No 2018 iMac?!

[Fishrrman]

Well, looks like this thread has run it course...!

Once more, a prediction (a guess, and nothing more):
2019 iMacs -will not- have an external "re-design", except "on the inside" (which is where it counts, anyway).
I do expect to see "space grey" as an option. Or that may become the "new color"...

 

[EugW]

An interesting side story on all this:

I currently have an i5 3.8ghz 27 inch iMac. I used to have a 2017 3.4ghz 21 inch iMac.
I regularly encode videos of plays that I videotape. The plays typically run around 2 hours.
When I exported these long videos with my 3.4ghz i5, it would very gradually get up to 90c but the fans stayed quiet, and it never got hotter than that.
With the 3.8ghz i5 27 inch the fans start ramping up about 5 percent of the way into the export, and the temps run up to as high as 98c, and about 10 percent of the way in, the ‘estimated time remaining’ starts to creep up, indicating its starting to thermal throttle.
HOWEVER, when I disable turbo boost for these super long exports, the 3.8ghz never get hotter than 90c and the fans stay quiet. So it behaves pretty much exactly like the 3.4ghz did, tho the 3.4 still had turbo boost enabled.
The kicker is that with turbo boost dosabled the encodes actually take LESS time - about 7 minutes less.

I’m guessing the 3.8 i5 throttles because its 91 watts, vs the 65 watts of the 3.4 i5.
At any rate, I’m better off for my narrow use case to turn off turbo boost on the i5 for super long exports/encodes.
And ... since the i5 thermal throttles in that use case, I have no doubt the i7 would be as bad, if not worse, since its using not only using turbo boost but also hyperthreading to try to eek more performance out of the poor chip.

This did make me conclude that benchmarks like geekbench dont tell the full story. My geekbench numbers on the 3.8 with turbo boost enabled are perfectly fine (15,400 multi, 5320 single)
But geekbench doesnt test the 91 watt processors long enough to get them to where they thermal throttle.

In some simple tests some of us did here, the 3.8 GHz i5-7600K behaves somewhat like the 3.5 GHz i5-7600, but I suspect this varies depending upon the specific individual chip, since in some third party non-Mac tests the 7600K does run hotter than the 7600, at least in extreme situations. In any case, that's why I bought the 7600, since it is nominally a 65 Watt chip, as opposed to the nominally 91 Watt 7600K. However, what most tests seem to indicate is that the 65 Watt 3.4 GHz i5-7500 behaves the best out of all of those four chips in the 27" iMac.

 

[Internet Enzyme]

That's just it. They don't actually always work that way. It takes a long time for the i5 iMac fans to ramp up, and when they do ramp up, it's a more gradual process before it hits max. In contrast, the i7 iMac's fans ramp right up to maximum in less than 30 seconds.

What bugged me about my i7 7700K was that even if I were exporting a short birthday video with a 2 minute encode, I'd hear the i7 fan ramp up to max for over a minute. This never happens with my i5 7600.

BTW, we did some testing with a file in Handbrake, and the i7-7700K ramped up in 30 seconds and stayed at maximum until the 10 minute encode was finished. The i7-7600 took over 9 minutes to ramp up to significantly audible and finished the encode in 12.5 minutes. IOW, the 7700K was 20% faster, but was much, much louder during most of the encode.

Even my 2010 Core i7 870 didn't ramp the fan up anywhere near as fast as the 2017 i7-7700K. I bought the 2017 i7-7700K hoping it'd behave like the 2010 Core i7 870, but nope the 7700K was much worse for fan noise, so I returned the 7700K and got the i5-7600 instead. I'm much happier with the 7600.

I suppose the problem here is that the thermals of the imac require higher fan speeds to cool a better processor—fair enough. I just don’t see how temporary fan noise as you export a video is bad if the video exports faster due to it. ****, I’ve purposefully made my fan curve more aggressive on all of my macs. Forcing fan speeds up is almost always beneficial anyways
[doublepost=1546190635][/doublepost]

Well, looks like this thread has run it course...!

Once more, a prediction (a guess, and nothing more):
2019 iMacs -will not- have an external "re-design", except "on the inside" (which is where it counts, anyway).
I do expect to see "space grey" as an option. Or that may become the "new color"...

