HardOCP recently had the opportunity to sit down with NVIDIA face-to-face and discuss its next generation GPU, codenamed "GF100" which is based on the "Fermi" architecture you have likely heard so much about for the last few months. Currently NVIDIA is not sharing GF100 based video card specifics. This is NOT a product launch! This is a look into the GF100 GPU's inner workings and how that relates to gaming. The "GF" in "GF100" stands for a "Graphics" solution based on the "Fermi" architecture. The "100" denotes that it is the high-end part of the current GPU family.
The GF100 is NVIDIA's next big investment, and it is yet to be seen if it will pay off for them. The GF100 is more than just a GPU for gaming; we all know that based on the recent information that has been given. However, don't let this GP-GPU nonsense fool you, NVIDIA made it clear to us...finally…the GF100 is built for gaming.
"Following a great event in November of last year, DevGAMM Lisbon is coming back to the beautiful and sunny Cascais region to catch up with old friends, connect with game developers from around the world, hear from seasoned professionals, and have a great time all around." - DevGAMM.
A look into the sad trajectory of indie games from high successful releases to complete irrelevancy in just a few weeks or months.
That's the thing with gaming there's always new experiences to have why spend months or years playing a single game when there's a new experience right around the corner.
Indie or AAA if your building your game expecting long term player counts you'll probably be disappointed as gamers often enjoy something for a few weeks and move on only to return if it's truely a classic.
Out of all the generations I've experienced there's games from 30 plus years ago I still dust off and play like super Mario bros, earthbound, vice city and san Andreas being games I treasure and revisit every few years but I'm not going back to play a game designed to keep me engaged for months on end because it's also designed to milk my wallet in most cases.
Build a great game that people love make it playable offline and ask does it matter if the concurrent player count is under 100 a year post launch more often that not it doesn't
The price of entry is too high to take chances like I used to. Was looking at V Rising and that ranges from $50-$130 CAD. That’s a lot for an indie imo. By the time it goes on sale, the player count might be dwindling. But that’s the trade-off, I guess.
"Nordic Game has revealed the winners of this year’s Nordic Game Awards, which took place on Thursday, May 23, during NG24 Spring and streamed live from Slagthuset in Malmö, Sweden." - Nordic Game.
I'm really excited after reading what Kyle Bennett had to say.
"We have come away more excited about GF100 (Fermi) than we have ever been. The design of the Fermi architecture is very innovative and NVIDIA engineers get big kudos for thinking outside of the box, or in this case, the traditional graphics pipeline. This thing has got several magazine covers in its future and an engineering award or two for sure."
Read the part where tesselation is said to be a 'function'. There is no dedicated tesselation hardware. When nvidia showed the uniengine benchmark and it was faster than a 5870, then it was no real surprise. It also didnt tell the whole story. GF100 is a much more expensive card to make and no doubt sell, more in the region of a 5970 which of course has twice the tesselation performance of a 5870.
In some ways it is similar to the scenario when shaders became unified. 7900 v 8600. You could run both vertex or pixel on unified shaders so when a unified card like a geforce 8600 was initially shown in vertex tests, it ran ONLY vertex. Thus the entire cards shaders were employed for its maximum vertex fillrate, thus making it appear considerably faster than a DX9 generation card with a split architecture and for example a mere 8 vertex shaders running the same test. However, its just an artificial benchmark. Ideal case scenario stuff. Games arent all vertex OR pixel shaders. More like a mixture. Which (partly) explained when running games of the time then 8600 was typically slower than say a 7900GTX with split architecture, because despite the unified advantage of higher maximum fillrates, typically games would require more of a balance alongside more bandwidth of course.
ANYWAY. To round this point off in this case GF100 running this bench is running ONLY tesselation at its absolute peak. It may very well have 8 times the MAXIMUM geometry performance of GT200, much like an 8600 boasted three or four times the maximum vertex performance of a 7900. In an actual real world DX11 game scenario, the 'polymorph engines' are employed for functions besides mere tesselation. They are not dedicated hardware, they are a shader subset....Thus although the Ati design appears significantly slower, it actually has full on dedicated hardware for the function. When deployed outside of a synth bench, nvidia will not be able to boast such impressive figures. Nvidia's implementation makes the design more flexible in that they will be scalable for midrange parts. Good thing? Maybe in ease of design for nvidia, but it also means that the midrange nvidia cards will have cut down tesselation performance as 'polymorph engines' are removed for the midrange/low end. ATi cards all have the same unit and tesselation performance. ATi's design means that the tesselator is always free to do its thing, not sometimes busy being something else....
As always though, the Real world tests will be the ones that actually matter
Well this is the thing. People need to see behind the numbers. Nvidia claiming more tesselation performance in a synthetic benchmark is one thing, but in the real world of games it looks much bleaker. As i have said in other posts most of the consumer money from gamers is poured onto the midrange parts, rather than high end enthusiast parts. Midrange cards derived from this architecture will almost certainly have sliced up tesselation performance for a start, three quarters or maybe only half the top dog. But something like a radeon 5750 retains similar tesselation performance as a 5870. The other thing is if nvidia actually manage to pull out some new midrange parts. The question is when? 5750 has been around three months or so now. How long til a proper midrange DX11 nvidia part after this? June? later?? More recycled DX10 parts will not be attractive against ATi's top to bottom DX11 lineup...
GF100 is going to be massive if the rumours and indications hold true and they seemingly are to date. It will cost nvidia an absolute fortune to manufacture just like GT200. I honestly have no idea how nvidia expect to make any money with this chip. 5870 is much smaller, much cheaper to make, almost certainly wont be massively slower especially in DX11 and is actually profitable for ati. That is quite important considering they are businesses trying to make money! Nvidia absolutely must have a successful midrange and low end lineup derived from this architecture asap. They cant keep afloat with only high end chips like this, you cant balance a pyramid on its point.
http://www.youtube.com/watc...
With all this news about fermi, I wonder if its damage control.