Pics of gt300

Printable View

Show 100 post(s) from this thread on one page

10-01-2009, 12:36 PM
Chumbucket843

Quote:

Originally Posted by Calmatory

Yes, because by the specs GF100 seems to be far from being superior compared to RV870. It's huge, its slow(GFLOPS wise), it will cost a ton, and there isn't cards out yet. So yeah, thats the "superior product" Nvidia is having. :rolleyes:

The only way I can see GF100 being superior is GPGPU applications. Thats where it's aimed towards. Gaming performance seems to be the secondary priority.

show me where a 1.1tflop 4870 beats a 1 teraflop 285. just because ATi's theoretical flops are higher doesnt mean the card is faster in the real world. hell, a 5870 has double the specs of a 4870 but that doesnt show. i dont see how 2.3tflops is slow either. it will be competitive as a gaming card considering it has 2x more shaders, bigger caches and faster memory.
10-01-2009, 12:38 PM
Clairvoyant129

Quote:

Originally Posted by marten_larsson

Can't say I'm sorry. A good joke is a good joke. Also, since the card was not the real deal and still was showed by the CEO I think it deserves some bashing. Had it been ATi that did the same thing we'd be hearing this from some of you instead (well, saaya probably would have been doing the photoshop anyway).

Why is it not real? He EXPLAINED THAT that it is a TELSA unit, not Geforce and that Geforce will come later :rolleyes: It's still the same GPU but for a different purpose. That means it's not the real card?

Also they did a physics demo with Fermi vs a GTX 280.
10-01-2009, 12:38 PM
initialised

I don't like the 6+8 pin connectors being at 90 degrees from each other, it will spoil the aesthetic. Also does it smell as good as a first gen GTX295?
10-01-2009, 12:40 PM
saaya

gt300 will be expensive, but nvidia did hint they will have several cut down chips quickly at or after launch...
so if the gpu works, and works efficiently, they can cut it down to the right price and perf points and compete with ati very well...
i hope thats going to happen cause thats what really speeds up innovation AND it reduces prices for us customers :D
10-01-2009, 12:41 PM
Manicdan

Quote:

Originally Posted by Clairvoyant129

Why is it not real? He EXPLAINED THAT that it is a TELSA unit, not Geforce and that Geforce will come later :rolleyes: It's still the same GPU but for a different purpose. That means it's not the real card?

Also they did a physics demo with Fermi vs a GTX 280.

we think its not real cause so far everything looks like they put a fake cooler onto a larger PCB, the cut the PCB to make it fit. basically its a mockup of what they expect the production to look like
10-01-2009, 12:43 PM
saaya

gt300 will be expensive, but nvidia did hint they will have several cut down chips quickly at or after launch...
so if the gpu works, and works efficiently, they can cut it down to the right price and perf points and compete with ati very well...
i hope thats going to happen cause thats what really speeds up innovation... AND it reduces prices for us customers :D
10-01-2009, 12:46 PM
marten_larsson

Quote:

Originally Posted by Clairvoyant129

Why is it not real? He EXPLAINED THAT that it is a TELSA unit, not Geforce and that Geforce will come later :rolleyes: It's still the same GPU but for a different purpose. That means it's not the real card?

Also they did a physics demo with Fermi vs a GTX 280.

Have you looked at the pictures? A sawed off PCB is probably not what I would want to run in my server if I had one. So I'm sorry if I'm not overly enthusiastic about this. I believe the GT300/G300/GF100 or whatever it will be called will outperform the 5870. That has been my projection since early this summer. But I doubt it will hit retail before Xmas.
10-01-2009, 12:47 PM
Katanai

Quote:

Originally Posted by Calmatory

It's huge, its slow(GFLOPS wise), it will cost a ton, and there isn't cards out yet. So yeah, thats the "superior product" Nvidia is having. :rolleyes:
.

