Any way to let the 9900K be 6 cores with Hyper Threading? Just curious what that freq would be.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
2019 iMac - i5 vs i9? Cooling and Throttling?
- Thread starter Black Diesel
- Start date
- Sort by reaction score
Any way to let the 9900K be 6 cores with Hyper Threading? Just curious what that freq would be.
No clue in actually trying to get the 9900K in acting like a 6 core with hyperthreading.
The only thing that you could do is limit how many threads the program can use if there is such an option like there is in Cinebench.
Do you even fully read what you post or anyone else's?
Look at the chart you used. A stock 9900K is capable of 5GHz on one/two cores. If Apple claims up to 5GHz than it can only mean one or two cores.
I didn't say Apple claimed its 95W restricted though. I've said they give capped 85W on full load, otherwise the 9900K will receive 120W spikes as seen in Max Tech's i9 iMac video.
Apple will be even less bothered to change iMac chassis if future 10nm chips will produce less heat. The 9900K was a knee jerk reaction to Ryzen anyway. Doubt Apple had a 14nm 8c/16t chip in mind when designing the current chassis.
Xcode lets you limit the # of cores and turn HT On/Off globally on a Mac - though been a while since I used it so have the check the HT part
Yes - it still does...
But no worries - I already see a 6C 12T test that gives the info I wanted
Last edited:
All that Apple advertise is that the Core i9 iMac “now clocks Turbo Boost speeds up to 5.0GHz,” nothing more. Apple do not explain what Turbo Boost is. Perhaps it would be better if they did, but as they do not, it’s the customer’s responsibility to look it up.
One could, for example, perform a Google search for Turbo Boost, where the very first result is this page on Intel’s website describing Intel® Turbo Boost Technology 2.0:
Due to varying power characteristics, some parts with Intel® Turbo Boost Technology 2.0 may not achieve maximum turbo frequencies when running heavy workloads and using multiple cores concurrently [emphasis added].
Availability and frequency upside of Intel® Turbo Boost Technology 2.0 state depends upon a number of factors including, but not limited to, the following:
Availability and frequency upside of Intel® Turbo Boost Technology 2.0 state depends upon a number of factors including, but not limited to, the following:
…
> Number of active cores [emphasis added]
> Number of active cores [emphasis added]
Or one could read the Turbo Boost article on Wikipedia, which, citing the web page linked above, says:
The increased clock rate is limited by the processor's power, current, and thermal limits, the number of cores currently in use, and the maximum frequency of the active cores [emphasis added].
Or one could go to Intel’s website and, with a just a little digging, look at their web page for Intel® Turbo Boost Max Technology 3.0, where it prominently says:
Intel® Turbo Boost Max Technology 3.0 is a combination of software and hardware that delivers more than 15% better single-threaded performance [emphasis added].
Or one could (as mcnallym pointed out above) find Intel’s technical specifications for the Core i9-9900K and click on the information button (the “i” in a blue circle) next to “Max Turbo Frequency”:
Max turbo frequency is the maximum single core frequency at which the processor is capable of operating using Intel® Turbo Boost Technology [emphasis added]…
The meaning of Turbo Boost is easily discoverable. Apple are not responsible for a customer’s mistaken assumptions.
For those still interested in this debate, here is an article that helps to clear up some confusion:
The Intel Core i9-9900K at 95W: Fixing The Power for SFF
It also helps to show that both sides of the argument have some validity.
This is especially highlighted in their conclusion of a 95W restricted 9900K:
Overall, it acts like a 9900K [i9] in single thread mode, and like a 9700K [i7] in multi-thread mode.
So, really, the new iMac i9—thermally restricted to ~85 watts—is really no better, and in some cases worse, then the 9th gen i7 chip. Though, to be fair, we need to wait until more directly comparable reviews are published about actual performace of the actual iMac.
Regarding thermals, bear in mind the results we've seen thus far are on a brand new machine. What happens when you start getting dust on the heatsink, for example? Thermal performance will never improve, only worsen throughout its lifespan. The i5 is also turbo boosting nearly at Intel's max spec, whereas the "i9" is far from. It's not unreasonable to see this gap widen.
There has always been some amount of variance between the implementations of a given processor, but, I think Apple just set the standard for the largest deviation with the "i9" iMac. This is my contention.
Why does Puget System's i9 achieve a roughly 15–20% better Cinebench R15 score than the iMac. R15 also doesn't load the processor as realistically as R20, either. So, as people have pointed out, it sounds like most, if not all business class machines are running the 9900k in an already restricted mode. So, if this can be confirmed, what's with the huge discrepancy in benchmarks of an already throttled 9900k vs the iMac?
Is this not due to thermal throttling?
The Intel Core i9-9900K at 95W: Fixing The Power for SFF
It also helps to show that both sides of the argument have some validity.
