|
|
 |
|
|
|
Following Acer and the GN245HQ, which should come onto the market this month, Asus has presented the VG278H, a 27’’ 3D stereo screen which supports this mode both via the DVI Dual-Link and the HDMI 1.4a. The HDMI 1.4a allows for wider compatibility in terms of the graphics cards supported but is limited to 30 Hz for 1080p throughput, while the DVI solution, a prorietary NVIDIA 3D Vision one, will support 60 Hz, which is what you need for gaming. Note that an infrared transmitter is built into the screen. This model should be with us during the month of April.
 |
| |
|
There were a few ‘special’ GeForce GTX 580s at the show. While most high-end graphics cards available in stores generally make do with the reference card design, it’s always interesting to check out the numerous customised designs, especially the most extreme among them.
 Asus were showing the GeForce GTX 580 Matrix Platinum. While it uses a triple slot design that at first seems similar to the one used for the GeForce GTX 580 DirectCU II, in fact the cooler is bigger. More importantly, the PCB has been entirely revisited. The power stage is equipped with no less than 16 phases, direct overclocking controls on the PCB itself, a ‘safe mode’ bios, which can be activated at the back of the card and read points for the various voltages.  Gigabyte was showing the GeForce GTX 580 Super Overclock. It has a sturdy cooler with a ‘king-size’ vapour chamber and three 8 cm fans underneath. The PCB has been revised and has 14 phases, two bios’ with a switch for going from one to the other as well as voltage read points.  MSI wasn’t going to be left out and was also showing an extreme GeForce GTX 580: the Lightning. MSI has gone for the new Twin Frozr III series cooler, with improved effectiveness thanks to new fans with better directed air flow. Here too, the PCB has been revised. It is now enormous, exceeding the standard format by around 2 cm vertically. There’s a 16 phase circuit as well as two bios’, one of which is optimised for extreme overclocking with liquid nitrogen and the NVIDIA energy consumption monitoring system deactivated. Neither pricing nor clocks have been given for any of these three cards. The three manufacturers will no doubt be checking each other out constantly right up to the last minute on this! These models should come onto the market at the beginning of April. |
| |
|
Two of AMD’s partners have confirmed that the Radeon HD 5700s are going to be renamed as the Radeon HD 6700s across the board. Remember, a few months ago, AMD told the press that this wouldn’t be the case, before changing their tune at the end of January, insisting on the fact that these ‘new’ cards would be reserved for OEMs.
Now we can expect the name change to be rolled out across all markets, as the early rumours were suggesting last October. The change will come into effect at the end of this month or beginning of April. Although AMD’s partners are saying that this will facilitate understanding of the range on sale, they hope to avoid the scenario whereby sales of the GeForce 9800 GTX+ fell after criticism of it being renamed as the GeForce GTS 250.
|
| |
|
 At CeBIT, Asus was showing a prototype (keyboard not working) of its forthcoming gamer laptop: the G74SX 3D. It uses a Core i7-2630QM processor and an NVIDIA graphics card, with a blanked out model number. A quick visit to the control panel however, revealed a GeForce GTX 560M (a model which doesn’t officially exist) with spec:  This mobile graphics card is based on a GF116 GPU, a new revision of the GF106 which equips the GeForce GT 555M as well as the GeForce GTX 460M. Compared to the GTX 460M, the GPU clock is up 15% while memory bandwidth is down 33%: though the GDDR5 is identical, the bus is down from 192 to 128 bits. Note that given the NVIDIA architecture, the GF106/116 can’t exploit the 192-bit bus to best effect, which should mean that the reduction in bus size has less of an impact than you might think on first analysis. Broadly speaking, the GTX 560M is thus very similar to the desktop GeForce GTS 450 which will soon be replaced by the GeForce GTX 550 Ti, which is also based on the GF116. NVIDIA is probably waiting for the launch of the GeForce GTX 550 Ti to launch the GeForce GTX 560M (shouldn’t be too long a delay). |
| |
|
At CeBIT AMD organised a demo of its forthcoming APU, Llano. To recap, it will include an Athlon II X4 and a GPU derived from Redwood, which equips the Radeon HD 5600s. Note that AMD has confirmed that this GPU will be equipped with a full UVD3 graphics engine, which will therefore support acceleration of Blu-ray 3D playback, in contrast to AMD’s C and E APUs.