Yeah, I’m starting to think that there really wont be a 2018 iMac. Surprised how large this thread became, too

 

[EugW]

I suppose the problem here is that the thermals of the imac require higher fan speeds to cool a better processor—fair enough. I just don’t see how temporary fan noise as you export a video is bad if the video exports faster due to it. ****, I’ve purposefully made my fan curve more aggressive on all of my macs. Forcing fan speeds up is almost always beneficial anyways

If you don't mind fan noise, that's fine, but a lot of us do. We like good looking quiet computers. The iMac i7-7700K is good looking, but it sure isn't quiet.

Unfortunately, Apple uses the same heatsink/fan setup for all the 27" iMac CPUs AFAIK, which is a problem. This is an issue on the PC side as well, but some boutique PC assemblers use higher end cooling with the i7-7700K, and it makes a huge difference in fan noise.

BTW, it can also make a difference in performance too. Some of the higher performance 27" iMacs will actually throttle under heavy load even with the fan running at maximum, whereas the entry level 27" i5 never throttles. @BigBoy2018 gave the example where his 7600K actually sometimes performs faster with Turbo Boost turned off because when you turn on Turbo Boost the chip throttles, whereas with Turbo Boost turned off, the chip doesn't get hot enough to throttle. In fact, with Turbo Boost turned off, his iMac never even reaches maximum fan noise, but performs better than his iMac at maximum fan with Turbo Boost turned on.

 

[BigBoy2018]

I suppose the problem here is that the thermals of the imac require higher fan speeds to cool a better processor—fair enough. I just don’t see how temporary fan noise as you export a video is bad if the video exports faster due to it. ****, I’ve purposefully made my fan curve more aggressive on all of my macs. Forcing fan speeds up is almost always beneficial anyways

But heres the problem: In my experience with the 3.8ghz 7600k imac with long encodes (2+ hours) the processor got hot, the fans sped up AND it thermally throttled. My long exports actually went FASTER when I disabled turbo boost - which kept the processor cooler.
So if youre doing super long exports/encodes like me, then the 3.8ghz i5 with turbo boost enabled is actually a detriment. The i7, with turbo boost AND hyperthreading both stressing the chip, i’m sure it would be just as bad or worse.

So maybe the i7 is good for short tasks, but then again, isnt it the long tasks where you would hope it would save you some time? From my experience, its the exact opposite.

 

[Internet Enzyme]

If you don't mind fan noise, that's fine, but a lot of us do. We like good looking quiet computers. The iMac i7-7700K is good looking, but it sure isn't quiet.

Unfortunately, Apple uses the same heatsink/fan setup for all the 27" iMac CPUs AFAIK, which is a problem. This is an issue on the PC side as well, but some boutique PC assemblers use higher end cooling with the i7-7700K, and it makes a huge difference in fan noise.

BTW, it can also make a difference in performance too. Some of the higher performance 27" iMacs will actually throttle under heavy load even with the fan running at maximum, whereas the entry level 27" i5 never throttles. @BigBoy2018 gave the example where his 7600K actually sometimes performs faster with Turbo Boost turned off because when you turn on Turbo Boost the chip throttles, whereas with Turbo Boost turned off, the chip doesn't get hot enough to throttle. In fact, with Turbo Boost turned off, his iMac never even reaches maximum fan noise, but performs better than his iMac at maximum fan with Turbo Boost turned on.

But heres the problem: In my experience with the 3.8ghz 7600k imac with long encodes (2+ hours) the processor got hot, the fans sped up AND it thermally throttled. My long exports actually went FASTER when I disabled turbo boost - which kept the processor cooler.
So if youre doing super long exports/encodes like me, then the 3.8ghz i5 with turbo boost enabled is actually a detriment. The i7, with turbo boost AND hyperthreading both stressing the chip, i’m sure it would be just as bad or worse.

So maybe the i7 is good for short tasks, but then again, isnt it the long tasks where you would hope it would save you some time? From my experience, its the exact opposite.

