ATI Radeon HD 4870
Published by Bogdan Alex, on June 24th, 2008, in the categories: Video Cards
Radeon HD 4870 will officially be launched tomorrow and. from what several sites report. It looks like this card is going to be up to 30% faster, while still cheaper than the 3870X2.
Apparently, the GDDR5 memory found on the 4870 model works at 900MHz basic frequency. However, the GDDR5 memories will benefit from the quad data rate architecture that results into an impressive bandwidth and speed of 3600MHz effectively. The 512Mb GDDR5 chips that ATI implemented in the new 4800 family have 3.6GB/s throughput or double what would GDDR 3 or 4 have at the same 900MHz speed.
The Radeon HD 4870 is going to primarily ship with 512MB onboard, amounting to eight 512Mbit chips. Later on, GPU integrators will be adding 1GB GDDR5 or as much as sixteen memory chips, but this will probably have to wait some more, as GDDR5 module availability suffers from early shortages.
Extrapolating what we know about the 4870 and the already available 4850 model, Fudzilla came to the conclusion that the 4870 will end up with almost twice as much bandwidth than the 4850, meaning around 115GB/s in total. Not to mention that 4870 and the entire 4800 family still supports 7.1 audio acceleration.
Apparently, the GDDR5 memory found on the 4870 model works at 900MHz basic frequency. However, the GDDR5 memories will benefit from the quad data rate architecture that results into an impressive bandwidth and speed of 3600MHz effectively. The 512Mb GDDR5 chips that ATI implemented in the new 4800 family have 3.6GB/s throughput or double what would GDDR 3 or 4 have at the same 900MHz speed.
The Radeon HD 4870 is going to primarily ship with 512MB onboard, amounting to eight 512Mbit chips. Later on, GPU integrators will be adding 1GB GDDR5 or as much as sixteen memory chips, but this will probably have to wait some more, as GDDR5 module availability suffers from early shortages.
Extrapolating what we know about the 4870 and the already available 4850 model, Fudzilla came to the conclusion that the 4870 will end up with almost twice as much bandwidth than the 4850, meaning around 115GB/s in total. Not to mention that 4870 and the entire 4800 family still supports 7.1 audio acceleration.
Improved NVIDIA GP-GPUs
Published by Bogdan Alex, on June 22nd, 2008, in the categories: Video Cards
ATI released its new FireStream GP-GPU and now it’s NVIDIA’s turn to announce their improved general purpose processing architectures.
According to Digitimes, immediately after ATI announced the FireStream 9250 model, NVIDIA revealed at the new Tesla 10-series computing solutions at the same International Supercomputing Conference. NVIDIA points out that the new products enable developers to solve computational challenges in a common and familiar development environment (mostly C-based) that scales effortlessly from one generation to the next with no re-coding required.
The new Tesla array includes the Tesla S1070 1U computing system and the Tesla C1060 computing processor and delivers up to four Teraflops per 1U system, plus there’s additional support for IEEE 754 arithmetic, 16GB of memory per 1U system and more efficient TDP specs.
The new Tesla family is compatible with the CUDA C-language development software for parallel computing, thus extending the reach of GPUs to any computationally intensive applications requiring double precision accuracy. NVIDIA boasts a bit, stating that over 70 million CUDA enabled GPUs have been sold into the market and over 60,000 downloads of the C-compiler have been recorded through the community web site.
Tesla S1070 1U and C1060 computing processor boards will cost $7999 and $1699 respectively and will be available starting August 2008.
According to Digitimes, immediately after ATI announced the FireStream 9250 model, NVIDIA revealed at the new Tesla 10-series computing solutions at the same International Supercomputing Conference. NVIDIA points out that the new products enable developers to solve computational challenges in a common and familiar development environment (mostly C-based) that scales effortlessly from one generation to the next with no re-coding required.
The new Tesla array includes the Tesla S1070 1U computing system and the Tesla C1060 computing processor and delivers up to four Teraflops per 1U system, plus there’s additional support for IEEE 754 arithmetic, 16GB of memory per 1U system and more efficient TDP specs.
The new Tesla family is compatible with the CUDA C-language development software for parallel computing, thus extending the reach of GPUs to any computationally intensive applications requiring double precision accuracy. NVIDIA boasts a bit, stating that over 70 million CUDA enabled GPUs have been sold into the market and over 60,000 downloads of the C-compiler have been recorded through the community web site.
Tesla S1070 1U and C1060 computing processor boards will cost $7999 and $1699 respectively and will be available starting August 2008.
GeForce 9800 GTX Price Down
Published by Codrut Nistor, on June 19th, 2008, in the categories: Video Cards
I know it sounds incredible, but it's true! The GeForce 9800 GTX cards from NVIDIA are going to become cheaper. In fact, the price is getting close to ridiculous! I know, models for enthusiasts like the BFG 9800 GTX ThermoIntelligence won't get that low for sure, but...
...what else can you say when NVIDIA announces its partners that the recommended prices on the GeForce 9800 GTX, from now on expected to be available for only $199! The main reason behind this is the fact that NVIDIA announced the GeForce 9800 GTX+, a slightly overclocked version of the 9800 GTX, which is planned to go up against the Radeon HD 4850.
The 9800 GTX+ is going to have the core/memory frequencies raised to 738/1836MHz, about 10% higher than the 675/1688MHz default speeds of the 9800 GTX. The price expected for the 9800 GTX+ is $229, and that's it.
