ATI to launch DirectX 11 GPUs in seven weeks (Sept 1x)

You're seriously going to fault Nvidia for their drivers when Microsoft crashes IT'S OWN SOFTWARE 18% of the time?

I'd like to see a similar graph for XP, we all know Vista was Microsoft's idea of a bad joke. Or, at least, I hope it was.

Quote:

Originally Posted by Killamus You're seriously going to fault Nvidia for their drivers when Microsoft crashes IT'S OWN SOFTWARE 18% of the time? I'd like to see a similar graph for XP, we all know Vista was Microsoft's idea of a bad joke. Or, at least, I hope it was.

Well, the chart makes a point. Vista was a pretty big joke, but NVIDIA drivers were causing more crashes than ATI drivers... While those crashes may have been the fault of the OS and not (or at least not entirely) the drivers, it does bring to question why ATI failures were far less frequent than NVIDIA ones.

Ok, there is entirely too much misinformation at this point...

The issue with the mobile divisions processors wasn't solder joint issues... it was fault cobalt doping. The solder issue you speak of was a completely different issue, and that was a result of our board partners, not us. The faulty cobalt doping resulted in higher than normal heat fluctuation, and then compounded with a soldering method that didn't compensate for that, things crumbled into that bad situation.

The cobalt doping issue is why we took a one time expense to fix the issues. We made good on fixing the issue, so it is really point moot. We never lied about the issue, but we did argue the end reason for the issue. People act like we released a product knowing full well that would happen. That is the farthest thing from the truth, and it is really insulting that you even suggest that. The people I work with are outstanding individuals that are working to advance the graphics industry, not ruin it.

Suffice to say, this nVidia bashing is rather silly.

As for that driver bashing... considering the lack of information leading up to Vista's release from Microsoft, it is no surprise the larger driver package has substantially more crashes. Our drivers are much more complex than ATi's, which leaves more to go wrong. It also serves to reason that we, a company with 2x the market share as ATi, would have approximately 2x the number of crash reports. The fact is, the appalling number of graphics driver related issues is a clear indicator of Microsoft's failure to distribute driver code and protocol to either company in the early phases of Vista. It is a shame you didn't think of market share before posting that... I also find it very disappointing you failed to mention that that graph isn't indicative of total time, nor in any given time frame. It stands to reason that as driver issues were corrected. Oh, and not to mention, that vague pie chart really tells us nothing, since it isn't specific (We produce many many chipsets as well, which are widely used, especially back during 2007, and so mainboard driver issues would account for a large percentage of that as well)

It took thousands upon thousands of man hours to fix issues after the fact, because Microsoft supplied little to no information in the early stages of Vista driver footprint development, aside from other areas. While there is no denying our drivers initially had many many issues with the Vista operating system, our driver development teams have been working endlessly to fix them. Within the last 9-12 months, dramatic improvements have been seen, and acknowledged by the tech community. It is unreasonable and absurd to suggest that without proper guidance from the OS's creator, that a 3rd party can hope to make flawless drivers. The shame for driver failures in regards to Vista belongs solely to Microsoft, and Microsoft only.


Now, as for your opinion on chip design... the large on chip structure of GT200 was simply a starting point for the course we planned to take.

GT200 and its sister Tesla units were released during GT300 engineering, that much is true. The large chip design does have drawbacks, but the largest chip you have seen was GT200, not GT300.

We have produced a much smaller package with GT300, while ATi's chip has grown substantially.

While you can expect ATi's offering to smaller in physical size, do not expect it to be nearly as small as the previous generation. They effectively doubled the amount of shader units, which isn't even mentioning other additions to the chip. The 40nm process TSMC already has low yields, and that is unfortunate. Do not believe for one second that ATi isn't having yield issues of their own.

You are also forgetting a very important fact: GT300 is radically different in design from GT200. You are attempting to compare an architecture that used similar technologies from G80+ to an entirely new architecture, similar in no way to anything you have seen in history. While your simple breakdown of memory layout and bus design is sound, it is still very simplified. While I cannot cite specific numbers, you can expect memory bandwidth on GT300 to exceed 225GB/sec.

This grandeur that people spew about ATi has been going on since RV600. Suffice to say, they have been wrong the last 2 generations about ATi completely destroying us in performance. ATi had the upper hand in production costs last generation, but that gap is going to be far more narrow this round. ATi is riding on their RV700 architecture, we aren't. AMD is in deep financial troubles, we aren't. I certainly don't hope their product fails; I hope it is a gleaming success. Competition spurs innovation and benefits the customer most. However, please try not to throw all your hopes into a company that is the same as its competition. Both companies are out to innovate, and to profit doing it. It would be unwise to support one blindly and throw stones at the other.

