Printable View
I'd like to see that. Making AA "free" would be a great idea. I'm still surprised that the XBox360 is able to do that with as simple a thing as 10MB of EDRAM in a "daughter" die. Speaking of which, wouldn't it be easier to take the same approach on a PC GPU?Quote:
Originally Posted by brandinb
I think I remember somebody saying a daughter die wouldn't work well in a PC, but damned if I can remember why.
LOL.. LOL.. (big breath).. LOLQuote:
Originally Posted by FghtinIrshNvrDi
if there is any truth whatsoever to that pie chart, its certainly profit, and not revenue. nVidia/ATI/Intel sell millions and millions of low end X1300/7300/integrated... the revenue of that makes up the huge majority, alongside with "mainstream" cards like X1600 and 7600.
X1900, 7900 are tiny minority. You know, not everybody has $$$ to buy the latest greatest all the time.
I absolutely assure you that the profit margins on high end cards are not small as you suggest. These high end cards are what finances all the R&D that went into the technology.
Think about it. 7900GTX, 7900GT, 7900GS.. very similar PCB, very similar components, very similar memory chips, same GPU, but drastically different prices. Yes, they are indeed making a big profit on 7900GTX, and correspondingly much smaller profit on 7900GS. When you're buying these high end cards, you're supporting the development of the technology and the industry.
Hmm, from what I have understood from the Vista design specs is that the new cards will have to deal with something no other video card has dealt within the past; 3D acceleration at the desktop, 24 x7. And, the current Vista design for Aero is 9.0c. So, what we are seeing is truly a dual level video card much like the current cards but instead of a seperate 2D and 3D clock, we have a seperate 3D desktop and 3D gaming GPU. The desktop unit is lower spec'd (aka 80 nm 7950 G73 core) with 256 MB of slower, cooler DDR3 and the second card is the DX10 ready part with the unified architecture and GDDR4. There could be some offload of function to the lower end part while in gaming mode, but most of the work will be accomplished by the main GPU.
As for the rumored spec, well, the 7950GX2 is a 1 GB card, right? More like marchitecture. So, maybe 1.5 Ghz total (1 GHz game core and 500 MHz desktop). Who knows.
New stuff from vr-zone:
"News : GPUs & Graphic Cards : ATi R600 To Clock Over 700MHz?
R600 will have high frequency and a 64 unified Shader units. ATI hopes to get the final chip at a frequency higher than 700MHz. At this point people don’t know the exact frequency as the final silicon is still not out but the estimation targets at more than 700MHz. This chip is 80 nanometre and scheduled for January/February time. You should be able to buy one in Q1 2007 and it will probably come shortly after Vista. "
its so incredibly plausable and feasable and practical in comparison to the earlier "G80 leak", that it boggles the mind.
HeavyH20,
how many OpenGL games do you play? Vs, how many DirectX? Since many people often play DirectX games, and seldom OpenGL games, why dont manufacturers make a seperate powerful DirectX core and a second less powerful OpenGL core?
This is similar analogy to yours with 2D, 3D desktop, 3D games.
1. Regardless if you're using Aero or not, 2D is still required. What is the GPU using when computer is booting up? What about if you use non-Vista Aero? What if you use a different operating system? What about virtual PC or X-Windows?
2. If your objective is to make as efficient high performing chip as possible, on as small die as possible so you can get as high profit margins as possible, why would you make redundant logic? Whats the point in that? What purpose would having separate 7300 logic for Aero desktop be?
3. No doubt you have played a wide variety of games. And probably monitored the temperature with rivatuner hardware monitor, and noticed that in some the temperature rises higher than in others. Especially if you try old school games like NFS3, or UT.
4. I have incredible faith in the competence of Microsoft programers. When just sitting idle at the Aero desktop, presumably minimal processing is being done.. just the contents of framebuffer being outputed. So, I think 95%+ of the time - the time between your keystrokes, and between your mouse clicks, there is little power consumption. For it to be otherwise, would be insane.
Regardless of what crazy rumours you see "Bob with the tinfoil hat" publishing about 1.5Ghz or mutli GPU or whatever, I think fundamentally things will be evolutionary. Thus, new video card will still look like a video card, will have output ports on the backplate, a single GPU in the middle, surrounded by memory chips, continue to use existing GDDR3/GDDR4 standards, and 8/12 layer PCB manufacturing techniques. The new GPU will still be able to run old school games, draw pixels as before, move vertexes, clip the planes, run simple fur shaders etc.. And even existing ALU's, ROP, etc, will still be there. I think the drastic difference (that perhaps is escaping people's minds), is how these logical units in the GPU will be arranged, and most importantly, the flexibility with which they will be able to work.
crude example:
Radeon - 3 TMU per pipeline (6 together)
Radeon 8500 - 2 TMU per pipeline (8 together)
Radeon 9700 - 1 TMU per pipeline (8 together)
X1900 - more pixel processors than TMU in 3:1 ratio.
progress is evolutionary and gradual. Advancement (performance) comes not necessarily because of huge brute force approach, but because of better more optimal for workload design.
