Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Intel Corporation (INTC) -- Ignore unavailable to you. Want to Upgrade?

To: Joe NYC who wrote (121284)	12/8/2000 5:14:37 AM
From: Joseph Pareti	Read Replies (2) \| Respond to of 186894

TO ALL please keep thinking clearly, despite of the instigation and misinformation. In times of crisis, rumors, shorts, saboteurs are popping up like mushrooms in a rainy, warm spring day. (i would have preferred a more explicit wording, but i experienced before that my posts were deleted by "referees") My opinion is they are paid to sling BS and they do it smart enough to even entice SI. The choice is yours : either - take cognizance Intel has made you rich and WILL keep making you rich or - believe blindly in these clowns, do exactly what they want YOU to do, and loose your shirt

To: Joe NYC who wrote (121284)	12/8/2000 2:18:03 PM
From: Tenchusatsu	Read Replies (1) \| Respond to of 186894

Joe, <The premise of AGP is that some textures are stored in main memory and accessed by graphics card from there, which is a complete joke, and shows a complete lack of foresight by the Intel designers.> That was the original premise, but hardly anyone uses that anymore, as I said before. Rather, AGP is used to download textures to local graphics memory. Didn't I mention this before? <Your 64 MB of textures are not realistic at all. Maybe the whole game may have that many of textures, not a scene, and a typical game on a PC with say 64 MB of main memory will have only a small fraction of this data in memory, so the theoretical transfer rate of AGP is irrelevant if the data is going from the hard disk. (Hmm, disk -> PCI disk controller -> PCI bus -> AGP bus -> AGP graphics card is probably even slower than disk -> PCI disk controller -> PCI bus -> PCI graphics card)> Data always moves from device to memory, or memory to device, (or sometimes device to CPU to memory and back again if you're talking about old PIO methods). Going from PCI disk controller to PCI graphics card is something that NO graphics card can do, nor is it something the graphics card is meant to do. (The graphics card is never meant to be a peer-to-peer I/O controller. It is only meant to bus master to gain access to memory, not another PCI device.) As for the 64 MB of local graphics memory, even that can use a fast AGP interface to memory. I don't know how games work these days, but I do know that the texture data in an entire game can easily exceed 64 MB. However, the textures are downloaded from main memory to the local graphics card on-the-fly as they are needed. The 64 MB of local graphics memory becomes a texture cache, but a cache still needs a fast bus to download data. I guess you could get around that limitation by downloading all of the textures in advance, but I don't think any application or game out there would ever do that. (By the way, a PC with only 64 MB of main memory will NOT have a graphics card with 64 MB of texture memory. Such a graphics card would only be paired up with high-end PCs, and they usually have at least 256 MB of main memory.) <Intel already had a spec for a general purpose bus in place that fulfilled all the realistic expectations of a graphics industry (not the idiotic ones of the AGP designers) in PCI 33/64 and PCI 66/64. Under the normal evolution, the graphics industry would drive the acceptance of these general purpose technologies, and a typical PC would today have a bunch of these slots instead of almost decade old PCI classic slots.> First of all, it's expensive to implement 64-bit PCI on a mainstream motherboard. Better to use the extra pins for a dedicated AGP slot, unless you are running a server. Second, 66 MHz PCI is limited to two slots on the bus, as far as I know. Plug in a graphics card, and all you have left is one remaining slot. PCI-X will allow for more slots on the bus at 66 MHz, but PCI-X wasn't around when Intel introduced AGP, and it still hasn't arrived yet. And third, PCI has remained "antiquated" because more and more functions are moving off of the PCI bus. This is better than cramming more devices and functions onto PCI, thereby necessitating a move to a more expensive and difficult-to-implement version of PCI. <From the technical perspective, AGP is an inexcusable piece of garbage.> I have yet to see ONE hardware enthusiast site agree with you these days. Meanwhile you might want to take a look at this Athlon workstation review: proe.com The Athlon 1.1 GHz workstation underperforms in graphics compared to the 900 MHz version. The review attributes this to the disabling of AGP in the 1.1 GHz workstation due to driver issues, and this results in a 50% drop in graphics performance. (Pro/E is a synthetic benchmark, but it's meant to reflect actual graphics workstation tasks.) In conclusion, I think your extremely negative view on AGP is based on faulty assumptions. I don't know if I'll change your view of AGP, but I hope I can correct some of your assumptions. Tenchusatsu