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: Greater Fool who wrote (65607)	7/15/1999 6:00:00 PM
From: Scumbria	Read Replies (2) \| Respond to of 1573458

GF, Without cache the Athlon would have pretty lousy performance, no? After all, it has a small onboard cache if I am correct. K7 has a large L1 cache (128KB). But without the L2, the system performance for integer benchmarks would probably be less than K6III systems. Scumbria

To: Greater Fool who wrote (65607)	7/15/1999 7:28:00 PM
From: kash johal	Read Replies (1) \| Respond to of 1573458

GreaterFool, Several points: 1. There is a minimum 12 week lag between wafer starts and product out. So they cannot jeopordize Q4 and will build a mix of K6 and K7 parts. Starting November they may well turn over hard to Athlon wafer starts. Depending on Dresden you could see 3-6m in Q1 2000 but I suspect Q2 is more likely due to slower dresden ramp. 2. Low cost Athlon. First producing 1 chip has major advantages as they can change product mix at packaging and adjust for yields and market demands very quickly. Secondly the K7 has a very large L1 cache so performance in integer should be close with similar speed Celeron and FPU likely will be faster than Celeron. Thirdly- once they get to 0.18 micron the die cost at 100mm2 will be in $10-30 range so who cares. 3. the next phase on AThlon will have integrated L2 cache. At that point they will likely market segment via L2 cache size. 4. The other major reason to run 1 design is the volume/yield curve. If you run 1M devices and 30% yield at 600Mhz then high margin parts may only be 300K units. If they run 3M and yield 1M at high speeds then the revenue stream is much more compelling and the higher margin on the extra 700K devices more than makes up for the extra die costs. Anyway, just my thoughts. Regards, Kash