EDIT
agree 100% with you on everything you said Alexio, EXCEPT your argument that 7900GTX is much more expensive to make than other 7900's.
I think this has to do with the subject of this thread so I'd like to comment on this.Quote:
Originally Posted by ***Deimos***
What you say is not true. The 7900GTX is so much more expensive to make than the 7900GS and 7900GT, the last two are about as expensive for a good reason. Only the higher binned GDDR3 and the bigger amount of it is so much more expensive. I bet that The total cost of the memory on the card at introduction was atleast 3 times as high on the 7900GTX. And if you talk about PCB and it's components. The 7900GTX uses like twice the power of what the 7900GT uses at stock. Go figure.
Profits in graphics sector are made in the midrange and low-end region, I really think this is the truth. So yes, ATI and Nvidia can make dual-GPU or dual-die cards, the last being more feasable. My problem with those specs are the 200W heat output, the 1.5GHz scaling, 700M transistors and the seperate AA die. Let's discuss them point by point.
1.)200W heat output. They could do it with watercooling, if they really wanted to. Still the card would run very hot and probably it would be noisy too. Fact is that it's cheaper to make the chip(s) bigger and put them on a lower vcore. Watercooling on large production scale is very complicated and expensive. As I said, more transistors and a lower vcore would be the better option.
2.) 1.5GHz scaling. A pointless remark in a spec list. Why? The Netburst architecture could in theory scale to +-10GHz. This remark is very unprofessional and makes the legitamacy of these specs doubtful. It sounds like something from the marketing department.
3.) 700M transistors. Again, they could do it. However, many of these transistors wouldn't be used for anything else than heat dissipating. Just like the G70 that has +-10% more transistors than the G71. Scaling down the clockspeed by a lot would be more economical. This rules out the possebility that both point 1, 2 and 3 can be true at the same time.
4.) The seperate AA die. The competition with ATI is big. No-AA benches are still very important in reviews. Unless this chip is also efficiently capable of doing vertex and pixel operation it wouldn't be a smart idea to slap such a chip on the card when power problems are so big already.
About supercomputer chips that dwarf the 700M G80.
These chips are made for best performance/Watt. Companies that use these CPU's look at performance/$ during the lifespan of the supercomputer. We only look at performance/$ the hardware costs. Vcore on these chips is scaled up untill watt:Mhz is only a little over linear.
Frequency scaling on chips is much more predictable and consistent if you keep to this rule than when you actually "push" the circuit. This is why it is feasable to make bigger chip for these super computers. Try it out with your Conroe at 1v. It's maximum frequency will be within a few % from others people's results at the same vcore. At 1.5V the difference is often huge.
Some about doubling of the trasnsitor count when changing to a new DirectX version. This has happend in the past yes. At that time fabrication processes were mving very fast. We are looking at only 90nm-->80nm. That is a very small difference. It's better to look at die sice in mm2. The difference from one generation of DirectX to another is much smaller then.
My conclusion: The specs in this list can not all be true at the same time. An example: Some people can still be a Christian while they agree that some things in the Bible are certainly not true. I don't. The same goes here.
I hope this thread doesn't get closed again. If I'm pushing it mods, just say so and I'll edit it untill it's okay.
Quote:
Originally Posted by ***Deimos***
1. Of course, 2D is required at the desktop if you are not running the Aero interface. Otherwise, if you want to run at higher resolutions (1920 x 1200 and above), the minimum requirement for Vista is a discrete 256 MB video card that is DirectX 9.0C compatible. You also have to run it at 32 bits per pixel. So, a older card like the 6800 GT just makes the cut to run the Aero desktop with a good sized monitor.
Running VMware or Virtual Server does not use the Aero desktop. It is limited to a desktop interface for workstations. So, you can run the non-Aero virtual desktops/servers within Aero.
2. What is easier? Putting a low end GPU on the board with 256 MB of some older 128 bit GDDR3? Cost for that, about $25. Marginal at best. And, what is easier? Creating a driver set to support DirectX 9 legacy pipelining on a cell based GPU or simply using current, inexpensive, well defined GPUs already available? It also serves to create a 3D environment to support Aero that does not produce wads of heat or use much energy.
