> In addition the k5 and cyrix all were way slow on MHZ but had better IPC than Intel.
This is a myth. Cyrix processors were much faster in some operations and much slower in others. K5 was a little better off in this regard, but there were still areas where P54 beat it in ipc.
> We had the infamous PR rating fiasco.
From a standpoint of realism, the PR rating was a fiasco for two reasons:
1) It was a lie, since it wasn't based on a general performance metric
2) It was a lie, because the manufacturers using it exaggerated to make themselves look better.
A Cyrix PR400 processor was slower than a 200MHz (equivalent) Mendocino/Celeron in some benchmarks. In the benchmarks that made the PR400 look good, it couldn't even meet or beat a Celeron-400 unless you crippled the Celeron.
If AMD and Cyrix processors truly and honestly met the Pentium in performance at their promised PR numbers, then I believe PR would not have been seen as such a fiasco.
> It seems likely that PIV will have raw Mhz lead.
Yes.
> Mustang may well win some benchmarks even with lower CLK.
> But its not clear to me why the PIV will not win the cream of the crop volumes as MHZ sells.
I think current conditions are looking better than you suspect for Mustang. I think it will be quite a while for P4 to be a big volume seller in that regard, and I think that overall performance will have an effect on the market.
Still, if P4 ramps as expected, Intel will keep their nameshare advantage.
> In addition high end buyers want a PC to be usefull for a while.
> With the world moving to sse2 - including hammer family - this on its own is a good reason to choose it vs an equivalent AMD chip.
I'm not sure I buy that. The world was moving to SSE last year, but I don't think forward-looking compatiblity was a hugely major element in PIII sales.
> So, AMD will have the middle of the market.
> But remember that mid 2001 intel is bringing enormous 0.13 foundry capacity to bear.
Edit: mid 2001 Intel is starting their 130nm ramp. You're making a big leap of faith in assuming that the first 130nm fab will suddenly do 130nm at breakthrough yields at 100% capacity. It takes time to ramp a new process.
> Couple PIV with SDRAM, DDR and u have a killer combination for 2H 2001.
A high performance combination, of course. But if the 130nm ramp starts midyear (that's when the first product is said to be coming out), then Q3 is way too early to start counting immediate AMDemise. Maybe AMD will see a little weakness in Q4. Of course, they'll be bouyed by increased Flash sales and -- if they're smart about it -- strong presence in the mobile market, in which Intel will likely be very weak in 2001.
-JC |
