2013 i5 3.4 vs i7 3.5 temperature for given load

[propower]

Update:

Me saying there is a thermal issue does not equal them saying it is so But no way I am keeping the BTO i7 with the way it has tested for me!

New i5/256SSD/16G ram ordered tonight i7 will be returned when the new one arrives.

Note: Getting the 16G ram option on the 2013 imac gets two 8G sticks instead of 2 4G sticks (if it was four 4G sticks that would suck!). $200 is not that far off from 16G on the retail market (OWC about that delivered). Still doesn't account that for the same dough yo could upgrade to 24G on your own... but -- now I'm only two sticks away from maxing it out someday

I sure hope the new i5 is as good as my test i5 machine was!

Well that was then !!!

OK - for those who don't know the free Apple app XCODE has a preference panel that lets you turn Hyperthreading on or off whenever you like! So: an i7 can be put in"i5" mode. After much reading and testing I found that my prime application isn't better with an i5, it has a HT bug that shows up right where I want to use it. Thus getting the i7 and putting it in i5 mode for this application (at least until a real fix comes along) gives me all that the i5 machine would and lets me have all the i7 capabilities when my prime software gets up to speed (and any other uses that can take advantage).

So: I cancelled my i5 order and now have this coming...
3.5GHz i7, 256SSD, 16G apple ram, 775GTX 2G

But back to the thread topic!!

I retested my first BTO 3.5GHz i7 with HT on and off and the "Yes dev/null" load (see first few posts).

The current i7 I have really runs a lot hotter than the i5 I tested whether HT is ON or OFF!

The results are even just a little higher than post #3.
One instance gets ~80degC peak
Two instances gets 95degC pk and fans around midway!!!
The Late 2013 3.4GHz i5 I tested earlier (don't have anymore) was 15 to 20degC cooler for the same load!
I still believe there is something wrong with the heatsink attach on this one but would love confirmation from other data points...

SO: Any one else with either the new i7 or i5 late 2013 imac willing to test at one & two instances of this stress test and report back temps and fans?

 

[Nirurin]

Well that was then !!!

OK - for those who don't know the free Apple app XCODE has a preference panel that lets you turn Hyperthreading on or off whenever you like! So: an i7 can be put in"i5" mode. After much reading and testing I found that my prime application isn't better with an i5, it has a HT bug that shows up right where I want to use it. Thus getting the i7 and putting it in i5 mode for this application (at least until a real fix comes along) gives me all that the i5 machine would and lets me have all the i7 capabilities when my prime software gets up to speed (and any other uses that can take advantage).

So: I cancelled my i5 order and now have this coming...
3.5GHz i7, 256SSD, 16G apple ram, 775GTX 2G

But back to the thread topic!!

I retested my first BTO 3.5GHz i7 with HT on and off and the "Yes dev/null" load (see first few posts).

The current i7 I have really runs a lot hotter than the i5 I tested whether HT is ON or OFF!

The results are even just a little higher than post #3.
One instance gets ~80degC peak
Two instances gets 95degC pk and fans around midway!!!
The Late 2013 3.4GHz i5 I tested earlier (don't have anymore) was 15 to 20degC cooler for the same load!
I still believe there is something wrong with the heatsink attach on this one but would love confirmation from other data points...

SO: Any one else with either the new i7 or i5 late 2013 imac willing to test at one & two instances of this stress test and report back temps and fans?

My i7 arrived today, and is still in its box cos I need to make desk space for it lol.
But when i have it up and running I will run this test (once I figure out how to)

 

[woddy]

Very interesting!

Could you possibly run only 4 and see what fans and temps you get? Be really great if you could run 2 as well but any additional data is great. Thanks!!

so again: i7/256SSD/gforce780

4 instances for 10 min: hovering around 1400 to 1700 rpm
2 instances for 10 min: 1200 rpm (minimum)

all measurements done at 20°C / 68°F room temperature. I guess this is an important fact to mention, as it undoubtedly influences the result.

