4870x4

TheBuzzer

TheBuzzer

HACK THE WORLD!
Joined
Aug 15, 2005
Messages
13,028
Come on, why are people comparing multiple chips gfx card to single chip gfx card?
Don't people remerber what happened to 3dfx when they did that approach?

There will be cheaper single chip that is faster than multiple chips. So why are people worrying about nvidia when their chip is a single chip. Well just because their chip size is bigger but if they strink it, they should have the lead still.

Unless ati gfx chip really does something when drivers get updated.

v56k.jpg
 
Come on, why are people comparing multiple chips gfx card to single chip gfx card?
Click to expand...

Because like the majority of the people who buy graphics card, nobody actually cares if the card has 1 chip or 1000 chips on it, they base there purchase on raw performance/available funds or bang for buck.
 
I compare them because they are still a single card. If a single card performs better with multiple GPUs over a single card with a single gpu, and they both cost around the same...my choice is obvious as long as its one card and micro stuttering isnt an issue.
 
Odds are the GT200 chips are bordering on 3x the cost of the RV770 chips. Even using two chips and with an Nvidia shrink they're still likely cheaper. Also keep in mind there are some additional yield benefits that come from smaller dies.
 
Multi-chip performance is not always an consistent or reliable than single chip performance.
 
TheBuzzer, 3DFX's death had nothing to do with multi gpu solutions, (they had been doing that since the Voodoo 1 quite successfully), and everything to do with:
MONUMENTAL mismanagement, refusing to open source or cheaply license glide til it was too late, not properly embracing opengl til it was too late, purchasing STB and essentially telling all of it's AIB partners to take a hike, failure to get OEM contracts, failure to adapt to the market place. Actually if it were not for the STB purchase and several years of gross mismanagement, 3DFX might have been able to bounce back from the late to market V5. It was only one bad product cycle after all.

As to why multi chip cards are being compared to single chip cards, as other have already said, most people and sites tend to compare products in a given price range against each other. As it stands right now, you can buy two 4870's for less than one 280. So it is sort of natural that the performance of the 2 be compared.
 
The RV770 is specifically designed to be scalable. Not only did ATI design it explicitly to be used in multiple-GPU configurations, but their technology is also far superior to 3Dfx's. When 3Dfx introduced multi-GPU solutions, they were hacked-together implementations that relied on software and inferior rendering methods, whereas current GPUs are being designed with hardware capabilities for parallel scaling.

Now, take a look at the relative performance between the monolithic GT200 and the smaller, scalable RV770. The RV770 performs almost as well, and when two are combined they outperform a single GTX280. The RV770 also benefits from a better manufacturing process, much higher yields due to the less complex design, and cheaper production costs. Considering two RV770s are both cheaper and faster than a single GTX280, which one is the inferior design?
 
Karma said:
So far as you know....
Click to expand...

Uhh, exactly. While future designs may change this, current multi-GPU implementations scale inconsistently, and provide much list consistent performance than equivalent single GPU solutions. We have no way of knowing how this will change with future incarnations of this technology, but that goes with out saying to anyone with even a simian level intelligence.
 
Well it is possible to make multiple chips be fast by working together.

It might reduce the cost for the company for making all the chips the same.



Just wait till gfx card becomes like

entry level 1 gpu chip
mid level 2 gpu chip
highend level 3 gpu chip
ultra 4 gpu chip



This way the cost wont be the different chips set amount but just the board that connect the chips and mem and stuff.
 
Obi_Kwiet said:
Uhh, exactly. While future designs may change this, current multi-GPU implementations scale inconsistently, and provide much list consistent performance than equivalent single GPU solutions. We have no way of knowing how this will change with future incarnations of this technology, but that goes with out saying to anyone with even a simian level intelligence.
Click to expand...

Actually, it's already changed, on the hardware level. You just don't know it yet.
 
