tdp and turbo on intel S chips

dosmastr

Junior Member
Feb 21, 2006
20
0
0
ok so google hasn't helped me.
toms forums also hasn't had any info
emailed anand and kyle bennet (no reply yet, they are both busy guys I imagine :) )
even intel's own tech support couldn't provide an answer (ugh... they got stuck on "TDP isn't the same as speedstep".... yes I know...)


TDP = maximum heat output
Maximum heat output would happen under turbo speed

so TDP= turbo speed heat output (most likely right?)

4570 TDP=84 watts, turbo speed=3.6Ghz
4570S TDP=65 watts, turbo speed=3.6Ghz

Its the same core, all that appears different is base speed which shouldn't effect MAX heat output, but rather only effect TYPICAL heat output. Everything I've read on TDP is that it mean MAX not typical.

I really would love to know how they do it.
All I can see is that the chip is either binned and the lower heat parts that could be K's (better overclocking if it just makes less heat to begin with) get marked as S
OR
The S parts throttle sooner to stay under 65watt output in which case the turbo speed is kinda a lie.
 

dosmastr

Junior Member
Feb 21, 2006
20
0
0
secondly, does lower heat equate to lower power usage since its doing the same work but converting less energy into heat?
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Different workloads generate different amount of heat. The turbo as such is not garanteed for anything, while it usually works at peak levels.

The S model is simply limited to 65W with a lower baseclock as well. That means it will run slower. But during some workloads it might run just as fast.

All power that goes into a CPU comes out as heat. Cept 0.0something% that goes out as radiation/emi.

The S and T chips are designed for cases where you need the lower TDP max. For example a tiny PC, passive cooling and so on.
 
Last edited:

dosmastr

Junior Member
Feb 21, 2006
20
0
0
I get all that, but it doesn't do anything to answer the quandary I have.

unless it drops to base clock more frequently than the non s part, how could it generate less heat on its worst day than a chip exactly the same except a higher base clock on exactly the same loads?
 

videogames101

Diamond Member
Aug 24, 2005
6,777
19
81
The max power draw won't be under a turbo load (one core), it will be when all cores are fully loaded.

Everyone knows this.
 

Concillian

Diamond Member
May 26, 2004
3,751
8
81
The max power draw won't be under a turbo load (one core), it will be when all cores are fully loaded.

Everyone knows this.

ding! Winner.

The assumption in the original post of 'max TDP at max turbo' is incorrect.
Turbo happens when all cores are not needed, one or two cores turbo. TDP is determined by the nominal speed, which is the speed when all 4 cores are in use.

So you're comparing 2.9 vs. 3.2 GHz when comparing the speeds for TDP.
 

dosmastr

Junior Member
Feb 21, 2006
20
0
0
ding! Winner.

The assumption in the original post of 'max TDP at max turbo' is incorrect.
Turbo happens when all cores are not needed, one or two cores turbo. TDP is determined by the nominal speed, which is the speed when all 4 cores are in use.

So you're comparing 2.9 vs. 3.2 GHz when comparing the speeds for TDP.

THANK YOU!

the reason I posed all the premises of the argument was so someone could plainly point out:
"well here's your problem, this is a wrong assumption"

I was figuring that the cpu only hits turbo under max load (also incorrect if multithreaded app in use) and max load = turbo=tdp

guess I forgot turbo isn't simply overclocking of the whole cpu.
 
Last edited:

dosmastr

Junior Member
Feb 21, 2006
20
0
0
so now that answers the second question, yes the S part will be using less juice... but may or may not be doing more work per watt
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,691
136
so now that answers the second question, yes the S part will be using less juice... but may or may not be doing more work per watt

Yes, that's about it. The reduced TDP versions have a lock that says to the CPU: "you can draw this amount of power, but what you spend it on is up to you". If you use Aida64 it has a feature where you can see what the CPUs maximum/sustained power draw is limited to.

I wouldn't bother with a reduced TDP CPU for desktop use, but if you're cooling/power limited (f.x. SFF/Pico PSU), they are very useful.
 

dosmastr

Junior Member
Feb 21, 2006
20
0
0
don't plan to be cooling limited (trying/failing to get 212 evo for under 30...or even 33) but do want to keep it as quiet and cheap to run as possible as I'm upgrading from a poor mans 2007 (e2180 3gb ram) pc and plan to keep this one just as long. (gpu and ssd and win 7 have been added though... didn't want to stay in dark ages)


LOL nice forum handle
 
Last edited:

ElFenix

Elite Member
Super Moderator
Mar 20, 2000
102,414
8,356
126
don't plan to be cooling limited (trying/failing to get 212 evo for under 30...or even 33) but do want to keep it as quiet and cheap to run as possible as I'm upgrading from a poor mans 2007 (e2180 3gb ram) pc and plan to keep this one just as long. (gpu and ssd and win 7 have been added though... didn't want to stay in dark ages)


LOL nice forum handle

don't get an S or T part unless you have severely restricted cooling capability.

the only thorough review of S and T parts i've seen done was silent pc review's

now, this is sandybridge rather than the latest haswell, but it all still works the same so this should be instructive. as you can see on page 3, at idle and for the same load (h264 playback), all the sandybridge processors draw the same power. at full load, the S and T parts will draw less power, but keep in mind that a regular part does more work in that same timeframe, so the full load for each part are different.

the interesting chart is on page 6. as you can see, the S and T parts actually draw more total power to do the same thing because they're slower (though, the difference is outside the significant figures so the parts might as well be the same).

if you want to save electricity and money up front, seems the best thing to do is to find a motherboard that allows undervolting and keep with the stock clocks.
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,691
136

BSim500

Golden Member
Jun 5, 2013
1,480
216
106
As others have mentioned, the turbo ratio settings are different for S & T chips. More of a "gap" between 1-4T loaded. The i5-4570T is also a DUAL-CORE (2C/4T) like an i3. Eg:-

4-thread loads = 3.4 (4570) / 3.2 (4570S) / NA (4570T)
3-thread loads = 3.5 (4570) / 3.3 (4570S) / NA (4570T)
2-thread loads = 3.6 (4570) / 3.5 (4570S) / 3.3 (4570T)
1-thread loads = 3.6 (4570) / 3.6 (4570S) / 3.6 (4570T)

http://www.intel.com/support/processors/corei5/sb/CS-032278.htm

No point paying a premium for that either as you can easily mimic the lower TDP by under-clocking & under-volting, manually adjusting multipliers per load in the BIOS (if your motherboard supports that), or playing around with the SpeedStep settings in Control Panel -> Power Options (Maximum processor state). If you were thinking about the T chips, there's also no point in getting an i5-4570T vs undervolting an i3-4340.
 
Last edited:

Qwertilot

Golden Member
Nov 28, 2013
1,604
257
126
Maybe I'd have to dissent slightly. From Sandy to Haswell they've been spending huge efforts in optimising for lower power operation and things could easily have changed. Far from impossible that 30/40W might not be Haswell's 'natural range' for power draw.

The 35W i7-4675T was benchmarked quite comprehensively, if not explicitly, in the Anandtech review here: http://www.anandtech.com/show/7604/asrock-m8/5

Definitely quite a low power draw under load (in one of the early entries) and quite nice CPU performance, not really seeming to bottleneck anything etc. If that carries over to the other Haswell S/T chips they might be quite a sensible choice.