Forums | MacLife

X-bit labs has a nice look at the 4xx0s up. E.g., the tesselator...

Back in May, 2007, it was a big surprise for us that the developer decided to program the tesselator using vertex shaders, not using DirectX 10 geometry shaders. At that time ATI explained that the ATI R600’s tessellation processor was taken from Microsoft Xbox 360 game console, which features DX9-class graphics chip developed by ATI Technologies. The tessellation unit of the ATI RV770 graphics processing unit can be programmed by both geometry and vertex shaders, hence, it is completely backwards compatible.

[img]   [img]

According to officials from AMD’s graphics product group, there is a number of games that take advantage of ATI’s programmable tesslation unit incoming and all of them were developed on ATI Radeon HD 2000 and HD 3000 hardware, which means that quite a number of end-users will be able to enjoy improved image quality and, perhaps, even feel themselves happy for choosing ATI Radeon over Nvidia GeForce.

But while the actual video games are enroute, those who have ATI Radeon HD 3000 or 4000 may enjoy The Froblins demo, which uses DX10.1’s global illumination technique, HDR, lighting and post-processing along with tessellation for frogs-goblins and terrain. Moreover, in the demo ATI Radeon HD hardware can even compute artificial intelligence, something new for graphics processors, isn’t it?

I see a couple of inaccuracies about history and DX8.1/PS1.3 & 1.4 et al next page, but interestingly, on 10.1 it's mentioned

Fortunately for ATI, not all PC games belong to the aforementioned program and there are at least two coming in within the next six months to nine months that take advantage of DirectX 10.1: BattleForge developed by Phenomic and published by Electronic Arts as well as an unnamed title from Sega.

BattleForge by Phenomic EA is a fantasy online real-time strategy (watch the video here), which involves loads of battle units. Image quality in the game is truly impressive: all the units have high-quality geometry, terrain and vegetation look pretty realistic and special effects are remarkable. According to officials from Phenomic Studio, the BattleForge game runs about 30% faster on DX10.1 compared to DX10 thanks to lower amount of rendering passes needed (perhaps, not in all types of scenes). The game is set to emerge already this year, but the release date is unknown.

Little is known about the DirectX 10.1 game title set to be published by Sega in early 2009 and formally unveiled sometime in July, possibly at the E3 convention. According to Chris Southall, technical director of Sega Europe, the title will only have DirectX 10 and 10.1 rendering paths, hence, will not work on DirectX 9-compatible hardware and will not function under Windows XP operating system. What is the most important, the unnamed title is PC-exclusive. According to Sega, DirectX 10.1 allowed the developer to create its title “easier”, make it look “prettier” and make it work “faster”, though, no actual details were unveiled.

Screenies here too. And not to forget D3's DX10.1 rendering.

http://www.xbitlabs.com/articles/video/ … d4850.html

"I knew there was something odd when my GPU tried to eat my hand!!" -Anonymous nVidia customer

CHIP FAB TSMC IS to delay its 40nm process, a bit. It was originally scheduled to start shipping this year, just barely, but now it has been pushed out.
...

The first one out of the gate is usually Altera, and the firm was hoping to ship 40nm parts this year, but that looks doubtful. The next big boy was Nvidia. Its next-gen part was set to tape out in September according to its engineering roadmaps. Still, if NV is as timely as usual with tapeouts, the 40nm delay won't hurt them at all.

Next on the road to pain is ATI. It was planning to have parts out weeks after Nvidia, but that likely won't happen either, because of the slip. In the end, all three big players will be out with 40nm parts at about the same time, late Q1.

If you need an FPGA to connect your next gen NV part to your next gen ATI part, there won't be any waiting before you get all the parts you need to make your boards. It will be a lively Q1. µ

So maybe the new/current-gen parts will be soldiering on longer than expected. 55nm die-shrinks at best for GT200 parts. We can watch prices drop longer before upgrading. Whee?

http://www.theinquirer.net/gb/inquirer/ … elays-40nm

Intel is fabbing 45nm chips, tho.

Your case rests uneasily.

‘The problem that we’ve seen over and over and over again in the computing industry is that there’s a cool new idea, and it promises a 10x or 20x performance improvements, but you’ve just got to go through this little orifice called a new programming model,’ Gelsinger explained to Custom PC. Those orifices, says Gelsinger, have always been ‘insurmountable as long as the general purpose computing models evolve into the future.’

Gelsinger used the Cell architecture used in the PlayStation 3’s CPU as an example to prove his point. ‘It [Cell] promised to be this radical new computing architecture,’ said Gelsinger, ‘and basically years later the application programmers have barely been able to comprehend how to write applications for it.’

This, according to Gelsinger, is one of the major reasons why Intel’s forthcoming Larrabee graphics chip will be entirely based on IA (Intel Architecture) x86 cores. ‘Our approach to this has been to not force programmers to make a radical architectural shift,’ explained Gelsinger, ‘but to take what they already know – this IA-compatible architecture – and extend the programming model to comprehend new visual computing data-parallel-throughput workloads, and that’s the strategy that we’re taking with Larrabee.’