In the future, if you care to bash any company or product for any reason, you should know one thing. It stands to reason to post data that is specific, being that is nature of the tech industry as a whole. Posting generic data with no real specifics is naive, silly, and only goes to point out your hatred or malice for Company A in respects to Company B.

I have proved on many occasions to be unbiased, recommended ATi for many builds. It is extremely distasteful that I am now seen as some nVidia Knight, simply in light of my soon to expire position within the company. I would hope that in the future, if you care to deface a company or a product, you provide extensive, precise, and accurate data that highlights the specifics of your argument instead of sloshing a bucket load of generic data onto the dart board and hoping your dart hits near the bulls eye.

hey you guys
i used nvidia thru the 6xxx ond up til the 7600 series -
now i have ati 4670
is there a temperature monitor program or something for this card?
nvidia had a temp / fan monitor so i could watch my gpu cook with bf2
(i actually build computers but prefer working on harware rather than the software end of things)

Tarun is right Rahja, I've got nVidia now but used to have ATI before and the lack of reactiveness and stability is tangible, although not an unsurmountable obstacle to usability. I very clearly get the feeling that nVidia's drivers, although not "bad" software, could get a lot better if programmed in a more "intelligent" and cautious manner as to asynchronous polling. ATI drivers cause less troubleshooting/complaints on online forums, while we hear quite a few on nVidia's.

Really? I've used both, and the only *Real* difference I noticed was in the UI - both are rather clunky in my opinion. Modifying settings took about the same amount of time, although ATI was a bit nicer about it. But games on both cards played as well as to be expected, considering the differences in the cards themselves.

Any problems I've had I've had with pretty much both, mostly due to user error or a motherboard manufacturer leaving out a set of cards that wasn't compatible. I've easily spent 50$ on Advil alone because of that

(All experiences here were under Linux, by the way. Haven't used an ATI card in a Windows box yet.)

http://www.theinquirer.net/inquirer/...misery-deepens

nVidia out-right lied about the entire issue. This isn't even a wild claim or some conspiracy, it's proven true by the INQ using science in the form of numerous engineers and technology.

And

http://www.theinquirer.net/category/...rlie-vs-nvidia

This is wrong in two parts. First, shrinking from 55nm to 40nm doubles your transistor count, so they can have twice the stream processors in the same die area. Secondly, ATI has included DirectX 11 features in it's chips since the HD2900 ...... 4 generations ago. ATI's chips are less than half the size of nVidias, they are each filled with millions of useless transistors because they are "DX11" transistors used for Hardware Tessellation and other features that DirectX 10 never supported, and those same chips still beat nVidia's in benchmarks.

How can something less than half the size with millions of wasted transistors beat an nVidia chip in performance? Because ATI concentrates on just graphics. nVidia has been pushing this mutli-use GPU crap for years, and it's not taken off. It wont take off either. And even if it does, it has nothing to do with gaming or performance.

ATI's been making DirectX 11 cards since the HD 2900, they've been making 40nm cards since earlier this year, and they've been working with GDDR5 for over a year. nVidia hasn't touched DX11, still hasn't began manufacturing with 40nm, and still hasn't began manufacturing with GDDR5.

And you think you stand a chance against ATI this round (58xx vs GT300)?

I find it laughable that you even cite the Inquirer for a source. After this post, I'm done even arguing this or any other similar topic, seeing where you get your "news" from. It is pointless to argue with anyone who actually cites that man (Charlie) as a source, seeing as how he is the most bias, most anti nVidia, anti physics, anti cGPU, anti everything else our company tries to do person to ever grace the universe. It's like citing a vegan for a study on meat... the logic is so totally flawed and backwards... perhaps if he took his head out of his ass for 10 seconds, he might actually be worthy of being, at best, a FOX News "reporter".

Your math on transistor count is wrong. Total die area is not directly proportional to process nor platter area. Primary and co process design is key here. You have to take into account cache levels, chip layout, access areas, and unit type. You aren't including any of that in your "math", and neither does Charlie (how interesting!). You are only thinking of area, not density. Transistor density and implementation is far more critical here compared to chip area. The doping methods for the 40nm node are radically different, and more heavy. This drastically cuts into transistor density, and that is a huge factor you have chosen to ignore (I have no idea why...)