3. Of course different games will heat up the core uniquely. If you use an HDR based game, it works more of the GPU than Pong ever would.
4. Microsoft sets the API, which in this case, is NOT evolutionary, but a whole new API. That's why there will be no DirectX 10 support on XP. Doesn't flow backwards, unlike prior DirectX implementations. In fact, the DirectX moniker was considered for obsolecence. And, each vendor can interpret and create a solution to meet the DirectX 10 requirements any way they see fit. ATI will likely tweak a driver and NVIDIA will throw hardware at it.
The dual GPU for the 8800 has been talked about for the past year. So, this is not really anything new. It is still all conjecture, but, in some ways, makes sense. More so than a 1.5 Ghz core clock. There have also been reports from the power supply vendors to allow for up to 300 watts to support these new video cards.
HeavyH20,
please go on ahead and believe what you want to believe. Dont let me or any other realists stop you from believing dillusions of 1.5Ghz GPU's, 300W video cards, and all sorts of contrived multi-chip setups. I think I've already clearly explained my point of view, and you're entirely allowed to present your view and what makes sense to you.
The 8800GTX was supposed to be about 700MHz core clock. I don't know how 1.5GHz clock got spread in the first place. The point is, that's absolutely rediculous.
There's quite a tizzy over Nvidia PR rep Mike Hera's partial interview with golem.de in which he states:
(translation from german by _xxx_ @ B3D)
+1 for the rumor specs imo.Quote:
As we're nearing the release day of a new product, the more rumours are making the rounds. This goes especially for GFX-chips, in this case nVidia's upcoming G80 chip. It will probably be capable of physics calculations as well, thanks to it's completely reengineered architecture.
This was announced by nVidia's PR-chief Michael O'Hara in one of the upcoming interviews by the colleagues from Golem.de and helped to clarify (some of) the rumours which made their rounds in the last couple of days. The initiator was a press release from the mainboard and GFX-card manufacturer ASUS about their „P5N32-SLI Premium“ mainboard.
The mainboard features the „nForce 590“ chipset for Intel-platforms and offers three PCI-Express 16x slots for GFX-cards. According to the PR the thirs slot should host an nVidia physics card. This statement yielded rumors about nVidia developing their own physics accelerator card. This seems to be a misuderstanding, as M. O'Hara explained to Golem.de.
So he said that nV does not develop their own physics card, but works on something else at the moment in which they have to slowly grow into. Further he said that with the next gen and the one after "the possibilities for physics calculations will be very good". This would be a reference to the upcoming G80 chip which should probably offer support for physics calculations.
Further he explained that nVidia thinks about using their new architecture in which the pixel and vertex shaders will be unified for physics calculations as well. Due to the unification of these two processing units, it's very simple to calculate physics within the GPU. NV firmly believes in this concept and is confident that they'll even outperform the upcoming quad-core CPU's.
edit: better translation.
Id really like some concrete hard info, from ATI/Nvidia on how much power, what kinda PSU's, we will need for the new cards, and in Crossfire.
I might go all out, and get Xfire R600.
Yanno, now I feel sooooo much better about backing up NVidia saying they'd go unified shaders even though all the rumors kept saying it'd be Fixed Function pipelines.Quote:
Originally Posted by turtle
Good find man!
umm.. the last nVidia card to have "fixed function pipelines" was gf2 ultra. all have been programmable ever since then.
The 128bit (64bit) part of mem controller won't be used only for AA. Why nVidia are willing to put 256MB or 128MB on this part of mem controller?? For xbox360 10MB of RAM is enough to do 4x AA in 1080p! For sure nV will use this to offload Physics, they can keep there Z-buffer data, etc.
One thing we can be sure --> We have something to wait for! :gamer:
Technically, pixel and vertex are considered fixed function.Quote:
Originally Posted by ***Deimos***
please read this:
http://en.wikipedia.org/wiki/Shader
and when you have finished reading it.. please read it again. No insult intended. Just want to make sure we're on the same page.
http://www.gamedev.net/reference/pro...d3d10overview/
This was written by a DX MVP incase you're curious.
Quote:
The Programmable Pipeline. Fixed function graphics are a thing of the past with Direct3D 10; this section explains what this means and why it's a good thing.
Technically, while somewhat programmable, a pipeline/shader that can only perform pixel shader or vertex shading is still considered fixed function. While the amount of functions has broadened, they still weren't fully programmable.Quote:
We've had some form of programmable pipeline for five years now – ever since Direct3D 8 back in the year 2000. Over the number of revisions since then it has become both more powerful and more flexible – and with Direct3D 10 it becomes the only choice. That's right – the fixed function pipeline is history!