Kind Sir, would you please lend me your crystal ball ? I've got some prospects of my own that I would like to confirm. Like if this forum will develop over time into an ATI clubhouse...
10-01-2009, 12:49 PM
jaredpace

I want the hollow model for like $4.95, I could put it in my PC and just use onboard video take to LANs and let everyone look throw my case window at it and drool. I bet they'd sell more of these than g92 key-chains.
10-01-2009, 01:14 PM
gamervivek

Quote:

Originally Posted by highoctane

A joke here and there is funny for sure but every Nvidia related thread gets run into the ground from the constant trolling.

I don't think it's any worse than a CEO trolling us with a mockup.And I don't even get how they are trolling? They called on the BS that was in jensen's hands and are having some fun at it's expense.:shrug:
10-01-2009, 01:19 PM
Calmatory

Quote:

Originally Posted by Katanai

Kind Sir, would you please lend me your crystal ball ? I've got some prospects of my own that I would like to confirm. Like if this forum will develop over time into an ATI clubhouse...

Read my another post.

3B transistors for what, 2.3(?) GFLOPS SP performance. Though the DP performance saw HUGE boost, some 8x the DP performance of GT200.

rv770 was 256mm² with 956 M transistors.
gt200b was ~480mm² with 1.4 B transistors.

956/256 = 3.73 M/mm²
1400/480 = 2.91 M/mm²

3.73/2.91 = 1.28.

Now, that means that RV770 packs 1.28 times more transistors to a defined area.

rv870 is ~330 mm² with 2.15 B transistors.
GF100 is supposed to be X with 3.0 B transistors.

3/2.15 = 1.39 = 39 % more transistors in GF100 than in RV870.
2150/330 = 6.51 M transistors per mm².

Now, if Nvidia managed to get the transistor density of RV870, that would mean that GF100 would be: 330 * 1.39 = 457 mm². Which is bigger, gives 39 % less dies per wafer, and increases the defect rate by 39 %. This means LOWER yields, and thus higher prices.

Let's assume that the transistor density ratio remains the same this round aswell:

457mm² * 1.28 = 584 mm². See. Things get even worse here.

That is how I prove the fact that GF100 WILL BE BIG. Also 384-bit MC requires more wiring which means more complex PCB, driving up the cost.

Of actual game performance I can't say anything. The GFLOPS difference(based on rumours on GF100's part) between RV870 and GF100 is higher than last gen, in favor of RV870, so based on that, it also favors RV870.
10-01-2009, 01:29 PM
LordEC911

Quote:

Originally Posted by Chumbucket843

show me where a 1.1tflop 4870 beats a 1 teraflop 285. just because ATi's theoretical flops are higher doesnt mean the card is faster in the real world. hell, a 5870 has double the specs of a 4870 but that doesnt show. i dont see how 2.3tflops is slow either. it will be competitive as a gaming card considering it has 2x more shaders, bigger caches and faster memory.

Where are we getting that each shader can do 3ops per clock?
From what I have read, the MUL is gone, and it can do 2ops per clock.
Depending on clocks it will be under 2TFlops for SP and under 1TFlops for DP, somewhere around 600-900GFlops DP.

Can anyone point out where they are getting 3ops per clock for GF100?
10-01-2009, 01:34 PM
highoctane

Quote:

Originally Posted by initialised

I don't like the 6+8 pin connectors being at 90 degrees from each other, it will spoil the aesthetic. Also does it smell as good as a first gen GTX295?

That is how the current Tesla cards power connectors, this probably won't apply to retail units.
10-01-2009, 01:35 PM
Calmatory

Quote:

Originally Posted by LordEC911

Where are we getting that each shader can do 3ops per clock?
From what I have read, the MUL is gone, and it can do 2ops per clock.
Depending on clocks it will be under 2TFlops for SP and under 1TFlops for DP, somewhere around 600-900GFlops DP.

Can anyone point out where they are getting 3ops per clock for GF100?

http://forums.nvidia.com/lofiversion...hp?t73056.html

At least that says TPU can do some operations per cycle, which adds one FLOP to the MAD, which counts as two FLOPs alone. Marketting talk.
10-01-2009, 01:37 PM
LordEC911

Quote:

Originally Posted by Calmatory

http://forums.nvidia.com/lofiversion...hp?t73056.html

At least that says TPU can do some operations per cycle, which adds one FLOP to the MAD, which counts as two FLOPs alone. Marketting talk.