Our company is not filled with bumbling idiots you would like to believe it is. nVidia is an extremely creative, brilliant company with the foresight and hindsight needed in this industry. The people I work with are unbelievably intelligent, and some of the best in their respective fields. We are not about the engineer a chip that loses the company millions of dollars as you suggest, that isn't competitive, and isn't a value to the customer segments we are targeting. We are also not in the business to be giving, just as ATi isn't. If you actually believe GT300 is going to be the failure Charlie paints it as, you are lost. GT300 is arguably the most powerful processor on the planet, and it will show that to you and all the critics. I will be leaving nVidia prior to its release in all likelihood, and I will get to watch and comment on its issues and successes as a third party, just like you. You can be assured, GT300 isn't some fat, inefficient bumbling beast that has no real direction. It will be the top performing card; I have no doubts of that.

There are not millions of wasted transistors in ATi's design either... you are off your rocker if you think ATi would ever waste that many transistors. ATi's chips are streamlined graphics powerhouses, with no waste. They are like a light heavyweight boxer, trim and fast. Our chip takes a different approach, and GT200 was but a test of that; GT300 will show the true side of our approach.

Here is some news for you... GT300 is so radically different from GT200 in architecture and design implementation, you can't even logically compare them. You are trying to compare a motorcycle against a car.... you don't do that... Charlie attempted to compare GT300 to Larabee... that is just hilarious and outright stupid. He is an idiot, and taking anything he says to heart is just as stupid. He is already known to be the laughing stock of the tech reporting community.

Did you know he is blacklisted by over 28 companies? No? Oh, well now you do. They won't talk to him, they won't give him any info. Why? For the simple reason that he twists and skews information to the point that it no longer is anything logical; he comes to his own deluded conclusions based on some sick belief that ATi is a good, righteous company that isn't in the business to make money...

Here is some really bad news for you... if ATi holds to that release date, and we (nVidia) continue to have yield issues with the TSMC 40nm process... guess what ATi is going to do with their prices... they are going to raise them to kingdom come. Why? Oh, because they can! With no competition, ATi will scrape up every single dollar they can to patch their already terrible quarterly reports. You think nVidia is the only company that does that? HA HA HA. You just wait... come the end of October, you will all be complaining that the prices of the new ATi cards is outrageous, just like you did with GT200 when it released. Companies are not here to please you, they are here to make money. Try to remember that before you go fanboy mode.

Aside from your Inquirer sources, all written by Charlie, is there any other actual source you would like to cite? It won't matter, I am done arguing on behalf of the company that employs me. If I was able to give you data that I have, I would. Unfortunately, I enjoy my steady paychecks and am not about to break an NDA to appease you; one who cannot be appeased as it is.


As for those that still agree with the driver report numbers, here are some interesting things you failed to know about those numbers:

• They are a percentage breakdown based on a set number of reports.
• Those reports are from an undisclosed time frame.
• Those reports include all possible driver issues, not just graphics (this means chipsets, mobile solutions, etc)
• Those reports only take what system information could collect on the crash (sound driver conflicts with motherboard driver, crashing motherboard driver, thus flagging it as nVidia instead of Creative, etc)
• Those reports do not take into account nor make any attempt at disparity between different systems (one person's PC could have generated 400 of those reports by itself if they had severe driver conflicts)

That data paints a picture of nothing but what YOU WANT TO SEE IN IT. If you are going to toss around fun data like that, at least point out the caveats in it. Posting a random chart with no background information is foolish. You just lost all credibility with any intelligent person who knows the difference between precision and accuracy.

Now, this will be my last post on this subject, as it is fruitless to continue to argue with someone with as little to no scope of the industry as you. So, bash away! Here is your chance to post all the misinformation you want, so take this opportunity and grind nVidia into the ground. But know this, I will lock any thread that complains about ATi's high cost cards when they release; this has never been the place for QQ threads, and it won't start to be.

You're behaving like I am oblivious to his bias. It doesn't matter what the source is, it provided scientific evidence, which can't be refuted. If Einstein's work was printed in Playboy, you are saying it would have been wrong. There's no logic in that statement.

Sure it's not 100% correct. It's a generalization.

I don't think it's that. I think nVidia has taken stupid gamble after stupid gamble and they're paying for it now. ATI gambled on GDDR5, nVidia decided not to. No one could guess the outcome, because sometimes there's manufacturing issues with new memory, and sometimes there isn't. Well it turned out ATI got lucky and GDDR5 was not only widely available, it was dirt cheap. GDDR5 was the major reason ATI beat nVidia. With GDDR5, they could use a 256-bit chip rather than 512-bit, which costs them far less and provides greater yields.