DX10 will change that.
do we even have any idea when DX10 comes out ???
if they are correct figures, they make for good reading ... :D :D
A little bit off topic here, but have any of you seen this?
http://www.thermaltakeusa.com/produc...0099/w0099.asp
Dedicated GPU Power supply.....More stuff for me to spend extra $$$ on.....provided i want like Dual R600s and like an X1950Pro for physics or something.
Been out for a while, other companies make them as well. An external brick (IMO) would be better, less internal wiring to deak with.Quote:
Originally Posted by Radeon117X
I know...but in any case, what i really was getting at was how many people will buy this? And Im wondering if it will be necessary when the G80 and R600 come out being very power hungry....Quote:
Originally Posted by Ominous Gamer
I think you picked that up wrong.Quote:
Originally Posted by DilTech
That states that while programmable pipelines have been around, pipes HAVE to be programmable for dx10 and further (you'd get away with fixed function pipes on DX9/8/7 etc, wont use the full API, but will work).
But now states that with DX10, the pipe MUST be programmable.
States nothing about unified shaders, which BTW, are NOT essential for DX10.
yes listen to him ^Quote:
Originally Posted by Dublin_Gunner
"Quote:
We've had some form of programmable pipeline for five years now – ever since Direct3D 8 back in the year 2000. Over the number of revisions since then it has become both more powerful and more flexible – and with Direct3D 10 it becomes the only choice. That's right – the fixed function pipeline is history!
Technically, while somewhat programmable, a pipeline/shader that can only perform pixel shader or vertex shading is still considered fixed function. While the amount of functions has broadened, they still weren't fully programmable.
DX10 will change that."
you completely misunderstood. the author is basically stating that DX10 is incompatible with early versions of DX. You must use the new DX10 API. Calls for DX7 fixed functions like DOT3 bump map wont exist. Everything goes through DX10 API.
maybe I missed something, but is "T&L" being replaced by the "Geometry Shader"?
from what I understand, now allows control of vertexes grouped into primatives, and create/destroy vertices!!!
(ex: me thinks of games where shooting monsters cuts them up into smaller monsters which are just that much more difficult to hit)
i'll take that as a yes, with the possibility of the T&L engines also remaining in a sort of fixed-function form to play along with the geo shaders if need be?
Quote:
Originally Posted by STEvil
They have nothing to do with each other.
T&L is what basically the Pixel shader / vertex shader portion of a 'modern' gfx card does. Before hand those calculation would have had to be done on the CPU, with the GFX 'accelerator' outputting the result to the screen.
So in essence, it hasnt been replaced, but its just a very basic, archaic term for what the shader engines do.
http://en.wikipedia.org/wiki/Transform_and_Lighting
http://en.wikipedia.org/wiki/Transfo...g_and_Lighting
from what I understand, "T&L" is essential, and not yet replaced by any shaders in any GPU. Regardless of the programmable manipulation of pixel values and vertex values, there has to be a mechanism to transform the object 3D world coordiantes into the view frustrum, and onto the 2D screen. Furthermore, all polygon on the boundary of the view frustrum need to be clipped (those not in view ignored). So, like triangle setup, TCL is and for forseeable future will continue to exist.
Now, not sure about the lighting part. In GF/GF2 series they were fixed function DOT3, gloss, specular, etc
that's pretty much all pixel shaders now.
please read:
http://en.wikipedia.org/wiki/Graphics_pipeline
wtf is the view frustrum? the edge of what can be seen?Quote:
Originally Posted by ***Deimos***
Peanuts anyone? http://aboutpeanuts.com/
I hope this card has enough power to fight the gfx violence from Supreme commander and Crysis.
be careful using wikipedia as a source for your knowledge. you wouldn't be the first person to own himself due to wrong articles on wikipedia.Quote:
Originally Posted by ***Deimos***
oops.. my post about view frustrum was lost (stupid wireless).
view frustrum:
it is the projected area being rendered. From the pinhole camera imagine 4 lines being projected out into the distance in a shape of a rectange.
At some close distance you have the front plane. In some games you might notice that you can go into objects.. and they look cutoff. This is the effect of clipping at the front plane. Things behind the front plane are not drawn onto the viewing plane.
At some large distance, typically where you have fog or obsucrity is the back plane. In some games thats not far, but others like Far Cry its really really far. But nevertheless, there is a distance at which enemies and objects magically disappear.. this is the effect of the back plane.