I just ran cinebench, which literally maxed out my 7700k at 100% and it maintained a steady 4.4 GHz the whole time—no throttling. The fan did become audible during the test, reaching 2000 RPM. I can see it getting up to 2700 RPM with longer loads. So I suppose if you want a completely dead silent build at 100% cpu utilization, a super thin all-in-one desktop is not exactly your best bet in the first place. It might get to that fan speed a little bit quicker, which is I suppose the crux of your argument, but I feel like every single computer apple makes besides the macbook will kick up the fans and get very very audible—at least this one doesn't throttle, giving you some of the best real world performance out of any modern mac.

 

[BigBoy2018]

I just ran cinebench, which literally maxed out my 7700k at 100% and it maintained a steady 4.4 GHz the whole time—no throttling. The fan did become audible during the test, reaching 2000 RPM. I can see it getting up to 2700 RPM with longer loads. So I suppose if you want a completely dead silent build at 100% cpu utilization, a super thin all-in-one desktop is not exactly your best bet in the first place. It might get to that fan speed a little bit quicker, which is I suppose the crux of your argument, but I feel like every single computer apple makes besides the macbook will kick up the fans and get very very audible—at least this one doesn't throttle, giving you some of the best real world performance out of any modern mac.

You didnt clearly read what I wrote. I’m talking about LONG exports/encodes. Thats when it thermally throttles, about 10 minutes into the endcode. The cinebench test is no where near that long, so that, and other benchmarks are useless tests of what I’m talking about.

Its maybe 20 mins into the long render that you start to see it lose ground to a cooler running i5

 

[EugW]

I just ran cinebench, which literally maxed out my 7700k at 100% and it maintained a steady 4.4 GHz the whole time—no throttling. The fan did become audible during the test, reaching 2000 RPM. I can see it getting up to 2700 RPM with longer loads. So I suppose if you want a completely dead silent build at 100% cpu utilization, a super thin all-in-one desktop is not exactly your best bet in the first place. It might get to that fan speed a little bit quicker, which is I suppose the crux of your argument, but I feel like every single computer apple makes besides the macbook will kick up the fans and get very very audible—at least this one doesn't throttle, giving you some of the best real world performance out of any modern mac.

Cinebench is basically irrelevant to what we are talking about. It's way too short of a test. Yet, even in that very short test, the fan ramps up on the i7-7700K iMac in Cinebench. Not so on my i5.

And as mentioned, I've owned both the i7-7700K iMac and the i5-7600 iMac.

 

[BigBoy2018]

Cinebench is basically irrelevant to what we are talking about. It's way too short of a test. Yet, even in that very short test, the fan ramps up on the i7-7700K iMac in Cinebench. Not so on my i5.

And as mentioned, I've owned both the i7-7700K iMac and the i5-7600 iMac.

Thats the thing, the i7 may work quicker for short tasks, so you save a minute or so out of 8 min render
A time saving ls sure, but if you have a render that lasts 2 hours, that i7 is a hinderence, not a help, since eventually during that render it will throttle several times.
If the i7 never theottled and ran at full speed the whole way through, then it would save a LOT of time and to me, be worth the extra expense.
But it’s not.

 

[EugW]

Thats the thing, the i7 may work quicker for short tasks, so you save a minute or so out of 8 min render
A time saving ls sure, but if you have a render that lasts 2 hours, that i7 is a hinderence, not a help, since eventually during that render it will throttle several times.
If the i7 never theottled and ran at full speed the whole way through, then it would save a LOT of time and to me, be worth the extra expense.
But it’s not.

Well, I think it depends on the individual unit. Some people have reported i7 iMacs that don't throttle (much) with long encodes, but that are just very loud.

OTOH, some people (like you) report i5-7600K iMacs that throttle, while others report ones that don't throttle.

However, I don't think I've read of anyone with a 3.4 GHz i5 iMac that throttles under any conditions, including encodes that last hours.

 

[BigBoy2018]

OTOH, some people (like you) report i5-7600K iMacs that throttle, while others report ones that don't throttle.

Gotcha, so you’ve actually heard of people with the i5-7600k that didnt throttle, even with super long encodes/exports?

 

[EugW]

Gotcha, so you’ve actually heard of people with the i5-7600k that didnt throttle, even with super long encodes/exports?

Well, I dunno about "super long" but with encodes longer than 10 minutes.

 

[MandiMac]

Gotcha, so you’ve actually heard of people with the i5-7600k that didnt throttle, even with super long encodes/exports?