With this round, both chips are using GDDR5, so I think things will be closer, but I also know ATI will "win". ATI may not have the single fastest chip, but that doesn't matter. People don't look at that anymore, or care. Who cares if nVidia's chip, twice the size of ATI's, is 2% faster in benchmarks. People aren't willing to pay $200 more for that chip. People are no longer willing to pay hundreds of $$$ for 2%. You claim nVidia will have the fastest chip, which I think everyone is expecting anyways, but it wont have the best chip.

ATI has been under the impression Hardware Tesselation would be included in DX for awhile.

It will be similar in size, but different in architecture. But you have to remember, ATI can use the same architecture because they've been building DirectX 11 capable hardware since the HD2900!!! They don't need to change anything because they've been doing DirectX 11 for years and years successfully!!! Just look at the Xbox 360.

ATI could do that, but then they will alienate people into waiting until your chip comes out, and that's probably not a good thing. In fact, that's exactly why the nVidia FUD (Fear, Uncertainty, Doubt) machine is going full tilt. You're best move at this point is to convince as many people not to buy ATI's cards in September.

What ATI will do, however, is release a very inexpensive chip, because they can, since there chip is tiny and cheap to make, and they've been stock piling 1GB GDDR5 for quite some time. They will do this, because everyone who buys ATI in September 2009 wont buy nVidia in March 2010.

ATI could only lose if they released expensive cards. There is no strategy in releasing expensive cards, especially when ATI's cards are so cheap to manufacture due to their tiny die size.

because Windows 7 s released in October and it comes with DX11 and Microshaft say that Vista will be updated with DX11 aswell, however XP won't as just as with DX10 it wouldn't work on some of the back end code of XP

In the world of fantasy land, where businesses that are in the red and have an advantage don't price gouge, sure! Unfortunately, we don't live in fantasy land, we live in the real world.

ATi is in dire straights, and they will price gouge like crazy having no competition. ATi isn't holier than any other company, and any other sane company would do the same.

Oh, and as I mentioned previously... ATi's newest upper tier cards don't have the smallest die size in the world... unfortunately. Theirs got significantly larger, ours shrank. They are riding their architecture... it will work this generation, but then it's game over.

I hope for your sake, ATi suddenly becomes more interested in making people happy than profits... but the truth is, any company that does that goes under. ATi cannot afford to be nicey nice to their customers. If they alienate them, who cares? You either buy a new ATi DX11 card, or you go without. Not much of a choice...

ATI isn't in the red.

Theirs got larger? No. It's stayed the same size. Their 4890 was slightly larger because they added a decap ring. You're manipulating information...again...

You guys lost the 4800/GT200 round. ATI's cards cost half as much and still out perform your cards. You had to undersell some cards and take huge losses on others, and then go as far as to glue two PCB's together just to win some arbitrary "fastest card available" award that nobody cares about. If any company is going under, it's yours. You've lose money already, you're about to lose more money, and Intel is going to pound you into the ground with Larrabee.

When the 4870 and 4850 first came out, nVidia had nothing to compete against them with, and ATI still made them affordable for all.

The results are out for graphics shipments and market share in the second quarter of 2009 and AMD appears to rule the roost.

That's according to Jon Peddie Research - the company specializes in tracking the graphics market and that's a barometer of the entire PC market.

In the latest report, JPR says that the channel - that is to say the distribution and reseller market - stopped ordering graphics cards and made sure their inventories of existing cards were depleted and hoped the first quarter of this year would be better.

The first quarter showed some improvement but the second quarter was very good for the vendors like AMD, Intel and Nvidia. That. says JPR, means the channel is gearing up for a healthy Q3, traditionally a time for good sales.

And now for the figures. Shipments jumped to 98.3 million units, quarter on quarter, year on year.

This is the shape of the graphics chip market according to the market research company.



The company cautions that things won't return to the normal seasonal pattern until the third or fourth quarter of this year and the market won't hit the levels of 2008 until 2010.

The PC market will be helped by Apple Snow Leopard and Microsoft Windows 7, plus AMD and Nvidia will bring in 40 nanometer designs with better performance at "surprisingly aggressive" prices.

The worst is over and there is pent up demand, hopes JPR.


