AMD "Piledriver" refresh of Zambezi - info, speculations, test, fans

[TESKATLIPOKA]

undone

(btw ES Zambezi frequency is 15% lower than fx8150, so we could expect turbo up to near 4.7Ghz with retail Trinity.)

ES 3100mhz, turbo 4.1Ghz
Fx8150 3600Mhz, turbo 4.2Ghz
default clocks increased +500Mhz but turbo just +100Mhz.
and now here we have an ES Trinity 3.8Ghz, turbo 4.1Ghz
I think the best case is we will see a model with default 3.9-4.1Ghz, turbo 4.2-4.3Ghz, basically Trinity with FX4170 clocks but TDP 100W instead of 125W.
I am pretty sure turbo 4.6-4.7Ghz won't happen even if they released a 125W model, too much voltage needed, I don't think they are willing to set ~1.5V for max turbo even if they stayed in TDP.

[informal]

When it comes to piledriver,we can expect good news since:
1)AMD lists IPC and power improvements(~10-15%;uncertainty probably comes from process node ability at the moment of starting the production). This can imply around 5% IPC boost and 10% clock boost. If process allows they can boost clock further.
2)AMD stated they are working with MS on Win7 scheduler update.Win8 already supports optimal thread scheduling for BD. This can lead to another ~5% more performance over current level of Zambezi.

All summed up: from 15% to 20% improvement is quite possible with top end PD(8270?) vs FX8150. This would be roughly enough to put new PD model at around 2600K or 2700K level. This probably won't be enough to beat 8T IB @3.5/3.9Ghz but should be pretty close overall. Remember that 2600K/2700K level ,on desktop workloads, is overall very close to 980x/990x Westmere(which is still number two desktop chip out there).

[BeepBeep2]

When it comes to piledriver,we can expect good news since:
1)AMD lists IPC and power improvements(~10-15%;uncertainty probably comes from process node ability at the moment of starting the production). This can imply around 5% IPC boost and 10% clock boost. If process allows they can boost clock further.
2)AMD stated they are working with MS on Win7 scheduler update.Win8 already supports optimal thread scheduling for BD. This can lead to another ~5% more performance over current level of Zambezi.

All summed up: from 15% to 20% improvement is quite possible with top end PD(8270?) vs FX8150. This would be roughly enough to put new PD model at around 2600K or 2700K level. This probably won't be enough to beat 8T IB @3.5/3.9Ghz but should be pretty close overall. Remember that 2600K/2700K level ,on desktop workloads, is overall very close to 980x/990x Westmere(which is still number two desktop chip out there).

What about SB-E? You failed to even mention it

[informal]

What about SB-E? You failed to even mention it

Yes I didn't mention it since 6C SB-E is 550-900$ product. There is 4C variant also,but it's not faster than 2600K and cost the same or more(+ the additional cost of s2011 board).
So looking at 6C SB-E models,you will be looking at 2x-3x the price difference(CPU only;board prices difference adds on top of that) and 15-20% more performance on desktop. Just doesn't make any sense.

[BeepBeep2]

Yes I didn't mention it since 6C SB-E is 550-900$ product. There is 4C variant also,but it's not faster than 2600K and cost the same or more(+ the additional cost of s2011 board).
So looking at 6C SB-E models,you will be looking at 2x-3x the price difference(CPU only;board prices difference adds on top of that) and 15-20% more performance on desktop. Just doesn't make any sense.

They have absolutely no competition

[dirtyd]

They have absolutely no competition

In an exceedingly tiny, albeit high-margin part of the market

[BeepBeep2]

In an exceedingly tiny, albeit high-margin part of the market

You realize, if they had good competition...that would be mainstream

That tiny high margin part of the market is rather quite large when you consider that Intel owns 100% of it and 80% of the overall market.

(That was after Phenom II caught AMD up in market share, I'm sure it will slip back a few percent)