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.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Patrick Grinsell who wrote (41020)	11/7/1998 7:52:00 PM
From: Scumbria	Respond to of 1574005

Patrick, the majority of current applications can fit all the textures in the local memory of a Voodoo2 The main memory -> graphics bandwidth issue is going to become critical in the near future. The arrival of multitexture applications will require significantly higher bandwidth. Running multiple 3D windows at the same time will stress memory bandwidth as well. for games such as Unreal and Quake2 the CPU is currently the bottleneck and not the accelerator. Don't be surprised to see hardware geometry engines starting to replace the CPU functions. Matrix multiplies can be done very efficiently on special purpose hardware. Scumbria

To: Patrick Grinsell who wrote (41020)	11/7/1998 9:53:00 PM
From: Steve Porter	Respond to of 1574005

Patrick, et al: RE: SLI Voodoo2 fill rate constrained.. I can fill rate constrain a voodoo2 on a PII 266.. it all depends on how much work you have to do for each polygon. I.e. do I do a lot of transformations on the CPU or just a couple and then hand it to the videocard.. Steve PS. If an application is fill rate constrained on a Voodoo 2 SLI it needs to be re-written, cause it's awfully in effective!

To: Patrick Grinsell who wrote (41020)	11/8/1998 1:09:00 AM
From: Ali Chen	Read Replies (1) \| Respond to of 1574005

Patrick, re <bandwidth and bottlenecks> The article you cited is a dilettante's rambling. I do not know what the "fill constrained" is, as per your reference to some Intel guy. The 3-D pipeline usually can be 1) "geometry-transform-constrained" (mostly CPU/FPU, where 3D-Now is of tremendous help), or 2) "triangle-setup-constrained" (where PCI latency on writes to non-cacheable areas are of major problem), or 3)"rendering-constrained" (where the accelerator paints colored and textured trianges. The balance between these three parts is heavily dependent on the size of triangles, or, to be more correct, on the spectrum of triangles for the particular scene content. AGP has nothing to do with most of the constraints except "rendering". The enhancements in K6-2 Atiq Raza was talking about are in the area of Write Combining, which helps to combine data directed to non-cacheable areas of video accelerators, mostly for "triangle set-ups". The Katmai has no visible advantages to K6-3 (with 3DNow and 256k on-chip L2) in any of these areas. - Ali