HTH

 

[propower]

so again: i7/256SSD/gforce740

4 instances for 10 min: hovering around 1400 to 1700 rpm
2 instances for 10 min: 1200 rpm (minimum)

all measurements done at 20°C / 68°F room temperature. I guess this is an important fact to mention, as it undoubtedly influences the result.

HTH

That's great Woddy! Gforce 740??? I think you meant 780m right?
Did you happen to have the CPU temp as well?? So far I think Temperature Gauge pro (demo) is all that will register Haswell iMacs - not istat yet...
4 instances would peg mine at max fan for sure!

 

[joema2]

...the free Apple app XCODE has a preference panel that lets you turn Hyperthreading on or off whenever you like!...
SO: Any one else with either the new i7 or i5 late 2013 imac willing to test at one & two instances of this stress test and report back temps and fans?

I've seen several posts saying that turning hyperthreading off in the XCODE preference panel no longer works on 10.8. Have you verified it works on the current OS X version?

My wife has a 2012 i7 3.4Ghz iMac and I have a 2013 i7 3.5Ghz. I'll be happy to run the stress tests but am busy so it make take a few days.

Re hyperthreading, no specific API calls or programming methods (beyond having multiple threads) are needed to use hyperthreading. It is automatically used for any multithreaded app.

That said, for programmers working in C++ at a low level, there are some methods and compiler directives which can improve utilization of hyperthreading. Also at compile time, code generation options can be used which improve performance on a hyperthreaded CPU. There is much detailed info in this Intel white paper: http://cache-www.intel.com/cd/00/00/01/77/17705_htt_user_guide.pdf

For a given program, the end user doesn't know how that was handled. So what counts is a real-world controlled test.

On most Windows PCs you can disable hyperthreading in the BIOS. This facilitates running tests with HT on and off.

Macs used to allow turning HT on/off via either a terminal sudo command or the Processor Pane in Xcode, however these apparently no longer work on 10.8. I personally tried the terminal command "sudo nvram SMT=0" on 10.8.5 and it definitely does not work.

On my Win7 PC with an i7-875K @ 4Ghz, 64-bit Geekbench 3 does about 9,000 with HT off, vs about 11,000 with HT on. Likewise exporting an H.264 high-def video in Premiere Pro CS6, the difference is 28 sec (HT off) vs 24 sec (HT on), roughly 15% difference improvement. All tests repeated three times for consistency.

In general if you examine Activity Monitor and it shows high CPU (over 60-70%) on all available cores, then HT might be of benefit. Lots of common activities such as Photoshop HDR, video transcode, video render, etc. are heavily multithreaded and CPU intensive. You can examine the CPU activity of all cores in Activity Monitor, by selecting Window->CPU Usage.

That said, the % of time your machine needs over four cores can be very small. Thus prioritizing an an i7-based machine for hyperthreading alone may not be the best choice. On a top-spec 2013 iMac 27, there's only a $200 difference between i5 and i7. But the difference between 1TB regular vs Fusion Drive is *also* $200, which is likely a much better investment.

In general the trend in software development is increased threads and parallelization. If your application profile falls into that region, paying $200 more on an top-spec iMac to get (maybe) 20% more performance is a good deal. But across many real-world performance scenarios, you'll likely see more noticeable differences from I/O improvements, which is what you already stated.

 

[propower]

XCODE application called instruments is the one I use (R click XCODE - show contents - applications - instruments - panel opens hit cancel - go to preferences and there you get the HT on off and how many cores).

Appears to work perfectly in 10.8.5 - dev/null test - activity monitor and my Application ProTools.

For pro tools there is a large advantage to using HT when using a lot of software instruments and dense sessions. For me, I am all about lowest possible throughput latency. This puts me at high sample rates (96kHz) and low what are called audio buffers. The difference between HT on and Off for this case is substantial (as in NO HT much better).