The area within is the viewing frustrum. Only polygons within the viewing frustrum are transformed (in 4D to allow both translation and rotation), lit per vertex, and clipped. There is additional Z buffer that can be used to help speed up the ordering of the objects within this space from front to back and render pixel only for the object on top. Otherwise, like back in the Riva/Rage days, ALL the objects were rendered on top of each other (back then memory was scarce with 2-4MB cards).
In addition ATI/nVidia have hierarchal early Z rejection, to quickly eliminate whole objects and areas from consideration for rendering (in early days the drivers had some quirks and sometimes there were holes in surfaces because of this). A variant of this is "ultrashadow" where large portions of shadow volume are skipped from rendering because not in view.
http://en.wikipedia.org/wiki/Occlusion_culling
http://en.wikipedia.org/wiki/HyperZ
What do you guys make of all the hyper-active folks getting overly excited that R600 wont have just 64 shaders but 64 pipelines!!
Am I the only one who no longer "cares" about melodramatic jumps in Mhz, or brute force approach with more execution units.
Have we learned nothing from P4 vs Athlon64 (and now vs Core 2 Duo)?
What about R300( 9700pro) agressive use of caching and compression as well as hyperZ to dramatically reduce workload?
I personally much prefer the "work smarter" vs the "work harder" approach. Both X1800 and 6800 have 16 pipelines yet look at the dramatic difference in efficiency. For all I care, R600 could have 16 pipelines too.
Sorry if it's old info but do you guys think these new cards will be bottlenecked in a 2*8X sli setup?
Or have we already seen this happen? I havn't used sli since 7800gt so I haven't been keeping myself updated really...
Are these specs confirmed?
Judging by the pictures of the card, with it's hybrid water/air cooler, and 12 memory chips(384-bit memory bus)... They look to be real
http://www.xbitlabs.com/news/video/d...929090635.html
and what do you think about this:
full length dense fin heatsink with water plugs
two PCIE power connectors
very long PCB.
I just hope in all this is that GT models will be "normal" This is just getting stupid... e-penis and geek-penis enlargements GPU companies have gone made this time around!!
________
Ipad guide
It has been said once and I'll repeat it. "It's not worth it to buy the 1st gen DX10 video cards..."
Quote:
Originally Posted by Eastcoasthandle
So shud i buy a X1950XTX or X1900xt CF now. n Wait for 2 yrs before Upgrading to 2nd or 3rd Gen Cards, Will i be good enuf to play at High settings uptill 2yrs ??. I m not rich, i cant afford to upgrade every year.
That's totally up to you at this point. There are no games intensive enough to warrant a next gen video card upgrade unless you plan on buying Vista ie. DX10.Quote:
Originally Posted by ANP !!!
Untill DX10 games are more mainstream then vaporware, save your money and keep what you have. I haven't notice gaming developers flocking to DX10 just yet.
many game developers have been actively using DX10 for game development for long time.. there will be many DX10 games in 2007.Quote:
Originally Posted by Eastcoasthandle
next gen DX10 hardware will be feature superior and much faster than existing hardware.
Games like Oblivion in some areas with all the features turned to maximum strain the most powerful single video cards today 7900GTX/X1900XTX.
BUT
It will take many months, well into summer for nVidia/ATI to get mature Vista DX10 drivers.. likewise, a while for chipset drivers as well as Windows updates to make stable platform. Likewise, games will see their share of patches and updates as they work the bugs out. And as for the performance, well it wouldn't kill you to run wiht 2xAA instead of 8xAA, or turn down the resultion a notch from 1600x1200 to 1280x1024.. I mean... are you really willing to go through all the headaches of being a virtual beta tester for months when you can stick with existing hardware?
Yep G80 will be faster no matter how fast the current cards are. So i think i ll have to wait for G80 Since i want something that can power my 2407, untill then i ll run my 6600 card as a Standby. :rolleyes: :rolleyes: :rolleyes:
IMHO
Its very well possible its the real thing.
But questions:
* 12 memomory chips... so 384/768MB, and 192/384bit memory interface.
* where are all the capacitors?
* 2 PCIE connectors.. lots of power being used.. but sensible to increase stability.
* Looks like multiple redundant voltage regulators for GPU and/or memory.
* Plenty of big inductors on the front and some on the back.. but where are the big power transistors that make up part of buck converter?
* SLI bridge! OMG putting two in a case together!
FYI IMHO they are using multiple-phase power circuitry like on recent motherboards.. smaller (but multiple) capactiors, inductors, and power transistors needed.
http://www.hardocp.com/images/news/1...hCAw_1_3_l.jpg
notice 12 memory chips, and multiple voltage regulators and additional power transistors on back
http://www.hardocp.com/images/news/1...hCAw_1_5_l.jpg
only 2 big inductors, and 2 big capacitors. Many many transistors all silk screened Qxxx. Notice the bank of smaller ceramic capacitors to the right...