Article

Tell me, all knowing Kuntz, what is the platter and die area of AMD's next gen chips and the area of their current chips?

Tell me all knowing Tarun, who has a higher overall market share right now, AMD or nVidia?

Let's let you answer some direct questions for once.

This is pretty funny... you condemn one company for practicing cGPU technology from one side of your mouth, then turn around and praise another? Wow... great work there!

Larrabee is hilarious, at best. It is going to be the biggest flop the graphics card industry has seen in years, and the laughing stock of the the tech industry as a whole.

If you actually knew anything about its architecture and Intel's pipe dreams with regards to that, you would laugh too. Sorry, now I am really done with this conversation! LOLOLOLOL This is too much! I told a friend of mine who actually works for ATi's chip integration division that you said that, and he laughed too. Absolutely classic example of naivete; assuming because something sounds good in context, it will be popular and work. Communism sounds great on paper too, but we all know how it plays out in real life....

When & where did I praise Larrabee?

I think both ATi and nVdia have decent enough drivers, i own a PC with a 8800GTX and one with a HD4870 both running windows 7 x64 and ive had no crashes in the last 2 months which is impressive considering the operating system.

The only serious driver problem i've had in the last year or so has been with sound equipment, asus onboard sound chips and creative soundcards are terrible for driver support.

I think its unfair to judge a company by crash statistics, most crashes are due to the end user being retarded, the majority of computers are so full of crap like search bars, "freeware" apps and other software designed to infest a plebs PC that its a wonder they still boot up.

Also il add a +1 to the shoddy game design argument, game programmers tend not to be the most talented people, STALKER is a shining example of letting clueless imbeciles infront of a computer.

ow ow nvidia, the die size of an HD4890 is 292 square mm
and the die size of the GTX285 is 490 squae mm
and yes nvidia got a bigger share of the market, 29.2 %
and ati got 18.4
but the share of ati is going up, and nvidia is going down.
that can change, but if ati got there directx11 cars 2 or 3 months faster on the market than nvidia that share is going up for ati.
there are still people who buy a new VGA card every year
if directx11 arrives there will be die hard gamers that switch to ati.
not so much, but they are willing to pay a nice price for it.
about the larabee, dont know, maybe its ok, maybe it sucks ass.

AMD's RV800 series chips will probably be significantly bigger. if memory serves me correctly (and the rumor source i read was correct), it will have around 2000 stream processors. if there's no significant architecture change, then the RV800 chip will probably end up around 350 mm^2 at least.

anyways, i think it's unfair to criticize one company and not the other. i love my radeon cards, but they haven't exactly been bug free. my HD4850 shipped with buggy vBIOS, and would routinely get stuck in low power 3D, until i flashed it with a fixed vBIOS. both my HD4850 and HD4890 cannot handle overclocking and running dual monitors at the same time. more specifically, the entire HD4800 series cannot handle dual monitors if the vRAM frequency changes. for the HD4850, it will never leave its max frequencies if two monitors are attached. AMD managed to get around this problem in the 4890 by making the vRAM never underclock, which accounts for the 4890's high idle power consumption.

http://www.anandtech.com/showdoc.aspx?i=3573

The 5870 will likely be:

-An 8.5 " long card
-Require just one 6-pin connector
-180mm^2 die size (4870 was 260mm^2, nVidia's was the size of the Titanic)
-Just over a billion transistors

http://www.techpowerup.com/75586/Pow...s_in_2009.html

The reason is simple. ATI figures if you manufacture a really cheap chip, most people will buy it, since the costerformance ratio will destroy anything nVidia's 550mm^2 chip could ever dream of. Then for the enthusiast/performance croud, they simply start gluing cheap chips together until they have enough to win some benchmarks. And it's going to work out quite well, I really think they've outdone themselves this generation.

ah, i guess my source was really off then. however, are we sure that picture is of the RV870, and not a more mainstream chip (RV840)?

i'm quite familiar with AMD's "small chip" strategy. however, using a dual GPU solution for the very high end can have its drawbacks. the HD4870x2 seems to be completely at the mercy of drivers; if a game doesn't have a crossfire profile, it won't gain the performance benefit.

of course, i still remember the HD4850 debut, and its +105% crossfire scaling in COD4. yikes.

funny one, but kinda true.