For most normal operations - even PS and imovie I don't think it hardly would be noticed. Over the next few years for what I am doing - not essential - but can very likely come in handy. For $200 extra - not an issue for me - as long as for some crazy reason all i7's don't run so much hotter that fan noise will be a bother (again - for me - I hate any fan noise!). Hence why this thread is about temperature at load Though I am fascinated by the intricacies of HT. As in - on my app - how much core bandwidth is really available for the HT engine to pile more stuff through the same core -- another thread though eh?

 

[woddy]

That's great Woddy! Gforce 740??? I think you meant 780m right?
Did you happen to have the CPU temp as well?? So far I think Temperature Gauge pro (demo) is all that will register Haswell iMacs - not istat yet...
4 instances would peg mine at max fan for sure!

yes indeed, mixed up the gforce

so 4 instances of yes at the beginning:

after 11 minutes:

HTH

 

[propower]

COOL -- VERY Helpful !!!! The proc is hot but is regulated right where apple wants it under load - ie no increase in fans till above 90degC. It is impressive how low your fan is running to maintain it there! For me I get to about this place at 2 instances!

At the risk of trying your patience /good will any way you could also post a final shot of where two instances gets to on yours?

----------- Four Yes's for me and then 2 Yes's (on 2Y the fan goes between 1500 and 2300)

-----------Last picture is 2Y on the 3.4G i5 imac! what a difference.... THIS is what I am hoping 2Y on a "good" i7 should look like!!!!

 

[Nirurin]

Ok here is my test

Not sure if I have uploaded these pics properly though, I've not added pictures onto this forum before.

3 screenies, one of temps before the test, one after 10 minutes of 4-Yes, and one after 10 minutes of 8-yes.

Hopefully this is helpful, and hopefully you guys can tell me if my iMac is defective in some way!

 

[propower]

Hey Nirurin ! Great stuff MUCH appreciated! It seems you are right in line with woddy.

I hate to ask but I would love to see what two instances looks like for you too. Look two posts up and see the difference between 2Y on the new i5 and the new i7. I am really hoping my first i7 is not normal. It's replacement may be here by Friday.

I am letting my four instance one go the full 10 minutes - maybe it will level out like yours - Nope - mine stayed at full fan peaked at 99degC

 

[Chippy99]

Here's my late 2012 i7/680MX with 2 Yes instances running for 10 minutes:

I have to say I am quite pleased with this

 

[Nazmor]

My 4 yes test
i7 780m

Looks like fan kicks in earlier and stay higher comparing to others (Zimmerman,woddy,Nirurin). Just wonder what that really means. Why if others got similar average 83degC and cpu cores at 90+ (close to max) their system did not tried to lower it using full fan speed? And when actually their fan would go full speed??, if at 100+degC that would be very small margin to max. Isn't it that cpu core high temperature actually speeds up fan? If so woody and Nirurin got 93-94degC on cores and their fan still did not went faster keeping low rpm. Really strange.

 

[Krevnik]

My 4 yes test
i7 780m

Looks like fan kicks in earlier and stay higher comparing to others (Zimmerman,woddy,Nirurin). Just wonder what that really means. Why if others got similar average 83degC and cpu cores at 90+ (close to max) their system did not tried to lower it using full fan speed? And when actually their fan would go full speed??, if at 100+degC that would be very small margin to max. Isn't it that cpu core high temperature actually speeds up fan? If so woody and Nirurin got 93-94degC on cores and their fan still did not went faster keeping low rpm. Really strange.

Out of curiosity, do you have the fan boost option on in the app? It will alter fan behavior and push them to max much earlier.

The default behavior seems to only start ramping up at 90C, reaching max before 100C.