Please define what you mean with throttling:
a) not reaching the frequency the processor is rated for in the first place or
b) not reaching the full Turbo Boost frequency that is over the regular frequency

Because if you mean a) then the i5 won‘t throttle.

 

[EugW]

Please define what you mean with throttling:
a) not reaching the frequency the processor is rated for in the first place or
b) not reaching the full Turbo Boost frequency that is over the regular frequency

Because if you mean a) then the i5 won‘t throttle.

AFAIK, at least the 3.4 GHz i5-7500 iMac will run at full Turbo Boost for extended periods.

Full quad-core Turbo for the i5-7500 is 3.6 GHz. Non-Turbo is 3.4 GHz.
Full quad-core Turbo for the i5-7600 is 3.9 GHz. Non-Turbo is 3.5 GHz.
Full quad-core Turbo for the i5-7600K is 4.0 GHz. Non-Turbo is 3.8 GHz.
Full quad-core Turbo for the i7-7700K is 4.4 GHz, but with HyperThreading. Non-Turbo is 4.2 GHz.

Given these numbers (and the fact the non-K chips are nominally classified as 65 Watt and the K chips are nominally classified as 91 Watt), it's no surprise that the i5-7500 can (in some tests) run at full Turbo Boost essentially forever. Its Turbo Boost is only 0.1 GHz higher than the base clock of the i5-7600.

Also, given these numbers, it's not a surprise the i5-7600 performs very similarly to the i5-7600K, even in short benchmarks.

The main advantage of the i5-7600K iMac over the i5-7600 iMac is the faster GPU, but that faster GPU is also more power hungry when in heavy use.

In terms of @BigBoy2018's i5-7600K, it would be interesting to compare that in non-Turbo mode (at 3.8 GHz) vs the i5-7600 with full Turbo (at 3.9 GHz) for extended encodes.

 

[Internet Enzyme]

Cinebench is basically irrelevant to what we are talking about. It's way too short of a test. Yet, even in that very short test, the fan ramps up on the i7-7700K iMac in Cinebench. Not so on my i5.

And as mentioned, I've owned both the i7-7700K iMac and the i5-7600 iMac.

At 2700 RPM, the cpu never goes past like 98ºC, just barely missing the 100ºC T-junction. It’s kind of a super impractical thing to test out if the cpu for some reason throttles a few times over an hour of extended encoding, but I suppose a test like that could be arranged. Also, kind of curious, what did your computer score in cinebench? I suppose it doesnt matter if the i5 really does fare better in lengthy video encodes and real world scenarios then the i7 does, as you purport, but I’m just curious. I suppose we just have a different tolerance to fan noise. Obviously there is a happy middle ground, but I suppose I just think its reasonable to tolerate loud fans as I export a video if I’m getting substantially improved performance.

 

[MandiMac]

Then again, many forum users here are so hellbent on an i7 they‘ll never gonna use to their full extent - the processor upgrade is in that case just wasted money IMO.

 

[BigBoy2018]

Then again, many forum users here are so hellbent on an i7 they‘ll never gonna use to their full extent - the processor upgrade is in that case just wasted money IMO.

My things is I’m encoding videos that run 1 to 2 hours. The turbo mode of the 7600k doesnt offer any advantage with a render that long, so I think its a fair guess the i7 wouldnt either.

So, in my use case, I WOULD be using he i7 to its ‘full extent’ IF it didnt get so hot and throttle under loads like that.
So, as far as im concerned the i7 would be borderline useless for any use case.

it may save you 6 seconds on a 2-minute render, but something that minor you’d barely notice. Is that worth the extra expense and the extra heat stress the i7 puts on the entire system?

 

[wardie]

My things is I’m encoding videos that run 1 to 2 hours. The turbo mode of the 7600k doesnt offer any advantage with a render that long, so I think its a fair guess the i7 wouldnt either.

So, in my use case, I WOULD be using he i7 to its ‘full extent’ IF it didnt get so hot and throttle under loads like that.
So, as far as im concerned the i7 would be borderline useless for any use case.

it may save you 6 seconds on a 2-minute render, but something that minor you’d barely notice. Is that worth the extra expense and the extra heat stress the i7 puts on the entire system?