Yes, that is for G200, it has a date of 2008 on there.
People are thinking that GF100 has the same setup as G200?

Edit- Found the whitepaper where I read it. http://www.nvidia.com/content/PDF/fe...Whitepaper.pdf

SP-512FMA ops per clock
DP-256FMA ops per clock
10-01-2009, 01:38 PM
flippin_waffles

Quote:

Originally Posted by Katanai

Kind Sir, would you please lend me your crystal ball ? I've got some prospects of my own that I would like to confirm. Like if this forum will develop over time into an ATI clubhouse...

Im not so sure about the ATi clubhouse. Consider that AMD's official response on the Batman AA issue was locked long ago, while nv official response is fron and center on the main forum, for everyone to see. ;)
10-01-2009, 02:13 PM
DilTech

That was because there didn't need to be 2 threads on the same thing, considering the discussion continued from one thread to the other flippin. That was a decision by the site owner.

This weekend I'm going to be doing a major site clean-up, forming a single thread for all info on the GT-300 will be on the list so there won't be 50+ threads on it and locking the rest. There won't be any favoritism on my part.
10-01-2009, 02:39 PM
illectrik

Why don't you all just stop freaking out and have some patience until it actually releases?
10-01-2009, 02:40 PM
LordEC911

Quote:

Originally Posted by illectrik

Why don't you all just stop freaking out and have some patience until it actually releases?

LoL. Not possible. We are freaks. :ROTF:
Enthusiasts and patience usually do not go hand in hand.:welcome:
10-01-2009, 02:44 PM
flippin_waffles

Quote:

Originally Posted by DilTech

That was because there didn't need to be 2 threads on the same thing, considering the discussion continued from one thread to the other flippin. That was a decision by the site owner.

This weekend I'm going to be doing a major site clean-up, forming a single thread for all info on the GT-300 will be on the list so there won't be 50+ threads on it and locking the rest. There won't be any favoritism on my part.

Of course, just pointing out the irony. However it doesn't make alot of sense to close either, and have one side of the story not being told. But that's the way the cookie crumbles. That's ok, I for one will happily debate their side if that's what it is going to take. :p:

/off topic
10-01-2009, 02:49 PM
Vit^pr0n

Quote:

Originally Posted by illectrik

Why don't you all just stop freaking out and have some patience until it actually releases?

Sadly Nvidia's snail-like pace is making everyone here a little gitty. I think the forum will be like this for about 6 months when the card actually releases :rofl:
10-01-2009, 02:55 PM
BababooeyHTJ

Quote:

Originally Posted by flippin_waffles

Of course, just pointing out the irony. However it doesn't make alot of sense to close either, and have one side of the story not being told. But that's the way the cookie crumbles. That's ok, I for one will happily debate their side if that's what it is going to take. :p:

/off topic

OMG, a Nvidia driver thread gets posted here that has some changes that might be considered newsworthy and it gets moved quick. Someone posts a thread for AMD's monthly driver in the News section and it takes forever to get moved. How long did that thread stay in the news section touting like 100 points in Vantage and a couple fps in the RE5 benchmark? :ROTF: I'm sorry I love this forum but it is the ATI clubhouse, at least in the news section.

Edit: It never got moved.
10-01-2009, 03:15 PM
gamervivek

Quote:

Originally Posted by illectrik

Why don't you all just stop freaking out and have some patience until it actually releases?

that's like keeping your heartbeat in check while watching a striptease.:p:
10-01-2009, 03:55 PM
Chumbucket843

DIE SHOT!!!!
http://www.theinquirer.net/IMG/304/8...jpg?1254387021
you can clearly see the difference from the smaller bus and all 16 tpc's. this thing looks complicated.
10-01-2009, 04:55 PM
GTSRboy

Quote:

Originally Posted by Bobbylite

ouch @ the power connector on the end of the card

@ the end of the card would work better on my computer since right now the cables are jammed up against the sidewall of my case

Show 100 post(s) from this thread on one page