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 : 3DFX -- Ignore unavailable to you. Want to Upgrade?

To: Joe C. who wrote (5257)	7/13/1998 7:48:00 AM
From: Joe C.	Read Replies (1) \| Respond to of 16960

CPU Madness has an article on the TNT that's basically saying that the TNT really does have the potential nVidia claims and people should not be so quick to jump to conclusions. Michael Fleming questions their article altogether by saying that some of their technical assumptions may be incorrect. CPU Madness cpumadness.com Michael "Got2MuchTime" Fleming Response: <<I just mailed this off to the people at CPU Madness and thought you might be interested. ----------- Hello, my name is Michael Fleming - I write on 3D issues on my site, 3D Over-Time (over at Voodoo Extreme). I read your article (lies.htm) on the TNT and latency issues, and a few things seemed odd to me. In particular, the claim that, since the TNT performs about as well in the Massive Q2 benchmark as in demo1, while V2 SLI drops by around 18% (it's slower on Massive), V2 SLI is fillrate limited and the TNT is limited by drivers or something else external. The quote is: "Looking at Anandtech's benchmarks of the RIVA TNT at 1024x768 in demo1 and massive, we will notice that the performance of the TNT decreased 2 fps when moving to massive, while the performance of the Voodoo2 SLI 3Fingers ran at 1024x768 decreased ~12fps. This is a perfect example of how the TNT really does have the potential nVidia claims. The Voodoo2 falls a significant amount of FPS, showing that it is indeed being fill-rate limited as soon as the jump from demo to massive is being made. TNT, on the other hand, seems virtually unaffected. The slight difference in fps when switching from demo1 to massive (with the TNT) shows that the RIVA TNT is not being fill-rate limited, but being limited by some exterior resource, i.e. latencies, caused, most likely by poor drivers." I believe that whoever came up with this conclusion is, in fact, completely incorrect. The correct interpretation of these numbers, in my opinion, is quite the opposite. Massive is a harder benchmark on most systems because it _requires more CPU_ in the form of geometry and lighting, clipping, collision, etc. It is not significantly harder on a card's fillrate than demo1 at the same resolution. The V2 SLI slows down going to Massive from demo1 because it becomes _CPU LIMITED_, not, as you state, fillrate-limited. The TNT, on the other hand, is clearly _not_ CPU limited, since Massive's increased CPU load has virtually no effect on the numbers. Although other factors may also play a role, it appears that it is the TNT which is fillrate limited, being unable to keep up with the CPU (which can clearly transform, light, clip, etc. at least 57 fps-worth of material - _much_ more than the 32 fps displayed by the TNT). In summary, I request that you change your article to reflect this, since you base a fair part of the article (and thus its conclusions) on this erroneous (and, in fact, completely backwards) piece of logic. Of course, if I have flipped a cog, just write me back explaining where I am wrong :-). Thanks! - Michael http://voodooextreme/3Fingers/3D_OverTime -------- p.s. (I didn't send this in the email) Their general point about latency issues also doesn't make sense to me - The CPU hardly sends pixels, it sends triangles, and the number of pixels per triangle varies with resolution (all other things being equal), so I don't see how their calculated "CPU cycles/pixel" can be valid. sigh >> Joe C.

To: Joe C. who wrote (5257)	7/13/1998 9:00:00 AM
From: louis mason	Read Replies (2) \| Respond to of 16960

Maybe this is an explanation for what wallstreet is thinking in not treating TDFX like it should: Shift to Low-Cost PCs Will Hurt Suppliers of Processors and Graphics Chips, New Report Says Market for New Class of Integrated Products Will Total $1.6 Billion in 2000, Force Shakeout in Traditional PC Components Markets, According to New Report From Mercury Research SCOTTSDALE, Ariz. July 13 /PRNewswire/ -- The combined markets for traditional PC components like microprocessors, graphics accelerators, system logic and sound chips peaked in 1997 and will decline in '98 and beyond, according to Rearchitecting the PC, a new report released today by Mercury Research, Inc. The shrinking market is due largely to the fact that the dramatic shift to lower-cost PCs is forcing average selling prices for PC components dramatically lower. In 1998, component suppliers will ship $21.7 billion worth of processors, graphics controllers, chip sets and sound chips to PC suppliers. That figure is 4 percent lower than '97 shipments of $22.6 billion, according to data found in Rearchitecting the PC. The lower prices and declining markets are spurring investment in a new class of more cost-effective, highly integrated PC components -- a market that will grow to $1.6 billion in the year 2000, according to the report. Unfortunately for existing suppliers, however, the additional revenue generated by the integrated components will not be enough to reverse the decline in the market for traditional PC chips -- making a shakeout inevitable. ''The performance treadmill the component suppliers were on through much of the '90s is reaching a dead end,'' said Mike Feibus, a principal analyst at Mercury Research. ''With the high end of the PC market shriveling, chip makers can no longer profit by designing high-performance parts and riding them down the price curve.'' ''Going forward, the survivors will offer low-cost, highly integrated PC components which enable computers that are both attractive to buyers and profitable for suppliers,'' added Dean McCarron, also a principal analyst at Mercury Research. ''Now that the PC platform is stabilizing, integration is finally a viable avenue for achieving lower cost and higher performance.'' l mason