Have you got some comparative benchmarks to back this up around performance and stress impact? There is some info here. The multi-core performance of the i7 tend to be about 27% better than the i5, I expect due to a combination of the hyper-threading plus clock speed. Depends on your use case, but for long CPU bound jobs this is going to make a difference to some people e.g. 1/2 hour productivity on a 2h batch job.

https://www.geekbench.com/blog/2017/06/imac-27-inch-performance-june-2017/

On the noise thing yes the i7 ramps up the fans and it noise when at full tilt with turbo-boost on, but my experience is that is can sustain the turbo-boost speeds albeit noisily. However by disabling turbo boost (freeware) I’ve found the fan noise almost completely disappears (down to < 2000 rpm) at the expense of only c. 5% performance, as the non-turbo clock speed is high anyway and it runs over 4GHz continuously and slightly cooler at around 90C not 95-98C.

On high temps... seems to be a variety of views on this. IMHO as long as it keeps itself within the temp range it is designed to operate it at then fine I don’t care (I’m not into custom overclocking etc). I’ve replaced many computer components in my machines over decades but never a cooked CPU.

 

[BigBoy2018]

Then again, many forum users here are so hellbent on an i7 they‘ll never gonna use to their full extent - the processor upgrade is in that case just wasted money IMO.

Have you got some comparative benchmarks to back this up around performance and stress impact? There is some info here. The multi-core performance of the i7 tend to be about 27% better than the i5, I expect due to a combination of the hyper-threading plus clock speed. Depends on your use case, but for long CPU bound jobs this is going to make a difference to some people e.g. 1/2 hour productivity on a 2h batch job.

https://www.geekbench.com/blog/2017/06/imac-27-inch-performance-june-2017/

On the noise thing yes the i7 ramps up the fans and it noise when at full tilt with turbo-boost on, but my experience is that is can sustain the turbo-boost speeds albeit noisily. However by disabling turbo boost (freeware) I’ve found the fan noise almost completely disappears (down to < 2000 rpm) at the expense of only c. 5% performance, as the non-turbo clock speed is high anyway and it runs over 4GHz continuously and slightly cooler at around 90C not 95-98C.

On high temps... seems to be a variety of views on this. IMHO as long as it keeps itself within the temp range it is designed to operate it at then fine I don’t care (I’m not into custom overclocking etc). I’ve replaced many computer components in my machines over decades but never a cooked CPU.

As we’ve already discussed here, Geekbench is no where near. a long enough test to measure he effect of prolonged heat stress on performance.
My example is that for a 2 hour video export, the total time taken was actually about 7 mins shorter with turbo boost disabled.
With turbo boost enabled, the cpu runs at full speed for a while, (but hot, 98c or so - with fans running higher) but then about the 10 minute mark the ‘estimated time remaining’ starts to creep up.
That does not happen with turbo boost disabled. It runs around 88-90c throughout and the fans never ramp up and the 2 hour video render actually takes less time.

So again, geekbench is a worthless test of what I’m talking about here.

 

[wardie]

As we’ve already discussed here, Geekbench is no where near. a long enough test to measure he effect of prolonged heat stress on performance.
My example is that for a 2 hour video export, the total time taken was actually about 7 mins shorter with turbo boost disabled.
With turbo boost enabled, the cpu runs at full speed for a while, (but hot, 98c or so - with fans running higher) but then about the 10 minute mark the ‘estimated time remaining’ starts to creep up.
That does not happen with turbo boost disabled. It runs around 88-90c throughout and the fans never ramp up and the 2 hour video render actually takes less time.

So again, geekbench is a worthless test of what I’m talking about here.

The comparison I was looking for was i5 vs i7 over that time span... granted turbo on/off on i7 is not much % difference - hence me preferring to switch it off for a lot less noise and slightly lower temps... I’ve not got an i5 to compare to on a long job hitting multi-core performance... do you have some comparative data? The statements I find hard to understand above / trying to clarify is why it is said that the i7 has no practical benefit over i5 in any use case?

 

[BigBoy2018]

The comparison I was looking for was i5 vs i7 over that time span... granted turbo on/off on i7 is not much % difference - hence me preferring to switch it off for a lot less noise and slightly lower temps... I’ve not got an i5 to compare to on a long job hitting multi-core performance... do you have some comparative data?

