|
|
 |
|
|
|
 ATI seems to be OK with yields of the 80nm fabrication process and should release early February a Radeon X1950GT, third declination of the RV570(LE), after the Radeon X1950Pro and the recent Mobility Radeon X1900. With GPU and memory frequencies of 500/600Mhz, 36 PS Pipelines, 12ROPs, 8 VS Pipelines and 256bits memory bus, this card should be quite attractive compared to the X1650XT which is clocked at 575/675Mhz but only 24 PS pipelines, 8ROPs, 8 VS pipelines and a memory bus of 256bits.
The thing is that the Radeon X1900GT and X1900GT v2, even if their name indicates that they will be cheaper than the Radeon X1950GT, have actually the same number of units, an identical memory bus and higher frequencies respectively of 575/600 MHz and 512/660 MHz. In consequence, they will be faster for most of the possible uses. Apparently, the acquisition of ATI by AMD hasn't stopped the Canadian manufacturer to continue to defy all logic for the selection of the product names, just like Nvidia does sporadically.
 In Crossfire mode, the performances of the Radeon X1950GT should be closer of superior to the Radeon X1900GT. Indeed, unlike the Radeon X1900GT, the X1950GT has an internal CrossFire connector and doesn't need a master card or the CrossFire software which is less efficient. The US price announced isn't extraordinarily attractive: $169. This price is between the Radeon X1650XT and the Radeon X1950Pro but closer to the second. The usual partners of ATI aren't really enthusiast about this new product and only Palit confirmed to have the intention to produce cards based on this GPU. Sapphire could also follow Palit. |
| |
|
 Initially scheduled to be released in November 2006, the Athlon 64 X2 6000+ is finally postponed to the end of February. We remind you that AMD has chosen the rather ageing 90 nm fabrication process rather than the 65 nm for this frequency increase.
Like the Athlon FX 74, the Athlon 64 X2 6000+ is clocked at 3.0 GHz, has 2X1MB of L2 cache and a TDP of 125 Watts. It will be compatible with the socket AM2. Compared to the 5600+, it is clocked 200 MHz higher and has 36 additional Watts. During the third quarter, AMD will release a 65 nm version with a TDP of 89 watts. The price of this processor should be of $607 or respectively $102 and $77 more than the 5600+ and E6700. |
| |
|
In February 2006, or 4 months before the release of the Core architecture, Intel announced that their processors would be 20% faster than AMD's for « standards » benchmarks and that they would need at least "one year and a half to 2 years to catch up".
Today, it is AMD's turn to start teasing us and the manufacturer announced that the Barcelona would be 40% faster than the Clovertown for several types of tasks. We remind you that the Clovertown is the code name of Intel Quad Core processors currently on the market and the Barcelona is the code name of the K8L. This new quad core architecture has an L3 cache and a FPU outputting two times more than each core of the K8. AMD specified that the Barcelona would be 3.6 times faster than an Opteron dual core with floating point calculations or 1.8 times with an equivalent number of cores.
As always, this type of announcement cruelly lacks of precision and these data shouldn’t be taken as granted. To be more accurate, during the first demonstration of the Barcelona early December, AMD announced that the Barcelona would be 13% faster with TCP OLTP and 46% with SPECfp compared to a Xeon 5355. It seems that AMD has decided to emphasize the second figure for the pre-launch communication campaign….
This is a dangerous game that AMD has decided to enter: avoiding giving all information might lead to disappointments in the end. If that didn't happen with Intel, we hope for the manufacturer that it won't happen with AMD and the 40% (even if a much more reasonable figure would be altogether already really satisfying). |
| |
|
|
Copyright © 1997- Hardware.fr SARL. All rights reserved.
Read our privacy guidelines.
|
|
