what is the point of the 2.8ghz prescott??

[Hugo Drax]

It benches slower than the northwood 2.8, runs hotter and the L1/L2 has
higher latencies than the northwood cache. Why even FAB such a CPU? I
thought the point of prescott was to fab high MHZ processors like 3.4+ where
eventually RAW clock cycles make up for the slow architecture in terms of
instructions per clock cycle.

 

[JT]

It benches slower than the northwood 2.8, runs hotter and the L1/L2 has
higher latencies than the northwood cache. Why even FAB such a CPU? I
thought the point of prescott was to fab high MHZ processors like 3.4+ where
eventually RAW clock cycles make up for the slow architecture in terms of
instructions per clock cycle.

Got to do something with those chips that won't pass the tests at 3.4.
Sell them to unsuspecting consumers to pay for lower yeilds

JT

 

[somebody]

It benches slower than the northwood 2.8, runs hotter and the L1/L2 has
higher latencies than the northwood cache. Why even FAB such a CPU? I
thought the point of prescott was to fab high MHZ processors like 3.4+ where
eventually RAW clock cycles make up for the slow architecture in terms of
instructions per clock cycle.

You're right about 2.8GHz prescott not making much sense for the
customer.
There's several good answers to your question though.
The first is that my guess is that these 2.8s trickles out of the same
process that attempts to produce 3.4GHz Prescotts.
Might just as well try selling them.

The point of the 3.4GHz Prescott can be questioned too, of course.
But in the case of Intel, it's very little a case of performance or
value. It's all about brand name recognition, "GHz" and perceived
properties.
Consider Celerons and Xeons for instance. They're are grotesquely
outclassed by AMD. It's almost like comparing cpus from different
decades!
Why is _ANYBODY_ buying them?
The simple truth is that the market is mostly clueless, and Intel can
make money on that, so they do.

ancra

 

[Walt]

I am not sure myself, but wondering is it the HT being enabled and
the additional SSE3 instructions? If so, then does one just need
to wait until MS ships such support to make it worth while?

 

[Hugo Drax]

Got to do something with those chips that won't pass the tests at 3.4.
Sell them to unsuspecting consumers to pay for lower yeilds

JT

hehe I suspect you are right. Marketing is dictating engineering.

 

[Hugo Drax]

per clock cycle.

You're right about 2.8GHz prescott not making much sense for the
customer.
There's several good answers to your question though.
The first is that my guess is that these 2.8s trickles out of the same
process that attempts to produce 3.4GHz Prescotts.
Might just as well try selling them.

The point of the 3.4GHz Prescott can be questioned too, of course.
But in the case of Intel, it's very little a case of performance or
value. It's all about brand name recognition, "GHz" and perceived
properties.
Consider Celerons and Xeons for instance. They're are grotesquely
outclassed by AMD. It's almost like comparing cpus from different
decades!
Why is _ANYBODY_ buying them?
The simple truth is that the market is mostly clueless, and Intel can
make money on that, so they do.

ancra

Sad how marketing has dictated what engineering should do. If they added the
(Barrel shifter, improved branch predictor, IMUL improvements, smaller 90nm
fab) to the northwood you would have a real nice performing part and cooler.

 

[Hugo Drax]

I am not sure myself, but wondering is it the HT being enabled and
the additional SSE3 instructions? If so, then does one just need
to wait until MS ships such support to make it worth while?

Its slower because the pipeline was increased from 20 to 35 so 1
misprediction slows down the whole thing even more, also the L1 and L2 have
much more latency for some reason. I think 4.3Ghz is where Prescott will
have to go.

 

[somebody]

Its slower because the pipeline was increased from 20 to 35 so 1
misprediction slows down the whole thing even more, also the L1 and L2 have
much more latency for some reason. I think 4.3Ghz is where Prescott will
have to go.

- Well, branch prediction _does_ seem to be better on P4E. Larger
cache too. But I suppose it only averages at making the penalties
even.

ancra