I dont have that test, but the i7 hyperthreading in my mind is much like turbo boost. Its little tricks to get the processor to perform better ... for SHORT. periods of time.
My supposition is that if turbo boost eventually stresses the chip into thermal throttling during a really long render, then adding hyperhreading on top of that would make the problem as bad or worse.

At the end of the day, for my use case, I think I’d rather have a cpu that is rated for a high clock speed and can perform well for extended periods without these ‘tricks’ to eek out better scores in benchmarks, but are a hinderence with super long tasks.

 

[wardie]

I dont have that test, but the i7 hyperthreading in my mind is much like turbo boost. Its little tricks to get the processor to perform better ... for SHORT. periods of time.
My supposition is that if turbo boost eventually stresses the chip into thermal throttling during a really long render, then adding hyperhreading on top of that would make the problem as bad or worse.

At the end of the day, for my use case, I think I’d rather have a cpu that is rated for a high clock speed and can perform well for extended periods without these ‘tricks’ to eek out better scores in benchmarks, but are a hinderence with super long tasks.

Not my experience, I’ve watched long job runs for say 0.5-2 hours doing Lightroom batches and the processors are 100% on all 8 logical cores all the time at >4GHz (speed depending if TB is on/off). HT doesn’t change the clock speed and is not something that will be turned off my thermal throttling, I would expect to see the whole chip clock speed reduce in that case, which it does not.

HT is just a clever mechanism to run a couple of threads on a single core concurrently when they are using different parts of the core and thus increases net transaction throughput. Typically this seems to have a “real world benefit of 20% (certainly not double as some people would assume). Jury is out on whether Intel with bother with it in future though, interesting article below.

Either way, it does have a sustainable performance gain but unless someone can post some controlled conditions i5 vs i7 iMac benchmarks over a long time period CPU multi-core tests we’ll not really know - I certainly take your point about Geekbench and can’t find much else posted out there specifically on this.

https://www.intel.com/content/www/u...per-threading/hyper-threading-technology.html

https://www.extremetech.com/extreme...-intel-will-drop-hyper-threading-from-core-i7

 

[SecuritySteve]

Not my experience, I’ve watched long job runs for say 0.5-2 hours doing Lightroom batches and the processors are 100% on all 8 logical cores all the time at >4GHz (speed depending if TB is on/off). HT doesn’t change the clock speed and is not something that will be turned off my thermal throttling, I would expect to see the whole chip clock speed reduce in that case, which it does not.

HT is just a clever mechanism to run a couple of threads on a single core concurrently when they are using different parts of the core and thus increases net transaction throughput. Typically this seems to have a “real world benefit of 20% (certainly not double as some people would assume). Jury is out on whether Intel with bother with it in future though, interesting article below.

Either way, it does have a sustainable performance gain but unless someone can post some controlled conditions i5 vs i7 iMac benchmarks over a long time period CPU multi-core tests we’ll not really know - I certainly take your point about Geekbench and can’t find much else posted out there specifically on this.

https://www.intel.com/content/www/u...per-threading/hyper-threading-technology.html

https://www.extremetech.com/extreme...-intel-will-drop-hyper-threading-from-core-i7

Don't be fooled. Hyperthreading is still valuable, and it's not going away. Intel is just changing their i7s to be what i5s used to be, and i9s to be what i7s used to be. It's just marketing that you're getting something bigger and better, when the reality is it's still just the next generation of the top-class.

 

[BigBoy2018]

Don't be fooled. Hyperthreading is still valuable, and it's not going away. Intel is just changing their i7s to be what i5s used to be, and i9s to be what i7s used to be. It's just marketing that you're getting something bigger and better, when the reality is it's still just the next generation of the top-class.

I still think HT comes with the price of extra fan noise (at least in current iMacs) and of course extra heat stress. Not sure if that makes any difference to you that swear by the i7, but people like me and EugW prefer a cooler processor that runs without HT.

 

[SecuritySteve]

I still think HT comes with the price of extra fan noise (at least in current iMacs) and of course extra heat stress. Not sure if that makes any difference to you that swear by the i7, but people like me and EugW prefer a cooler processor that runs without HT.

Oh of course. That's why intel makes i5s (or the new i7s). Different folks have different needs for their CPUs.