Video Cards & GPUs - Page 456

Every day I try to somehow snag a Fermi sample from one of the many companies we work with here at TweakTown. The problem is for the first time over the weekend we heard the 'M' word come out of a partners mouth in regards to the release. What's the 'M' word? March!

NVIDIA has just quietly unveiled its first GeForce 300 series graphics chipset, dubbed the GeForce 310. But don't get too excited, it's said to be based on current architecture, not Fermi (which adds DirectX 11 support and inevitably far superior performance characteristics).

The GeForce 310 is positioned to directly replace the older 210 model as a replacement for systems limited to slow integrated graphics. As such, specifications aren't anything to wow yourself over with a 64-bit memory bus and 16 processing cores. Stock clockspeeds come in at 589MHz for the core and 500MHz on the 512MB of DDR2 memory.

Despite the lowly specs, the card is still said to be quite capable for most desktop tasks including the ability to fully accelerate 1080p video in hardware and make good use of NVIDIA's CUDA or PhysX technologies.

Yesterday I attempted to begin my guide on bitstreaming Blu-ray HD audio from movie discs to your amplifier knowing that the current batch of NVIDIA and AMD GPUs come with this type of support. Namely the GeForce 200 series and Radeon HD 5000 series.

Suspecting it may be an issue, I quickly discovered that while the hardware side is ready to roll, the software is playing catch up. Presently, the publically available versions of CyberLink's PowerDVD do not support bitstreaming with either and we are still waiting to hear back from ArcSoft about whether or not Total Media Theater (TMT) is capable of such a feat.

AMD is set to launch a new flagship video card called the HD 5970 and ahead of that official introduction, some of the specifications for the card have been leaked online. The card has the code name Hemlock and has two GPUs on a single video card.

[img]3[/img]

The GPUs are called Cypress, has an internal CrossfireX configuration, and are built on the 40nm process. Each GPU will have 1600 stream processors and a 256-bit GDDR5 memory interface. Actual VRAM will be 4GB total.

That means each GPU will get a massive 2TB of memory. The dual slot card will reportedly have a pair of DVI-D ports and a mini DisplayPort with the other slot having a large heat vent. The NDA will reportedly expire on the card on the 19th of November.

Wow there is so much spin out right now about Fermi/GT300/G100 that it is not funny. From the Anti-Green side we hear of delays and yield issues. We hear rumblings of problems with the actual design, that NVIDIA knows and is stalling, and that somehow they are pulling the wool over consumers' eyes on their DX11 parts. All in an attempt to undermine poor put upon AMD.

But what is really going on? The truthful answer to that one is, no one but NV (and maybe a few OEMs and TSMC) really knows. But still we do know that recent rumors and rumblings have some root in truth. The problem is with presentation and spin. For example we hear that Fermi is being delayed until Q1 2010. When you read this you automatically hear that everything related to the Fermi/GT300/G100 is being delayed. You think that this means no DX11 GeForce cards from NV this year, time to pull out the wallet and look to AMD right?

Well you would only be partially right.

You see Fermi actually represents several products; there is the GeForce, Tesla, and Quadro. So what are the new reports really talking about? Well they are talking about Quadro and Tesla. Traditionally new Quadro Products from NVIDIA are released in the April Time Frame right around NAB and Sigraph. So the news that this will not be launched until then is nothing new and just business as usual. As for Tesla, they are usually updated once a year so again that is not new and nothing out of the ordinary.

Now for most the big question is GeForce; so far NVIDIA is still saying that it will be launched this year. This could be very true and we will see DX11 performance from NV sometime in the next month or so. But will this be a general launch or a limited launch just to make the deadline. My thoughts now are different from what they were a couple of months ago. For starters at the time I was being told that the GT300 DX11 parts would make their debut in early to mid-November. This has now changed to late November, possibly after Black Friday. A launch this late with nothing in the stores will not improve NV sales but it might hurt AMD's if the performance is good enough. So I think that a limited product launch is very likely in the next few weeks. NV needs something on record before the end of the year to compete with AMD's current momentum. Anything less than a true DX11 part based on Fermi will not cut it with today's consumer so they had better be working very hard to get those out.