Larrabee, according to Gelsinger, will simply expand on a standard programming model. ‘It’s an IA-compatible core,’ explained Gelsinger, ‘and we’re extending it with a graphics vector visualisation instruction set that has full support for native programming models, full support for the graphics APIs like DX and OpenGL, and then this broad set of new programming models to go with it.’

Vector ISAs, we have heard of them.

http://www.custompc.co.uk/news/602868/i … story.html

ATi Round 2 is about here. Diamond is readying its XOC 4870 (pic). Diamond PR sez

The HD 4870 is a smoking gun dual slot card, PCIE 2.0, with 512MB of DDR5 memory and a clock speed of 800 MHz. The memory speed is 1100 MHz and is designed with 800 stream processors. The HD 4870 provides plug-and-play ATI CrossFireX™ upgradeability with up to quad-GPU support. Continuing with ATI’s Power Play and 55nm processing technology, this card is the fastest and efficient. “The Diamond Radeon HD 4870 XOC Black Edition was clocked to kick some ass”. We didn’t just want a fast card out in the market, we wanted the fastest card that could kick the living daylights and bust some performance records, say Mario Gastelum, Director of Product Development & Engineering. “we wanted a card that kicked the competitions teeth into the curb”, and that’s exactly what our engineers accomplished”. “The firmware was custom designed to enable end users to go beyond the normal over clocked speeds and allow them to push their cards for higher performance via the catalyst control center.” The GPU’s custom firmware has been unlocked to push cards to GPU settings of up to 950 Mhz and Memory of up 1200 Mhz.

MaximumPC sez

The buzz is flying about AMD’s “Super R770” and the possibility that it will snatch the GPU crown from Nvidia’s GeForce GTX series. As Editor-in Chief, Will Smith reported at the end of June, “ATI eschewed the huge, hot monolithic GPU for a more compact, but modular core. With twin goals of decreased power consumption and more efficiency per die area, ATI looks poised to dethrone Nvidia” and later said, “The Radeon 4870 runs nearly as fast as a GTX 280 in most benchmarks for about 60% of the cost.”

The "Super RV770" will arrive with water-cooling pre-installed and an unlocked BIOS, which enables the GPU to be pushed all the way to 950 MHz and the memory to 4.8 GT/s According to some sources, you may be able to push the GPU beyond 1 GHz, using TEC elements, and keep the temperature of GPU low. Don’t look for this unit in retail; it is an AIB/OEM-only product.

Oh, do not talk about 4870s OCed to 1+GHz via TEC. I can't afford my OCing juices firing up again...

::dreams of H3 tesselated FSAAed and filtered thru 4870s on calibrated 30-bit displays::

Bat wrote:

Bat wrote:

Books, keep an eye on that 8600M.

http://www.theinquirer.net/gb/inquirer/ … ass-itself

Worse. Books, really keep an eye on that 8600.

THE BURNING QUESTION on everyone's mind is what Nvidia parts are failing in the field? No GT200 jokes here, NV personnel are still quite sensitive about that, but our moles have told us about the bum GPUs.

The short story is that all the G84 and G86 parts are bad. Period. No exceptions. All of them, mobile and desktop, use the exact same ASIC, so expect them to go south in inordinate numbers as well. There are caveats however, and we will detail those in a bit.

Both of these ASICs have a rather terminal problem with unnamed substrate or bumping material, and it is heat related. If you ask Nvidia officially, you will get no reason why this happened, and no list of parts affected, we tried. Unofficially, they will blame everyone under the sun, and trash their suppliers in very colourful language.
...

When the process engineers pinged by the INQ picked themselves off the floor from laughing, they politely said that there is about zero chance that NV would change the assembly process or material set for a batch, much less an EOL part.

On the less technical side, multiple analysts also told us that NV specifically told them that this problem is confined only to HP. I wonder why Dell is having failures in huge numbers for their XPS lines and replacing them with ATI parts? Why is Asus having similar problems? Go check the message boards, any notebooks that came with G84s and G86s have boards filled with dead machine problems. Most of these, especially on the NV forums are being quashed and removed by admins, so act quickly and take screenshots of your posts.

All Nvidia G84 and G86s are bad

And to think, I just helped a friend order a Dell XPS with an 8600M GT in it... whee

To all you guys at nVIDIA who knew about this problem yet remained silent, smurf you!  I'm happy that I dodged the bullet, though.

Pink?!


(...his name isn't Ringwald, is it?)

ExtremeTech has a 4870X2 preview. Looks like maybe driver immaturity is the main reason it's not out yet.

3DMark Vantage is the latest, DX10-only test from Futuremark. As with their previous tests, it's not an actual game and does not necessarily stress the system the way a game would. On the other hand, it's far more forward-looking than games on the market today, and uses graphics techniques and levels of detail we won't commonly see in PC games for six months to a year.