as i dont work at nvidia or ati, i cant say anything about the cards.
as i work at a computer shop i can say this.
we sold more ati cards than nvidia cards for the last 5 months, and that's not about the cards in computers, so only the cards we sell separate
3 months ago we stopped putting 8400 and 9400 cards in cheap computers, but ati cards.
nvidia drivers = 100 MB or so
ati CCC = 40
just ati drivers = 20
just nvidia drivers = O_O not possible.

for the rest, for about the same price we have a card with the same or better performance and less driver downloading time spend.
complains so far, none.
and the big business is in the smaller cards, the amount of people buying a big ass 400 dollar/euro card is not so big.
if dell or so starts to sell your low end card, that's a nice profit.

i cant wait to see the HD5870, gt300 and larabee, will be fun to see who wins, and i hope its the best one, gogogo 3dfx(wait nvidia buyed that huh?)
(found my old 3dfx today, still works and still good for quake 3 with that blazing fast pentium2)

and rahja, maybe i should PM this, but what kind of collage have you done, working at ati or nvidia looks like a great job as i do way to much with electronics already.
and i guess its not an easy collage but i can still try

It's unknown what chip that was for, but it is assumed it was the "performance entry" chip/card at that show.

When two chips are placed inside the same MCM, it wont count as Crossfire X, so you wont need "driver support" for those cards. They will behave as one chip on a 512-bit bus. But ATI also said in another article I can't find for the life of me performance cards will come with 2 MCM's (4 chips total) in Crossfire X, that can combine with another card (8 chips total on 2 PCB's).

Although SLI/CFX is at the mercy of driver support, it typically comes out 2-3 months after the retail release of a game, depending on the popularity of the game. There are ways to force support for a game too, such as renaming your game to "crysis.exe".

To date, the only major driver issue's I've heard of were for the 4850 X2, since the card came out 5 months before ATI planned driver support for it haha.

no, i'm sure the HD4870x2 is simply treated as two HD4870. this is why you have to enable crossfire to gain the additional performance. otherwise, it will function as a single HD4870.

I don't think anyone has said otherwise???

well, you did.

The X2 cards are PCB's with 2 chips, so yes you need to enable CFX. Two chips in an MCM is like Intels Quad-Cores where they took Dual-Cores and glued them together. ATI hasn't released any MCM cards yet, since they are something new they are doing with the RV870's.

And new info from ATI's CTO just now:

http://www.fudzilla.com/content/view/14847/1/

ah, my bad.

although i'd question whether fudzilla is really trustworthy or not, what that article said makes sense. if AMD actually makes good on that September release date, the yields better be good.

Well in that particular article, they did speak with the CTO and that other dude, so what it says has to be true. When FUD posts a rumor, which they do often, they clearly mark it as such.

I'd think the RV870 would be done around mid-August since they still need to ship it out to manufactures to put onto boards, and do the final debugging/testing of them, and all that junk. Maybe we'll start seeing ES's towards the end of August, and some early benches.

Oooo looks like someone already has an ES:

http://www.chiphell.com/2009/0728/89.html

Two 6-pin connectors, what a beast.

it's longer than a HD2900XT? ouch.

What I don't get is that 2 6-pins = 1 8-pin......

And a 6-pin can easily handle 24 amps (288 Watts) yet the stupid PCI-E spec limits it to just 75 watts.....

And 8-pin adds no more voltage lines, and yet the rating magically grows to 150 Watts....

lol

Basically a 6-pin connector could power all the worlds GPU's, including the X2's.

And apparently GT300's are in their first swing:

http://www.xtremesystems.org/forums/...d.php?t=230847

I did my undergraduate as physics, and my graduate studies (doctorate) as electrical physics with specialization in particle and nano-electrical physics.

All wattage calculations are done before e- loss and before any inversion takes place. At wire vs at source is pretty different. You also have to take into account fluctuation and cooling systems on the cards. You don't go with the bare minimum, you go with a surplus to make up for resistance and inversion.

most powersupply's have multiple 12V power lines.
so they make VGA cards with 2 times 6 pins.
and an 8 pins connector is an 6 pins with 2 extra ground lines.(3 yellow, 5 black)dont know why they made that 150 watts.
its nice if you want to know how much your VGA cards needs, no connectors = max 75 watt.
1 x 6 pins = max 150 watt, 2 x 6 pins = max 225 watt, 1 x 8 and 1 x 6 = 300 watt and 2 x 8 = 375 watt.
dont know if there are cards with 2 times 8 pins, maybe that weird asus card with 2 times an GTX285 on 1 board.
they called it an GTX295 but normally thats 2 GTX275 on 1 card.
aw well, both overkill vor GW