Damn, and I bought a Core2Duo not realizing that I could have just upgraded to a faster single core! *head slap*

Tech changes, rendering techniques need to change too. The video card manufacturers need to drive it with their tech solutions as well as the middleware people like Microsoft. Of course, even still 90% of the time my C2D/and/or/quad chip spends its life with only one core maxed with the rest sitting idle.

Later when someone figures out a better way to compute than pushing electrons around we can celebrate the return of the fast monolithic processor.
 
kllrnohj said:
I laugh at you because multi GPU cards are closer to dual CPU systems than a dual-core CPU :rolleyes:
Click to expand...

Dual-CPU systems don't scale much worse than dual-core systems. It's only when you get into higher numbers of CPUs that their scalability is limited by the bus bandwidth, and even then, with technologies like Hypertransport and the upcoming Quickpath, the buses will be able to handle quite a lot of traffic. That situation is similar to the rumours about ATI's new interlink tech, which is supposed to drastically reduce the scaling issues for the X2 card. Obviously we won't know for sure how successful it is until the card actually comes out, but it would be foolish to rule out the possibility that it really is all that.
 
multi-gpu is just the way things are going. Why build an expensive, massive, hot power hog of a gpu when you can build smaller, cooler, and more efficient smaller ones and stack 4 of them together? One of the fastest super-computers on the planet uses many crappy processors to accomplish one big, awesome result.

makes sense to me... dunno about you.
 
The thing is, ATi is using a new kind of interconnect technology that basically makes the two GPU's behave as one, nearly doubling performance (roughly 80% increase).

Traditional methods of linking two GPU's together through SLI/Crossfire aren't effective, but the direction ATi is going makes perfect sense.

Cheaper, and scalable.
 
People are just talking shit on ATi because they know nVidia's 280 won't beat the X2.
 
RADEoN said:
People are just talking shit on ATi because they know nVidia's 280 won't beat the X2.
Click to expand...

in crysis, two 4870's and one 280 are very close in performance. if everything stays the same, the new x2 will not be faster than a 280 in this game, though i doubt ati will remain idle. hopefully ati will make improvements in driver so we can really see what this new cf interconnect technology can do.
 
ATI's x2 was made to be dual gpu, and they had that in mind since the drawing boards. You cannot discredit it as not being successful or a viable solution to high end gaming by arguing that it is a dual gpu solution. It scales very well, is efficient, and is built on a simple and small process. When it all comes together, the bottom line is that if you are buying from ATI, you are getting a helluva lot more performance for 3/4 or 1/2 the price of a competing NVidia product.
 
I love when people assume that the 4870x2 is going to have the same performance as 2x 4870.

CF doesn't use the "new interconnect".... therferor, theoretically speaking, if the GPU's and RAM are clocked the same, the 4870x2 should be faster than 2x 4870 in CF.

Guess we'll just have to wait and see.
 
rennyf77 said:
in crysis, two 4870's and one 280 are very close in performance. if everything stays the same, the new x2 will not be faster than a 280 in this game, though i doubt ati will remain idle. hopefully ati will make improvements in driver so we can really see what this new cf interconnect technology can do.
Click to expand...


i really can care less what card is faster in the ill coded CRYSIS.... Crysis should not be the game of choice for testing... its bloat code
 
xdivenx said:
ATI's x2 was made to be dual gpu, and they had that in mind since the drawing boards. You cannot discredit it as not being successful or a viable solution to high end gaming by arguing that it is a dual gpu solution. It scales very well, is efficient, and is built on a simple and small process. When it all comes together, the bottom line is that if you are buying from ATI, you are getting a helluva lot more performance for 3/4 or 1/2 the price of a competing NVidia product.
Click to expand...

i don't think you're right about the interconnect issue. in the forum following the [H] evaluation of the 4800 series, kyle noted that one of the reasons they chose not to include cf tests was because the interconnect method was very different from what ati was using before. that would suggest to me that the way two ati gpus communicate with one another; whether on a single pcb or through cf bridges is pretty similar. who's discrediting anything. i can only go on what's available at hand, and until the x2's are in the market i can only surmise their performance by observing the performance of two 4870's together.
 