3DMark Vantage shows enormous potential for the new ATI card. Scaling over the Radeon HD 4870 is about 80% in High mode, and a whopping 95% in Extreme mode. It's also quite a bit faster than the $650 GeForce GTX 280. So far so good, right? Well, there's a reason the card isn't out yet, as some of our other benchmarks demonstrate.
...

Still Some Work To Do

Our early look at the performance of the Radeon HD 4870 X2 (code-named R700) is at once exciting and disappointing.

When we look at these early benchmarks, we want to consider two things. First, how much faster a dual-RV770 card is than a single RV770 card. Our R700 engineering sample appears to use GDDR5 and run at the same clock speeds as the Radeon HD 4870, so we want to see how much faster two of them on a single board can go. Second, we want to compare ATI's flagship high-end product against the Nvidia's, the GeForce GTX 280.

Tests like 3DMark Vantage and Company of Heroes show amazing potential. The new card scales very well over a single Radeon HD 4870 and is quite a bit faster than the GeForce GTX 280. Crysis, on the other hand, shows almost no scaling at lower resolutions.

When ATI finally ships the Radeon HD 4870 X2, the scaling over a single Radeon HD 4870 needs to range from "good" (50% or more) at mid-level resolutions to "fantastic" (80% or more) at high resolutions with AA/AF enabled. It looks like it's there on some applications, and not on others. If dual-GPU solutions are going to be the norm for ATI's high-end graphics from here on out, they really need to nail it.

http://www.extremetech.com/article2/0,2 … 444,00.asp

The AT Preview starts to clarify what's been known for awhile, ATI is progressing towards sharing resources (like VRAM) between chips, reducing or eliminating waste and redundancy. The PCIe2 bridge chip between the two on-card chips should go a long way to reducing latency, etc.

Building a Better CrossFire
When AMD began talking about no longer building high end hardware using single monolithic GPUs a few weeks back, we let them know that improving CrossFire support would be incredibly important going forward. AMD told us that they are putting a lot into that but also that they have some exciting technology up their sleeves with R700 to help out as well. Unfortunately, we haven't gotten as much detailed information on how it works, but the new technology is GPU to GPU communication.

Until now, CrossFire has done zero GPU to GPU or framebuffer to framebuffer communication. As with the first iteration, each card fully renders the parts of the screen for which it is responsible (be it a whole frame in AFR, the top or bottom half of a screen, or alternating tiles). These results are sent to a combiner where the digital signals are merged and output to the screen. This is the only communication that takes place in CrossFire at the moment. R700 will change that by allowing GPUs to communicate.

[img]  [caption]RV770 has a CrossFire X Sideport...we assume that the two RV770s on a single R700 board somehow connect Sideports and make fast. AMD hasn't told us how yet.[/caption]

It is not clear how extensive this communication will be, what information will be shared, or how much bandwidth requirements are increased because of this feature. And while it is a step in the right direction, the holy grail of single-card multi-GPU solutions will be a shared framebuffer. Currently both GPUs need a copy of all textures, geometry, etc., and this is a huge waste of resources. [..]

CrossFire started life trailinging SLI. Now it's clearly ahead, and NV is begging Intel mobo-makers to include support via their so-so bridge chip.... how things change.

Maybe 4870X2 looks just too good to NV. Maybe desperation, but a consumer win if you fancy the 'monolithic megachips' and have the PSU for them.

GeForce GTX 280 price reduced further

Apparently NVIDIA isn't finished trimming GeForce GTX 280 pricing. Early this morning online retailers began selling GeForce GTX 280 cards for as low as $449.99 without mail-in rebate. Factory OC'ed cards are now selling for $499.99. This marks the second price cut for the GeForce GTX 280, which originally launched 30 days ago priced at $650. Who is carrying the card at $450? For the sake of my sanity, I'll limit the links to Newegg and Zipzoomfly, who seem to be carrying more cards priced at $450 than other online etailers at the moment:

[lists/ quotes]

http://www.firingsquad.com/news/newsart … chid=20682

THE PAIN JUST KEEPS coming for Nvidia, this time in the partner space. Two of their key 'alpha' partners have just defected.

It is nearly impossible to make money selling Nvidia cards right now, NV squeezed partners' margins too much, and when the going got rough, prioritised its own margins over partner survival. This gave partners the choice of popping like a zit financially or moving to less green pastures.

Like Gainward before them, two of the largest, XFX and EVGA, have defected. Want to bet Nvidia doesn't know yet? In any case, partners leaving a company like rats leaving a sinking ship is never a good sign, let's see what happens in the coming months when the pain really piles on at NV.

The really interesting part is who they defected to, and it isn't ATI. Paperwork has been signed though, and it is a done deal.

http://www.theinquirer.net/gb/inquirer/ … ers-defect