Obviously, all this can be described with a single word: lame. I am sure that, if ATI wouldn't have had the HD 48xx breathing down NVIDIA's neck...you know what would have happened, I don't have to say it!
...what else can you say when NVIDIA announces its partners that the recommended prices on the GeForce 9800 GTX, from now on expected to be available for only $199! The main reason behind this is the fact that NVIDIA announced the GeForce 9800 GTX+, a slightly overclocked version of the 9800 GTX, which is planned to go up against the Radeon HD 4850.
The 9800 GTX+ is going to have the core/memory frequencies raised to 738/1836MHz, about 10% higher than the 675/1688MHz default speeds of the 9800 GTX. The price expected for the 9800 GTX+ is $229, and that's it.
Obviously, all this can be described with a single word: lame. I am sure that, if ATI wouldn't have had the HD 48xx breathing down NVIDIA's neck...you know what would have happened, I don't have to say it!
ATI Prophecies
Published by Codrut Nistor, on June 18th, 2008, in the categories: Video Cards
Pretty soon, the world as we know it is going to end. This has nothing to do with any natural disaster, because I am talking about the computer hardware landscape, and not the Earth. Why? Well, according to ATI's senior VP of marketing, Richard Bergman, NVIDIA's GeForce GTX 200 GPU is the last high-end monolithic "megachip." So... no more single-chip GPUs??? Good question!
According to EETimes , Mr. Bergman said "I predict our competitor will go down the same path for its next GPU once they see this. They have made their last monolithic GPU."
Looking at it closer, he may be right - NVIDIA's GTX 200 GPU has a die size of almost 600mm², because its dimensions are 24X24mm, while the 55nm GPUs from ATI are smaller, with only 16X16mm. The catch? We'll leave power consumption aside, and mention one big factor - pricing, because the less chips fit on a wafer, the more expensive is to produce them, so that's why we'll soon get to see incredible performance from ATI for about $300(I hardly wait!).
In the end, it seems AMD doesn't need to go with large monolithic GPUs. If they may not get the best monolithic GPU for desktop computers, which may not be true anyway, because the current test results look amazing, ATI will end up with the most powerful GPU ready to use in notebooks, while there seems to be "no way this new Nvidia core will be in notebooks this fall," according to Bergman.
...what else should I say? I just have to ask what do you think - will NVIDIA shrink the die size and go with the flow, or keep hanging to those large DirectX 10 GPUs and DDR3, while ATI already stepped into the DirectX 10.1 and GDDR5 era?
According to EETimes , Mr. Bergman said "I predict our competitor will go down the same path for its next GPU once they see this. They have made their last monolithic GPU."
Looking at it closer, he may be right - NVIDIA's GTX 200 GPU has a die size of almost 600mm², because its dimensions are 24X24mm, while the 55nm GPUs from ATI are smaller, with only 16X16mm. The catch? We'll leave power consumption aside, and mention one big factor - pricing, because the less chips fit on a wafer, the more expensive is to produce them, so that's why we'll soon get to see incredible performance from ATI for about $300(I hardly wait!).
In the end, it seems AMD doesn't need to go with large monolithic GPUs. If they may not get the best monolithic GPU for desktop computers, which may not be true anyway, because the current test results look amazing, ATI will end up with the most powerful GPU ready to use in notebooks, while there seems to be "no way this new Nvidia core will be in notebooks this fall," according to Bergman.
...what else should I say? I just have to ask what do you think - will NVIDIA shrink the die size and go with the flow, or keep hanging to those large DirectX 10 GPUs and DDR3, while ATI already stepped into the DirectX 10.1 and GDDR5 era?
NVIDIA Starts Price War
Published by Bogdan Alex, on June 18th, 2008, in the categories: Video Cards
The battle for desktop computer graphics is set to begin once again, but this time, there’s going to be a price war as well. NVIDIA has recently launched the GTX 280 and GTX 260 cards, and we can already spot a €283.89 GTX 260 version coming from Asus. Of course, these are not yet available, but by the time ATI volumes in its new HD 48XX cards, things should be getting tighter. However, ATI is very proficient at price wars, as the company started this nasty game with RV670, Radeon 3870 and 3850 cards. On the other hand, NVIDIA learns pretty fast and it’s willing to put some early pressure on the Canadians.
Fudzilla estimates that the Radeon HD 4870, which is supposed to cost $329 at launch, will probably drop to $299 very fast and at the same time NVIDIA will go ahead and slash prices for the GTX 260, way under the initial $399. Further speculations point out that the GTX 260 will end up costing around $349 in about two weeks, putting pressure on its direct contender, the Radeon HD 4870.
Moreover, NVIDIA plans to cut prices for the 9800GTX to under $299 to complicate things a bit more for ATI’s Radeon HD 4850. Tough calls for ATI, but all this price war is going to prove beneficial either way for us end users.
Fudzilla estimates that the Radeon HD 4870, which is supposed to cost $329 at launch, will probably drop to $299 very fast and at the same time NVIDIA will go ahead and slash prices for the GTX 260, way under the initial $399. Further speculations point out that the GTX 260 will end up costing around $349 in about two weeks, putting pressure on its direct contender, the Radeon HD 4870.
Moreover, NVIDIA plans to cut prices for the 9800GTX to under $299 to complicate things a bit more for ATI’s Radeon HD 4850. Tough calls for ATI, but all this price war is going to prove beneficial either way for us end users.
PAGES