 

[propower]

My 4 yes test
i7 780m

Looks like fan kicks in earlier and stay higher comparing to others (Zimmerman,woddy,Nirurin). Just wonder what that really means. Why if others got similar average 83degC and cpu cores at 90+ (close to max) their system did not tried to lower it using full fan speed? And when actually their fan would go full speed??, if at 100+degC that would be very small margin to max. Isn't it that cpu core high temperature actually speeds up fan? If so woody and Nirurin got 93-94degC on cores and their fan still did not went faster keeping low rpm. Really strange.

Yours looks like mine!

First - could you please repeat this at 2-Yes's?

Second- my fear is that the higher fan speed like what I have is an indication of the processor not being attached to the heat sink as well as it can/should be. Not bad enough to be a fail but not good enough to make it run where the fan does not have to work near as hard. I am ALL about low fan noise (and keeping the thermal stress on the proc as low as possible) so this is very important to me!

 

[woddy]

At the risk of trying your patience /good will any way you could also post a final shot of where two instances gets to on yours?

here we go: 2 instances of yes: after 1 minute

after 10 more minutes:

 

[Nazmor]

Yours looks like mine!

First - could you please repeat this at 2-Yes's?

Second- my fear is that the higher fan speed like what I have is an indication of the processor not being attached to the heat sink as well as it can/should be. Not bad enough to be a fail but not good enough to make it run where the fan does not have to work near as hard. I am ALL about low fan noise (and keeping the thermal stress on the proc as low as possible) so this is very important to me!

2x yes, screenshot taken after 9 minutes, temperature was going up and down between 67-70degC.

 

[propower]

First off - THANKS SO much for the great on point replies!

From comparing my 2 yes test with two similar machines now and one IvyBridge LAte 2012 I can see that my temps are indeed at least 10degC more for the same load. I made the right choice to order a replacement! It seems about 2 yes load is all the new i7 imacs will do before the fans start coming on. Not a complaint - just an observation.

What is TOTALLY fascinating to me now is that I originally tested a 3.4GHz i5 late 2013 imac and at 2 yeses it was another 10degC cooler. But the really interesting thing is that at 4 yes's (100% load for the i5) it only peaked at 86degC and still no fan! My simple logic said the i7 under equivalent load should be the same temp (maybe a little higher - 3.5/3.4 = 3% or 2 to 3degC at load. I suppose Woddys 4 Y test is close to this.... but the 2Y test is way off --- I was scared when I returned that one that it was a particularly well built one thermally speaking... LOL

 

[Nazmor]

Out of curiosity, do you have the fan boost option on in the app? It will alter fan behavior and push them to max much earlier.

The default behavior seems to only start ramping up at 90C, reaching max before 100C.

Fan control in the app is off. I may be wrong but it looks like fan will start ramping up when average temp is above 80degC or any core temp is above 95degC. There are posted 4yes tests with low fan rpm and max core temperature below 95. In my test when core temp went above 95 I got warning message from "temperature gauge pro" and at the same time fan started to ramping up to max. So quite possibly 1degC can make a difference between noise and silence and temperature in the room can affect it.

Another observation is that fan never stays at the same rpm unless it is max around 2600rpm so there is no chance it will stay at 1900rpm for a long time, in my tests it goes up to let's say to 1900rpm and then goes down to 1200rpm and it will repeat if temperature will go up again. So it would be good to actually get feedback about maximum fan rpm for 4xyes in the 10 minutes period.

Last thing, we are testing using very artificial procedure putting cpu all cores to 100% load, how often in real application it happens? To be honest before I started to test my system with "yes's" I heard fan ramping up only once for maybe 10 sec when I was testing game for few hours, and other than that fan is very quiet.

 

[propower]

Nazmor -

All good and relevant points.

- THe artificial YES test is just a convenient and easy access test for different people to use so at least the test is the same.

- 100% load probably only matters to hard core HAndbrake and very serious video encoders.

- Constant 25% to 50% load is very real for me (pro audio). Knowing how hard I can push a machine before the fans kick on is absolutely key (again - just for me).