AMD used two laptops for its demo. The first was equipped with a Llano prototype, the AMD A8-3510MX, which corresponds to a quad core version clocked at 1.8 GHz and equipped with a Radeon HD 6620M GPU clocked at 450 MHz (this clock may be slightly different on the final model). The second was based around a Core i7-2630QM clocked at 2 GHz (2.9 GHz in turbo) and equipped with an HD 3000 graphics core clocked at 650 MHz (1.1 GHz in turbo). It has a TDP of 45W and AMD said that this was the same for its Llano prototype. Note that AMD seems to be imitating the Intel mobile CPU nomenclature, by using larger figures.
AMD was of course looking to avoid a straight fight in terms of CPU performance, a fight which the Core i7 would clearly have won. AMD therefore accentuated a usage scenario. First of all a game was launched (benchmark Final Fantasy XIV) and was then left running in the background while a CPU load (Excel script), the playback of an HD video and an OpenGL rendering (SPEC Viewperf) were launched one after the other. While no precise performance figure was given, you can see that performance during processing of the Excel script was similar on both machines, while the 3D renderings were fluid on Llano and very jumpy on the Core i7. More importantly, the Llano machine remained very responsive, in contrast to the Core i7 one which was apparently pushed beyond its limits.
How best to interpret these results? We see several factors as coming into play: a) the Intel drivers are far from being effective in OpenGL and more generally when several graphics applications are running at the same time, b) loading the CPU and GPU in this way means that the Core i7 doesn’t benefit from Turbo and c) Sandy Bridge graphics performance can be significantly affected during high CPU use, as we observed in our report on this. On the Core i7, the last level cache and communication bus are shared between the CPU and GPU cores, whereas in Llano, the CPU and GPU are only linked by the memory controller. The scenario chosen by AMD was therefore very probably chosen to show Llano off in its best light and it’s therefore impossible to draw any general conclusions from the AMD demo.
Llano on the left, Sandy Bridge on the right.AMD also set up monitoring for total laptop energy consumption, saying that they did everything they could to avoid any differences external to the platform (SSD and identical memory). The screen, though similar, may however have impacted on the results. That said, overall, energy consumption levels were similar, with a slight advantage to the Llano. This advantage became much more significant when SPEC Viewperf was launched: we observed 56W for the AMD solution against 77W for the Intel. In the guise of a pure graphics test, AMD launched Furmark, which gave Llano a clear advantage. Although the energy consumption display was deactivated during this test, we were able to see, on the wattmetres behind the laptops, that this time the Llano consumed a lot more power: 52W against 37W for the Core i7. Obviously AMD didn’t care to dwell on this, but the score makes sense given their integration of a much more powerful GPU, which is after all the strong suit on this APU. The good news here is that AMD seems to have been able to attach sufficient CPU perforamnce to this GPU within a thermal envelope that is comparable to that of Sandy Bridge. Of course, this demo leaves us wanting more because, in fact, all it does is confirm what we know already: the Llano GPU is a good deal faster than the Sandy Bridge one. All this still has to be confirmed during real practical examples, as, after all, nobody will be doing their gaming while watching a film and carrying out 3D modelling at the same time! |
| |
|
Intel has just announced a new Atom for netbooks. Clocked at 1.66 GHz, the N570 comes in at the high end of Intel’s Atom Mobile range, superseeding the N550. A dual core version with hyperthreading (2 cores/4 threads), the Atom N570 is equipped with one MB of cache memory and has a TDP of 8.5W, like the N550 which was clocked at 1.5 GHz. The Intel Intel press release talks about maximum support for DDR3 667 MHz, a standard which doesn’t exist. The previous Atom models are limited to support for DDR3 800 MHz, the lowest clock authorised by JEDEC. |
| |
|
|
Copyright © 1997- Hardware.fr SARL. All rights reserved.
Read our privacy guidelines.
|
|