NVIDIA's third instalment of their SLI Profile updates has just been released which expands upon the list of game titles to support the company's SLI multi-GPU technology and in some cases also optimizing support for SLI-ready titles.

The latest additions to support SLI using this update are as follows :-

• Borderlands
• Cloudy with a Chance of Meatballs
• G-Force
• FIFA Soccer 2010
• League of Legends: Clash of Fates
• NHL 2009
• Order of War
• Pro Evolution Soccer 2010
• Race On
• Star Trek: D-A-C
• Champions Online (Enhanced support)
• Dragon-Age: Origins (Enhanced support)

The GPU race is heating up. This is evidenced by the latest round of FUD circling the internet over the weekend. You see NVIDIA used a Mock Up card for its introduction of the Fermi at their GTC event. Now this is a typical setup. Manufacturers often show off mock up cards at trade shows - nothing new here really.

So why has this become a big topic in the news? Simply put because Fermi represents a real threat.

Even if it is not ready for launch right now (all of our sources still say first week of November for the launch and availability at the end of November). It is still a serious enough threat to AMD that much of the "information" getting out can be traced back to them.

This is like the rumors about Larrabee and its yields. With Larrabee, AMD was claiming that Intel was on their 3rd or 4th version of silicone. With GT300 yields, it was that they were less than 2%. Neither appears to be true according to the information we have about this.

So, what of this latest one? Well, we have heard that the first information about the mockup card was sent around by an AMD employee. Despite the truth that it was a mock up card, the mere fact that AMD is sending this out shows that Fermi represents a real threat to AMD and their new 58xx series. After all, why throw mud if you know you have the upper hand?

I do not know how the two GPUs will stand up next to each other in terms of performance. After all, I have not seen the cards in action side by side. But I do know that I am more eager than ever to find out, knowing that the information about the mock up came from a source at AMD. The sad part is that this will probably backfire on AMD as many enthusiasts will do the same thing and wait to see what all the bother is about.

So get this, after all the people claiming the GT300 was a pipe dream or failure were prove wrong yesterday with the demo the first Fermi based Tesla cards we now hear some people claiming that NVIDIA's launch shows a commitment towards GPU computing and a move away from Graphics.

This argument is about as bad as the people that claimed the Core i7 would be horrible for gaming because it Looked to be aimed at the high end desktop. At the time when these comments were made not much was known other than what Intel showed at IDC. This is the same thing we have here.

NVIDIA did a great job of keeping GT300/ Fermi under the radar. There was enough ATi news and NVIDIA bad PR to keep much of the details hidden from the public. As such it was not until the last few weeks that the GT300 was even known by the vast majority to be real and ready for launch.

However, NVIDIA has chosen to show off the power of Fermi in a very clever way. This is a top down method. We see the pinnacle of the line here with full GPU computing power. It is from this cPGU (as some are calling it) that the Gaming parts will be made.

People making claims about Fermi's performance as a GPU is the equivalent of only driving a car at 100MPH and saying it handles poorly at slows speeds. We all will have to wait and see how well it does in gaming performance. Personally with the information I have, I think everyone is going to be shocked at how well it does in gaming and where it might go in the near future.

Word's come in from the horse's mouth that NVIDIA's first DirectX 11 (GT300) GPU is shaping up okay and should make it for a Q4'09 launch, despite previous reports indicating delays and yield issues.

Remember all the rumors about bad yields of the GT300? Well nVidia has come out to confirm what we told you before. GT300 yields are fine and they are on schedule for a Q4 2009 launch. This launch should be at GTC in November with retail availability by the end of November.

Fudo over at Fudzilla is also confirming this with a direct statement from NVIDIA on the subject. They are claiming that the reports of less than 2% yields were based on a mistranslated report from AMD's competitive analysis team.

Also they say that the reports that the GT300 is a rebranded/respun GT200 are baseless. According to the report the GT300 is a ground up redesign and a new GPU. This is good news for the NVIDIA fans out there as recently AMD has been hogging all the press with the DX11 40nm Radeon HD5870.

Still we will have to wait and see what the GT300 bring to the table in terms of performance. Traditionally AMD has been a little better at implementing new versions of DX while NVIDIA has been better at the brute force approach. Either way the last month of the year will be interesting to be sure.