I just that this is not the retail version and that you are not forced to have to use both PEG jacks.
I note that while they've possibly enabled water cooling they didn't use a copper cored air heatsink. That seems a bit out of place, to have a use for water as a top end option, but not enough need that a good copper air cooler was used. Maybe this will be the GT GTX difference, or perhaps the water will be some kind of ultra type of card, with both GT and GTX being air based. I'm not going to start on the whole gtx vs ultra issue. I'm getting at the heart of what I'm seeing doesn't seem to speak of being super hot in all formats.
I can say I'm glad if they are blowing the hot air out the back again. That's better than letting it collect inside the case. But there are no heatpipes? Yet we have a water version? This all seems odd in a way, not wrong, but just odd.
Eastcoast touches on a good point however. The 2nd refresh could easily come with a process shrinkage 7800 vs 7900 springs to mind, and that might make the first edition very hot and the second one cooler. So there is that to think about.
Lastly, some folks are running 1920 and 2560 resolutions and while not needed AA necessarily, driving that many pixels with a lot of modern eye candy can really drag down a single card. So I'm thinking we aren't quite out of the "need" range in powerful gpu's yet. There are still ways to bring the most powerful gpu's directly to their knees.
:)
using up to 250W full load for GTX
the water cooling G80 is going to cost around 900$ (not sure but it is what they are talking about today)
for GT is about 180W
short supply for G80
using GDDR3
I think you're completely wrong... I'm so sure, that I'm willing to bet that if I lose I'll post photo of me in panties.Quote:
Originally Posted by guess2098
LOL..... That would be kind of funny..... I also believe they will be using GDDR4
~Mike
So do i assume that this card is a ES and nV will modify it when they start shipping retail cards.
I love your avatar, with the cat coming out of computer case...Quote:
Originally Posted by arisythila
Long time ago, in galaxy far far away, in school we started computer programming on 286's. Since we never used the floppies, we would cram candy bar wrappers into the floppy slot, and every nook and crany. Boy, were they surprised when they tried to save some guy's onto a floppy!!
Make sure it is a thong .. this is extreme after all.Quote:
Originally Posted by ***Deimos***
You are a girly, right ? :confused:
:D
Regards
Andy
DX10 reference whitepaper, which I use as my reference:
http://download.microsoft.com/downlo...indexing%22%22
I like this
"Operation Direct3D 9 Direct3D 10 (reference) Draw 1470 154 Bind VS Shader 6636 416 Set Constant 3297 916 Set Blend Function 787 530 Table 3: Command cycle counts on Pentium IV."
no longer limited to just specular maps, bump maps, normal maps, or even parallax maps (photon mapping)... now you can do displacement mapping (in hardware).
old 2048x2048 texture size limit be gone.. now 8192x8192.. just try to imagine the size of that compared to your monitor's resolution.
each invocation of geometry shader can produce no vertices or up to 1024 in a primative. Where would you need new to "grow" vertices? Dynamic growing trees. A dragon that grows a second head. Realistic non-predetermined debris from collisions/explosions. Displacement mapping. A transformers video game?
And finally, sampling of shadow maps!.. No more jaggy shadows!!
for the bet, does it matter?Quote:
Make sure it is a thong .. this is extreme after all.
You are a girly, right ?
Regards
Andy
I got the perfect pink panties in mind.. ;)
Yay!!!!
So DX10 will fix the dodgy shadows in BF2!!!!
lol
Prolly not in BF2, this new stuff most likely is only on DX10 games, which still have to be released.
Yeah. BF2's dodgy shadows suck like hell. Not even super high resolution's fix it.
Quote:
Originally Posted by Helmore
I was jokin m8, while giving an example of crappy shadows!
Being a DX9 game, obviously DX10 wont fix it!
sometimes hard to tell if you are joking.Quote:
Originally Posted by Dublin_Gunner
but, I suppose if a company like Bungie updated Halo, then Dice can update BF2 too ((HL1 source also comes to mind). Get it all Vista "compatible" and showcase those dark sily smooth shadows.. oh yeah baby.
Likewise with 3DMark2005.. especially apparent on the deck when the captain is looking out for the beast in game test 3.
But, ofcourse you'll only see these benefits if the hardware supports it... and I doubt folks would buy G80/R600 to relive the "classics".