POPEGOLD said:
i really can care less what card is faster in the ill coded CRYSIS.... Crysis should not be the game of choice for testing... its bloat code
Click to expand...

crysis on a 280 or two 4870's is pretty cool. for the first time since the game was released, you can actually play it with much of its visual splendor enabled. both of these gpu solutions possess the raw processing power to make the game far more fluid than we've experienced in the past.
 
TheBuzzer said:
Well it is possible to make multiple chips be fast by working together.

It might reduce the cost for the company for making all the chips the same.



Just wait till gfx card becomes like

entry level 1 gpu chip
mid level 2 gpu chip
highend level 3 gpu chip
ultra 4 gpu chip



This way the cost wont be the different chips set amount but just the board that connect the chips and mem and stuff.
Click to expand...


I don't know where someone is coming off with saying 3dfx didn't have any model and just slapped chips together. That was excataly 3dfx's model with the VSA100 line of chips
1 chip voodoo4
2chips voodoo5 5500
4 chips voodoo5 6000

They have a similar plan for the next gen(rampage). Again people need to stop complaining about how many chips something has. Its bang for your buck/ price/performence
 
ApaThy_NeXT2 said:
I don't know where someone is coming off with saying 3dfx didn't have any model and just slapped chips together. That was excataly 3dfx's model with the VSA100 line of chips
1 chip voodoo4
2chips voodoo5 5500
4 chips voodoo5 6000

They have a similar plan for the next gen(rampage). Again people need to stop complaining about how many chips something has. Its bang for your buck/ price/performence
Click to expand...

mY iNTEL Q9300 is 4 chips glued together on one die.... dont hear people complaining about its not 1 chip?

Only hardcore blind nvidiots argue over how many chips are used

just dumb
 
Karma said:
Actually, it's already changed, on the hardware level. You just don't know it yet.
Click to expand...

But until this new technology is released, your point is irrelevant. I assume you're talking about the 4870X2... Which, developed or not, is still surrounded by myth and rumor dammit!

POPEGOLD said:
mY iNTEL Q9300 is 4 chips glued together on one die.... dont hear people complaining about its not 1 chip?

Only hardcore blind nvidiots argue over how many chips are used

just dumb
Click to expand...

QFT. So long as the chips work together, and work without a performance or feature hit (over a monolithic design) who the fuck cares?
 
Arcygenical said:
But until this new technology is released, your point is irrelevant. I assume you're talking about the 4870X2... Which, developed or not, is still surrounded by myth and rumor dammit!
Click to expand...

It's already been released. People just can't use it yet.
 
Arcygenical said:
But until this new technology is released, your point is irrelevant. I assume you're talking about the 4870X2... Which, developed or not, is still surrounded by myth and rumor dammit!



QFT. So long as the chips work together, and work without a performance or feature hit (over a monolithic design) who the fuck cares?
Click to expand...


Well u will see Nvidiots post ad nasuem that the 4870X2 vs the GTX280 is not fair beacuse its 2 gpu's versus one
 
xdivenx said:
ATI's x2 was made to be dual gpu, and they had that in mind since the drawing boards. You cannot discredit it as not being successful or a viable solution to high end gaming by arguing that it is a dual gpu solution. It scales very well, is efficient, and is built on a simple and small process. When it all comes together, the bottom line is that if you are buying from ATI, you are getting a helluva lot more performance for 3/4 or 1/2 the price of a competing NVidia product.
Click to expand...
well said. ati is really producing amazingly with this tech.
 
POPEGOLD said:
Well u will see Nvidiots post ad nasuem that the 4870X2 vs the GTX280 is not fair beacuse its 2 gpu's versus one
Click to expand...

They'll probably be the same people gloating about how awesome the GTX280 GX2 is when it comes out for $999 and causes brownouts in your neighborhood.

I agree with everyone who says "who cares" about the number of chips. If the 4870x2 can consistently provide better performance than the GTX280 for a lower sticker price, why would I care how many chips it uses?
 
You must log in or register to reply here.
Back
Top