]

Wow... 400mhz.. thats fairly low for NV... hopefully they can get more work done per cycle....

[Demirug]

Wow... 400mhz.. thats fairly low for NV... hopefully they can get more work done per cycle....

With 16 shaderpipes they should be able to do more work per cycle. ;)

[Richthofen]

Depends how you look at it. You could say it's great engineering from NVIDIA to get all that working on a standard 0.13 process, or you could say ATi are better for doing a similar part with 50~ million less transistors ;)

(depending on how they count, obviously).

Yeah but why 50 mill transistors less? You should count in that ATI is lacking a lot in the feature department and maybe still in the AF implementation too. That is where most of the transistors are being used. Dunno about NV40 AF. Maybe NV implemented some software tweaks but maybe they spend less transistors in that department too.

My position is clear. I will not support saving transistors for a crappy angle dependent AF nor will i support a company playing the feature blocking game.
If Nvidia is competitive in the performance department i will try my best to critizise ATI for leaving major DX9 features out. It is not something to be proud of nor is it something very intelligent in the engeneering department.

[Geforce4ti4200]

ati leaving dx9 stuff out? they have been fully dx9 since their 9500 and 9700 cards, nvidias fx5950 is less dx9 than the little 9700 pro LOL. however ati did leave out pixel shader 3 to save 30 million transistors. no game will use it for a long time, nvidia wasted those 30m transistors. by the time games use it, the next generation will be out, the r500

[Spags]

From FarCry 1.1 Patch


- Improved shader speed (up to two times faster) for NVidia Geforce FX cards on very high settings
- Fixed fog bug with NVidia drivers
- Added support for NVidia NV40 chipset
- Fixed realtime shadows from blended trees
- Made improvements to lighting shaders on very high settings
- Fixed bug with lake reflections
- Added support for PS3.0 shader profile


What does that mean as far as not seeing PS3.0 anywhere?

[Paul]

My position is clear. I will not support saving transistors for a crappy angle dependent AF

Should be fun in a few days then.

[GlowStick]

Here is some more brand spankin new INQ news heh

http://www.theinquirer.net/?article=15268

Dosent touch on GPU clock speed, but they repeate some modes NV40 can do.

[Geforce4ti4200]

Geforce 6800 Ultra has two power connectors and is, surprisingly a one slot card for its reference card.

It uses GDDR3 memory clocked at 550MHz but some partners might go even higher.


being single slot is a very good thing they did, seeing atis solutions have always been single slot. memory clocks arent even up to nvidia, so we could see golden samples with 1200-1400MHz ram, but may not be worth it since itll be so expensive and wont gain very much, maybe 10% if that. It being 1 slot hints the fact the core may only be 400MHz. They probably:

1. couldnt get good yields
2. didnt want 2 slots
3. didnt want a loud dustblower fan
4. saving the 475 cores for the refresh when their yields improve and they will use less volts so itll end up getting no hotter than at 400MHz

[nobie]

The MaximumPC article says the core will be 500-600 MHz, with 600 MHz memory. It's possible the early samples with the smaller cooler had to be underclocked, and it's also possible they didn't hit their clock targets...we'll know soon!

[volkskrant]

Well, I hear nVidia's doin Pert Plus and high clock speeds... or was it Pantene?
:retard: :rolleyes:
The rumor that Ati is more aggressive with memory and core speeds does not necessary imply the are less effective per clock thanthe last generation. So again.. What if they are going for both?

Valid question in my opinion :rolleyes: :retard:

[Richthofen]

Should be fun in a few days then.

as you might have read i already guess that Nvidia is going the same route like ATI in the AF department and i can tell you that i don't like that at all.

But who is to blame - Nvidia?
Certainly not. The problem in the AF department is ATI and those crappy reviewers not critizising the crappy AF enough. Why? Simply the only thing those stupid guys saw were AA quality differences and some crappy DX9 synthetical benchmarks.

I can tell you that i don't like the way it is going in the AF department but it's not Nvidia's fault. They have done a lot of things wrong but definately in the AF department they are the last one i would blame for that development.

[Richthofen]

however ati did leave out pixel shader 3 to save 30 million transistors. no game will use it for a long time, nvidia wasted those 30m transistors. by the time games use it, the next generation will be out, the r500

Same story as everywhere. If no one offers it no one will use it. The hardware support has always to be there in advance. If no one would have offered DX9 in hardware no one would have used it.
The same argument you are using concerning SM3.0 i could use about SM2.0
If ATI leaves SM3.0 out reviewers must critizise them the full way because that's what they deserve. If every company is thinking that way we can send developers to Hawai or to the Bahamas for going swimming and spending their spare time over there.They would have a lot of it.