FROM THE ARCHIVES:
I got so excited to see 4-5 trees at a time in Unreal2 preview.. I though wow, this is next generation technology. Yet, all those teaser clips and screenshots of HL2 and Doom3 looked absolutely incredible (back then). At least half the people discredited them as pre-rendered, or photoshoped. Then, I think it was March/APril 2004, OUT OF NOWHERE, Far Cry. My hippie pot smoking friend, eyez ablaze, turned down a reefer to barge into my room, and get me to download this "demo".. pff.. another bunch of screenshots or some stupid clip of "supposed gameplay".
My heart skipped a beat. OMG, I couldn't take my eyes of the screen. I spent hours literaly crawling through the vegetation. Exploring the island. Taking long luxurious swims. Probably took more in-game screenshots that day then all the years before combined! For such a long time, in all sorts of games, you "imagined" you were in a jungle (ex Goldeneye N64), or in a forest, or just walking on gravel or through grass... now, you no longer had to imagine.. it was there. Probably the 3rd religious awakaning, after seeing 3D for first time (Quake), and ofcourse the enchanting music and that memorable warthog in Halo.
Well, stop being so serious all the time Deimos ;)Quote:
Originally Posted by ***Deimos***
How much power Will this Tower really Eat up ?? They Say 300w, But i seriously doubt it. If yes, then they would fry like hell. Cant say about the R600 since they will be very hot. Will my OCZ 700w be fine powering 8800GTX.
Also was wondering, since this Card IS SUPPOSED TO BE VERY LONG, how long will be R600, (ATI say they will be the biggest Cards ever) OMG. :D :rolleyes: :rolleyes:
I don't like it when they release these school bus size cards. They are just bleh. I remember the 6800ultra, that was a huge card, then the 7900GTX came out and that was pretty big. Pretty soon we are going to need a case just for a GPU. It will be linked with a floppy cable haha.
Pretty soon you will buy a video card and plug the motherboard into it!!
Quote:
Originally Posted by ANP !!!
Your psu should handle one at least. Quad Sli might not work though ;)
I doubt they would try quad again.. That whole project was doomed from the start. Direct3D games see little benefit over SLI, since typically only triple buffering. No limitation on buffers in OpenGL, but then again, not many games either. Quad SLI is just way too impractical, and inefficient.Quote:
Originally Posted by Poodle
However, even SLI G80 will surely be big burden. That's probably where they came up with the 250-300W figures. But, also remember you need at least 100-150W for the rest of the system, as well as demand response, and margin for stability. All of a sudden 500W+ PSU requirement doesn't sound crazy anymore.
And one other important thing.. heat. And not just the challenge of cooling G80 (dual slot heatsink or aftermarket water cooling). G80 SLI will surely stress other components too. Some special applications may generate huge PCIE bandwidth load. Chipset probably hotter than ever. Likewise, more CPU-limted scenarios, and more stress on the CPU. If PCIE slot is relied on for power, significant strain there too. And ofcourse, with nearly 500W total being used by the system, even an efficient 80% 500W powersupply will be putting out nearly 100W of heat, all by itself.
i wonder if there will be a gx2 version of the g80. aka a more effecient sli solution? who knows? it would be cool tho.
feeling better about my $230 7900gt(x) LOL
Ah, so T&L is more a weaker less efficient form of what STMicro used in the Kyro series, though limited to the "frustrum". That makes more sense now.
Having mentioned STMicro, anyone else been keeping tabs on the things they've been doing? I poke over there once in a blue moon and saw a few rather interesting things..
EDIT - Looking forward to the panty pic, though there are bound to be more "entry level" consumer cards as well.. so maybe not.
Perhaps you misunderstood meQuote:
Originally Posted by STEvil
there IS NO SUBSTITUTE for T&L (or TLC). Its an integral part of the graphics pipeline. Doesn't matter if you're using pixel shaders or vertex shaders.. those are add-ons. You're always doing T&L.
ST Micro Kyro chips had no T&L hardware acceleration. It was done in software, ie the CPU did the calculations. Games which dont have compatibility fall-back.. games that require hardware pixel shaders or T&L.. they will not work at all with the Kyro.
the die is huge!!!!
using 1.1ns gddr3
it has something like HT build-in
Is it at least small enough for a MCW60 to make contact with the whole thing?
ur kidding right?Quote:
Originally Posted by ***Deimos***
Bungie didnt update halo2 they are just gonna put some code init so it will only run on Vista no other changes are done to it. It wont use Dx10 and it wont look anything fancier then ont he old Xbox. So to be short its a very easy job for a developer.
and Dice/EA will have Bf2142 out when Vista comes so i highly doubt theyl waste any energy on fixing or upgrading anything about Bf2 by then.
ya i guess they are the kind of company that will slap a Vista compatible sticker onit for a value version of Bf2 just to sell to the suckers out there but i doubt theyl fix anything once bf2142 is out.
sometimes hard to tell if you are joking.
but, I suppose if a company like Bungie updated Halo, then Dice can update BF2 too ((HL1 source also comes to mind). Get it all Vista "compatible" and showcase those dark sily smooth shadows.. oh yeah baby.