- The fan cooling algorithm is what it is. They could have done this many ways but Apple has been using a similar algorithm for years. As an old PC builder I would never design a machine to run at 100degC. But it is within the Proc spec so there you have it.

- I still think there is variation in heatsink attach. Your (and mine) may be not as well assembled as others.

- If ones common use never puts the computer (or rarely) in these kinds of loads this can very easily be a NON-issue

 

[propower]

New i7 BTO ---

A little better than the first one I got. At 2 yes's the temps peak in the low 90's and the fan comes on for 15 sec or so (caught it in the picture). Better than last machine - nowhere near how well the i5 performed. I still can not claim that the load on the i7 at 2 yes's is the same as the i5 though... these temps are also 20degC higher than the late 2012 i7/680MX posted above !!!!

 

[Nismo73]

New i7 BTO ---

A little better than the first one I got. At 2 yes's the temps peak in the low 90's and the fan comes on for 15 sec or so (caught it in the picture). Better than last machine - nowhere near how well the i5 performed. I still can not claim that the load on the i7 at 2 yes's is the same as the i5 though... these temps are also 20degC higher than the late 2012 i7/680MX posted above !!!!

I'm trying to wrap my head around how much these tests are stressing the cores compared to real world scenarios such as BF3 or doing some video editing???

 

[Nirurin]

Here is my i7 780m 2013:

Seems to be stable at 67c/68c, with just 1200rpm fan speed.

This is about 5c hotter than last years i7 that was shown above, but then its also faster so I guess thats probably about right.

 

[propower]

I'm trying to wrap my head around how much these tests are stressing the cores compared to real world scenarios such as BF3 or doing some video editing???

Although not a perfect test (nothing replaces the real load) I have found these all totally consistent with my pro tools (pro audio) loading. For a given session density I get a certain fairly constant CPU load. On an i7 for instance two yes = 25% total cpu load. When I run a pro tools session at 25% it goes to the same place thermally.

----------

Here is my i7 780m 2013:

Seems to be stable at 67c/68c, with just 1200rpm fan speed.

This is about 5c hotter than last years i7 that was shown above, but then its also faster so I guess thats probably about right.

I am assuming this was a two yes test??

Let me ask also -- (not that it seems likely but....) the machine I have is 3.5GHz, 256SSD, i7, GTX775M. The other i7 I tested was 8Gram and 512SSD. Could you please detail the rest of your build (fusion drive??). Is it even remotely possible that having the spinning drive inside could be a thermal help somehow??? It is a large thermal mass - is it close or in contact with the processor heatsink?? THe i5 I tested that did so well was a simple 1TB spinning drive...

Regardless- your results are exactly what I was hoping for - and exactly what I have NOT gotten on now two BTO i7 machines. What the heck do I do now?

 

[woddy]

Did you ever consider the environmental factor? An iMac that "breathes" cold air may not need as much rpm as one breathing (slightly) hotter air?

 

[Nirurin]

Although not a perfect test (nothing replaces the real load) I have found these all totally consistent with my pro tools (pro audio) loading. For a given session density I get a certain fairly constant CPU load. On an i7 for instance two yes = 25% total cpu load. When I run a pro tools session at 25% it goes to the same place thermally.

----------



I am assuming this was a two yes test??

Let me ask also -- (not that it seems likely but....) the machine I have is 3.5GHz, 256SSD, i7, GTX775M. The other i7 I tested was 8Gram and 512SSD. Could you please detail the rest of your build (fusion drive??). Is it even remotely possible that having the spinning drive inside could be a thermal help somehow??? It is a large thermal mass - is it close or in contact with the processor heatsink?? THe i5 I tested that did so well was a simple 1TB spinning drive...

Regardless- your results are exactly what I was hoping for - and exactly what I have NOT gotten on now two BTO i7 machines. What the heck do I do now?

Yeh was a two yes test.

2013 i7, 780m, 256gb SSD