TechCombo
Technology News, Innovations, Opportunities and ReviewsThe Future of Graphics Cards
With the next series of graphics cards just around the corner there is a lot of speculation on what specifications they have and how good they will be. In this article I aim to fill you in on the specifications of both sides and what is right from wrong.
First to release their cards will be ATI and that is no surprise as nVidia have been shamelessly re-branding and failing to produce any ground breaking products i.e. the numerous failed attempts to put the GT200 core on a 40nm die. Where as ATI have already put a 40nm card out there in the shape of the RV740, even if there has been a huge shortage of 40nm wafers. There are a lot of reasons why ATI will be the first to the post, ATI have already, as I said above, got a 40nm card, they also have full DX10.1 which is only a stones throw away from DX11 and they also have a lot of experience with GDDR5.
In comparison to nVidia, who hasn’t got any 40nm desktop cards or any with GDDR5 and a major ice breaker is the fact that their DX10 is a very stripped down version, more like DX9.d. This brings a lot of issues for nVidia as their GT200 Die has been labeled a “monolith”. With more shaders comes more transistors and with more transistors comes bigger die area so the GT300 will be even bigger unless they can pull something magical out of a hat…or maybe defy the law of physics.
ATI are set to release in later this year, whereas nVidia are looking to be releasing in the middle of 2010, oh dear. Below I have made a table showing what the cards will most likely consist of. Enjoy.
Specifications Table
Now these figures are speculation but they are realistic figures compared to what I’ve seen some people claim on forums. One site which are so way off it’s unreal is this one. Claiming that a HD5870 will have 3200 SP + 80 TMUs comprising of 1 master chip + 4 slaves. Who comes up with this stuff, seriously. The big tell tale signs are, there is only 1 Xfire connector and the card is way to short (looks around 6 inches) But enough of the fan boy’s idiotic dreaming, onto the depths of the cards. Now given the table above is correct that is a lot of horsepower, no? But if we look at nVidia’s GT200 core, you’ll find that it only uses 4% – 10% of that core for GPU functions, the rest are be used for CPU functions, why nVidia did this I don’t know but it is a monumental waste of power, they seem to be hell bent on making the GPU take over the role of the CPU. But don’t think it’s over, oh no, the GT300 only uses ~8% of it’s core for GPU related tasks, there is no point in it, let the CPU do CPU tasks (apart from physics) it’s not like they can’t handle it specially if there are 8 cores on their way.
ATI on the other hand are slightly better as their RV770 uses ~20 of it’s core for GPU tasks, O.K granted it’s still not much but with ATI it’s understandable, the whole fusion project and all, which is a good idea in theory, putting a small low end GPU in the CPU to help speed communication between GPU and CPU is a lot better than making a all purpose GPU.
Now one thing I know most enthusiasts are wondering is Crossfire, or to be more specific, when will it scale after 2 cards. Well your answer is with the HD5 series, ATI have said they are spending a lot of time on drivers and making sure the cards scale well in 3 or even 4 card configurations, so not all bad news, that is until we come to nVidia. nVidia are changing their SLI and so far reports have said it’s not as good, but as this is very early nVidia have all the chance to get the 15% improvement they want, and I hope they get it.
The last thing I’m going to touch on in this article is the cooling of these soon to be beasts. Now with current cards struggling to keep the cores cool and be quiet your gonna have to lose one thing more than another, ATI seems to favour low fan speed over heat, where as nVidia prefer slightly higher fan speed and cooler core. But to combat this ATI have said they will be using a new cooling solution, now I’m guessing it’s a more advanced version of Sapphire’s Vapour-X stuff, but only time will tell. As for nVidia well my guess is they are just gonna throw more heatpipes and copper at it.
Well that was my first article, hope you liked it, please tell me if you did or didn’t below in the comments box. Till next time.
Impressively in-depth for a first article! I look forward to more!
Interesting, but purely speculative.Â
I would also like to know where did you come up with the figure that Geforce uses between 4 and 6% of its core for GPU related tasks and ATI 30%. Those numbers appears just plain wrong.
Nope, not a fanboy, my last 3 grpahics cards have been nVidia and the only ATI card i’ve had died on my withing the first month.
As for the info “Nvidia Investor Relations tells people that between five to ten per cent of the GT200 die area is dedicated to GPU compute tasks. The GT300 goes way farther here, but let’s be charitable and call it 10 per cent. This puts Nvidia at a 10 per cent areal disadvantage to ATI on the DX11 front, and that is before you talk about anything else. Out of the gate in second place.” at http://www.theinquirer.net/inquirer/news/1137331/a-look-nvidia-gt300-architecture.
Will change the 30 to 20% as i was a tad out.
You’re using the Inquirer as any sort of reliable source of information? Wow.
GPU compute likely delves directly into DX11’s “compute shading” capability, and likely has a set transistor count budget. If we said that DX11 compute shaders needed ~30MM transistors, than it would make up (by area) a smaller percentage of a BIG die (say, 300MM total) versus a higher percentage of smaller die (say, 180MM total.) Get the picture?
In reality, it looks like you’re just paraphrasing stuff you ripped off the Inquirer, and Charlie is NOT an insider. He throws mountains of fecal material onto the wall, and claims victory on whatever tiny percentage actually sticks.
He knows about as much about future graphics capabilities as my 60 year old mother.
I didn’t know they were a bad source, and so won’t be using them again. We all make mistakes and since this was my first article then I expect it to be full of holes.
Got any suggestions for reliable sources?
A good place to start is with these guys: http://forum.beyond3d.com/forumdisplay.php?f=13 (the 3D Architectures and Chips forum)
When it comes to unreleased hardware, the reality is that *nobody* actually knows. There are a few large, annoying, and loud players on the internet who like to spout wild claims, but the only people who DO know aren’t allowed to say, and the people who don’t know but like to make stuff up will only tell you that they have “their sources.”
The reality is, sources do leak, but how do you separate the legitimacy of a single leak amongst a flood of hyperbola? If people would actually spew less nonsensical crap, we’d probably actually know MORE about the product before it got to production, versus having 37 versions of someone’s fantasy with some minutiae that ends up being semi-correct.
I also noticed that you removed my prior post correcting about a half-dozen things you got utterly wrong, and yet didn’t correct those factual innaccuracies in your editorial piece either.
Damage control?
The owner did that not me, I read all comments. As for damage control well, I’d have to re-write the article which at the moment i don’t have time to do, but will do in the near future.
I also want to thank you for teching me (in a way) even if at the start you were a bit harsh.
Well I will say good job on your first article. Just be sure to get your facts straight.
Dear webmaster: I conclud u r a dumb@ss motherfocker stoopid sh1t. I despise you for the racist things you say in your blog!! LONG LIVE THE TURKISH FEDERATION!!