EDIT: its very well possible... I didn't say they WILL do it.
I had an odd thought
Regarding the power connectors.
Okay so instead running two seperate PCI-e cables to two cards, what if.
It would use one cable, have a "jumper" to the second card.
Running two cards off of one PCI-e cable?
Once there's sufficient current on that rail, I dont see how that would become a problem.
I still fail to see why they would need 2.
The only reason for 2 connectors is because the specs don't allow you to draw more then 75W from a single connector, but I'm sure that you can draw 150+W from a single connector without any problems.Quote:
Originally Posted by Dublin_Gunner
Quote:
Originally Posted by ***Deimos***
me asking if ur kidding is cause you say: bungie updating halo.
were did they update the upcoming Halo 2 for Vista.
the screenshots seen sofar look exactly like what it looked like a few years ago on the xbox. It hasnt been updated in anyway.
edit:
if ur updating is refered to directx or whatever.
dunno what Dx version Halo2 is/was but Vista will simply emulate it.
just touch a died 8800GTX and 8800GT
the 8800GTX is crazy long. about 1.5inch longer than 7950GX2
i am 100% sure it is using GDDR3 1.1ns!
very nice card and fan is better than 6800Ultra!!
but it dead!!!!! lol
gddr3? interesting
im thinking of getting X1900 or X1950 crossfire when G80- comes out, they better make it so i can do it in less than $600 heh
Bungie updated Halo for PC. I didn't say any more. I dont know why you insist that I prove something about Halo2 and Vista.. how did you come up with that?Quote:
Originally Posted by Starscream
Just saying, that sometimes, rarely, games DO get updated. Not always. Not always for Vista. Maybe not all Halo games. Please dont misunderstand.
Quite frankly, I'm not up to date on PCIE.. their website requries registration to view the whitepapers and data sheets... however, I recall about the same values.. 75W from PCIE slot,and 75W from PCIE power connector. Also other websites state 150W from PCIE total.Quote:
Originally Posted by Helmore
As for the 6 pin connector, I think it only supplies two 12V lines... so for 150W, that would be 12.5A.. 6.25A per cable pair. That is a LOT. Perhaps not impossible. But, certainly would affect stability. (note: there is 2 molex to 1 PCIE Y connector)
From what I recall, X800XTPE uses up to about 3A from 12A on power connector. An overclocked 6800Ulltra uses 4.5A and even 5A. But, it has two power connectors.
I think there are some good reasons/explanations to use two power connectors on G80:
1. Multiple voltage regulators and buck converters like on 6800Ultra. A connector/voltage regulator.
2. Power supplies now feature multiple rails.. this way you can draw power from both and not overburden any single one.
3. More even power distribution, less signal noise, and higher stability.
4. Scare the competition!?
Quote:
Originally Posted by guess2098
using up to 250W!!! full load for GTX ( single)
the water cooling G80 is going to cost around 900$ OMG!!)(not sure but it is what they are talking about today)
for GT is about 180W !!??
short supply for G80
using GDDR3???
Quote:
Originally Posted by ***Deimos***:
I think you're completely wrong... I'm so sure, that I'm willing to bet that if I lose I'll post photo of me in panties.
no pink panties for you
* it turns out in many games GTX uses about the same power as X1900XX/X1950XTX. And even in most power demanding situations hits 142-149W (depending on website review).
* G80 doesn't REQUIRE water cooling, let alone $900 worth, and uses a quite conventional heatsink.
* The GT in many cases uses less power than XTX, far far from 180W.
* short supply, reason why GDD3... only time will tell.
FYI: I find it very interesting which of my guesstimates were right/wrong.
80nm=NO. ~7900GT OC clocks at 1.25-1.3V to compensate for complexity=YES, 125-150W=YES, water cooler not required=YES, presumed still using vector units and hence 1.35Ghz ridiculous = DEAD WRONG... even though expecting the unimaginable, nVidia shocked even me... they really went all out!
Actually, the reason is quite simple.Quote:
Originally Posted by ***Deimos***
PCI-E spec allows for 75w per 16x slot, and 75W per power connector.
= 150W max for 1 power connector, but 8800gtx needs closer to 180W or so I believe, hence the need for the second PCI-E power connector, for a total max wattage of 225W
some sites are reporting:
8800GT = 3 phase power
8800GTX = 4 phase power