
News
by Anna Filatova
Well, it’s no secret that to connect the chipset’s Bridges HyperTransport bus licensed by AMD will be used. So, we won’t focus on that. Meanwhile, the second issue in the agenda provokes a lot of thoughts. That’s what we will dwell on.
NVIDIA said that it would equip Crush with a two-channel access to the memory bus. The 128-bit bus will offer a theoretical bandwidth of 4.2GB per second. Actually, this phenomenon has already been mentioned by some non-official sources, so it makes no surprise now.
As you remember, first based on NVIDIA Crush mainboard by MSI has three DDR SDRAM DIMM slots (we reported it early). We would like to draw your attention to the number of memory slots: there are three of them instead of traditional two or four. Considering the bandwidth of one DDR SDRAM module being 64-bit, it won’t be hard to estimate that at least two modules will be needed to provide 128-bit access. Then, what is the third «single» slot for?
So far we can bring only one reason to explain it. Perhaps, the third memory module will be «privatized» by the graphics. This way, if the assumption is true, Crush can’t be called a SMA chipset (Shared Memory Architecture, i.e., using part of the system memory as graphics memory).
What are we driving at? On the one hand, the architecture implemented in Crush will probably yield some performance gain in comparison to SMA. But on the other one, it will hardly make any sense to buy a pricy (presumably $45 for the chipset) mainboard with integrated graphics which is invalid without an additional DDR SDRAM DIMM. It sounds more reasonable to have a board without integrated video and couple it with a regular GeForce2 MX-based graphics card. Besides, these cards shouldn’t be too expensive by the time when this chipset is launched (the third quarter of 2001).
Furthermore, a «non-onboard» graphics card built on GeForce2 MX with 128-bit 166MHz SDRAM is sure to be faster than Crush’s on-board core with 64-bit DDR SDRAM clocked at 266MHz (PC2100 DDR SDRAM). It’s a proven fact that a mainboard with 64-bit DDR SDRAM even working at a higher frequency of 286MHz (Creative GeForce2 MX) fails to catch up with a standard board with 128-bit 166MHz SDRAM in performance.
Our only idea as to why NVIDIA has resorted to such a freaky approach in developing its integrated chipset is that architecture of the kind is comparatively simple. It is much easier to introduce a «private» memory module for the graphics than to take the trouble sharing them among the CPU, the graphics chip and RAMDAC.
Anyway, the future of NVIDIA Crush has turned much vaguer.
Discussion
Latest News
Friday, November 21, 2008
2:29 pm | IN BRIEF: Qimonda Accuses Seagate, LSI of Patent Infringement. Qimonda Asks ITC to Investigate Alleged Patent Infringements by LSI, Seagate
6:05 am | Sales of Blu-Ray Disc Players Below Expectations, Movie Prices Should Get Cheaper – Sony. Sony Admits: High BD Movie Prices Curtail Sales of Blu-Ray Players
4:12 am | AMD to Release Quad-Core Processor for Notebooks in 2010. AMD’s “Champlain” CPU Set to Emerge in 2010
Thursday, November 20, 2008
3:21 pm | Foxconn Rumoured to Take Over Pegatron, Quit Channel Business. Foxconn May Sacrifice Channel Business for Contract Manufacturing Capacity Expansion
1:43 pm | Intel Plans to Release Discrete Larrabee Graphics Processors in 2009 or 2010. Intel Unsure When to Release Its Standalone Larrabee Graphics Chip
12:09 pm | S3 Graphics Reveals New-Generation Graphics Processor. S3 Unveils GPU with Built-In Audio Core, Starts to Sell Chrome 530 GT Graphics Card
8:29 am | Blu-Ray Disc Players’ Prices Fall Below $150. Relatively Inexpensive Blu-Ray Players Finally Emerge
4:24 am | JEDEC to Set Solid State Drive Standards in 2009. JEDEC Plans to Standardize SSDs Next Year
3:00 am | Elpida May Gain Manufacturing Capacity or May Slowdown Its Further Expansion. Elpida May Plan Surprising Acquisition of Powerchip or ProMOS





