What would it be like if we shrunk an old CPU down to modern manufacturing dies?
How much better would a P4 be at 14nm or 22nm?
>>55243765
It would perform exactly the same, as it's literally the same CPU just smaller.
>>55243778
Addition: The only benefits might of course be lower power consumption and better temps.
The basic benefits of shrinking dies is of course that you can manufacture higher complexity and more cores per square unit of space, at lower power and temperature ratings than would be possible with a larger die.
![Xeon_Phi_Family-2d76e04d73344ee3[1].jpg](https://i.imgur.com/wtmgtfs.jpg "Xeon_Phi_Family-2d76e04d73344ee3[1].jpg")
>>55243765
I've heard the Xeon Phi (pic related) basically has about 70 Pentium cores (don't remember if Pentium 3 or 4)
>>55243841
I thought they where small atom cores
>>55243765
Might be an increase in fmax depending on the given process, but over all it would be almost exactly the same, just at lower power.
There would be an impact on SRAM performance, as that is a property of the given process as well, not strictly design architecture. So the caches on a CPU could end up moderately faster.
>>55243841
They're Atom cores. The latest 14nm Phi is Silvermont arch.
>>55243857
They are.
>>55243778
>>55243810
>>55243882
Could it be overclocked to higher frequencies?
>>55243778
But wouldn't performance increase since electrons don't travel as far?
>>55243911
That depends on more factors than process. The architecture of an IC itself can inherently limit clocks.
If the arch could handle it, and the process could deliver an uplift in fmax, then yes.
>>55243841
Only the old Knight's Corner modified Pentium P54C-like cores. Newer ones are all Silvermont/Airmont.
>>55243911
I'm no expert but usually smaller = less heat
and with less heat you can overclock better and less power consumption but what's the point? You can get a Pentium G4500 that uses 14nm?
>>55243841
Original Xeon Phi was Pentium cores, the newer ones use Atom cores.
>>55243765
What do you think Intel has been doing? They haven't done anything major since the Pentium 3. They just add instruction sets, overclock it slightly, shrink it a bit, and call it new.
>>55244114
>>55244134
>bait when its the truth
>>55244134
He's right though
>>55243765
Might have a 4.8ghz base clock speed, yet may only performs as shit as a 3.5 ghz dual core vishera or 2 ghz celeron haswell
>>55244320
>>55244286
yeah gotta love those Pentium 3´s with, for example, hyperthreading
>>55246305
if you think a p4 at 4.8 ghz is anywhere near a 3.5ghz dual core vishara then you are retarded.
a p4 at 3ghz is slower than a single core of a 2ghz c2d. at 4.8, it might be able to keep up with 1 core of a bulldozer.
>>55247585
>heavily overclocked house fire Pentium 3
Kek
>>55247862
wew lad
>>55247585
>>55248106
>>55248135
Netburst DID have a few significant differences, for example the double pumped add/sub ALU's.
Yes the top SKU 3.8ghz Pentium 4 had a part running at 7.6ghz.
Yes the Pentium 4 that hit 8.5ghz had a part running at 19ghz.
>>55247585
>>55248135
NetBurst and P6 are not even remotely similar under the hood
>>55243951
.1/100000000000000000 is basically equivalent to 0 for all purposes.