This is especially highlighted in their conclusion of a 95W restricted 9900K:
Overall, it acts like a 9900K [i9] in single thread mode, and like a 9700K [i7] in multi-thread mode.
Regarding thermals, bear in mind the results we've seen thus far are on a brand new machine. What happens when you start getting dust on the heatsink, for example? Thermal performance will never improve, only worsen throughout its lifespan. The i5 is also turbo boosting nearly at Intel's max spec, whereas the "i9" is far from. It's not unreasonable to see this gap widen.
There has always been some amount of variance between the implementations of a given processor, but, I think Apple just set the standard for the largest deviation with the "i9" iMac. This is my contention.
Why does Puget System's i9 achieve a roughly 15–20% better Cinebench R15 score than the iMac. R15 also doesn't load the processor as realistically as R20, either. So, as people have pointed out, it sounds like most, if not all business class machines are running the 9900k in an already restricted mode. So, if this can be confirmed, what's with the huge discrepancy in benchmarks of an already throttled 9900k vs the iMac?
Is this not due to thermal throttling?
A good example of thermal throttling was the 2018 MacBook Pro that dropped the processor speed BELOW its base clock speed. This was first from an update though.
Processors running above the base/minimum clock speed is NOT thermal throttling.
I really do not understand this logic that "not achieving the max turbo boost frequency" means it is thermal throttling. That was never the intention of turbo boost. The fact the Intel's spec sheet lists the minimum/base clock speed to be 3.6Ghz and these iMacs so far are getting a minimum of 3.8Ghz is NOT thermal throttling.
You guys are making the term "thermal throttling" useless at this point. As every single desktop from Dell, HP, Lenovo, Apple, ... is thermal throttling with this logic of never hitting sustained max turbo boost.
No is not due to thermal throttling.
Is the same argument I have with my dad. Here in the uk then when calculating pension increases then they changed from rpi to cpi. Retail Price Index vs Consumer Price Index, with cpi being lower rate.
The argument is that whilst dad says he is losing money, I say he simply isn’t getting as much as could have. Ie the increase is smaller then would have been however didn’t have the larger increase to begin with, you just didn’t get as much as would have done if using rpi still.
Same as the reports stating savers are losing money by not constantly moving money around. Again not losing money, simply not going as much as could do if moving money around.
The 9900k in the iMac isn’t slowing down due to overheating, it is never boosting to the point that starts to slow down due to heat.
In effect the cpu is not being slowed in the iMac it never boosts above the point that would generate the heat to start having to slow down.
Thermal throttling with the definitions that I have seen out there is where a cpu overheats and so is slowed to generate less heat so the cpu can cool down. That isn’t what seeing being reported in reviews.
What seeing in the reviews that bother to look at heat/noise is that the cpu temp is kept below where starts to overheat and slow. Ie it doesn’t boost as much as could potentially. It however isn’t slowing the cpu down due to it overheating.
You aren’t seeing the cpu slow down, you simply don’t see it boost to the point that it then needs to slow down as kept within the thermal envelope of what the cooling can deal with.
Is being constrained in that not being allowed to get to the point where will generate too much heat and start to throttle. It isn’t getting to the point where starts to throttle the cpu speed downwards as simply isn’t being allowed to get to the point where
When you look at the puget systems with a 9900k then they have a LOT more cooling in that case which is also thicker then an AIO like the iMac.
Consequently the cpu is able to run higher clock speeds as the heat is removed from the system quicker. You provide more cooling then have more headroom to boost. The iMac simply not having the headroom to boost as much as a system with more cooling. However that doesn’t appear to be the definition of thermal throttling.
You are’t being slowed down when overheats you simply don’t get fast enough to hit thermal throttling in the first place. I dare say that if did an extended 48 hour render that may change however if doing 48 hour renders would argue that bought the wrong system for your need if bought an iMac.
Not saying that if the iMac was thicker with more cooling that couldn’t get higher boost speeds which is effectively what getting when buy other machine styles compared with an AIO.
Thermal throttling though is where a machine actually slows down, not that it never reaches the higher speed.
Is the compromise of an AIO style machine. If was something like the much requested xMac ie consumer tower with pci-express slots then would have more cooling and more room for boosting the speed higher then the iMac does.
Tired so apologies if rambled on a bit here.
The thermal throttling accusations are soooo tiresome. With the exception of MBP i9, which I believe was fixed by an software update, it is a way to through dirt using non-relevant systems as comparisons.
It is more intersting to know how much the iMac/i9 can be pressed until the fan noice is intorable.
It is more intersting to know how much the iMac/i9 can be pressed until the fan noice is intorable.
Those who continue to refute the massive throttling of the i9 are delusional and ignorant of fact and physics.
What I find most interesting, is that the general consensus seems to be that the turbo function will only be a short bust, and that it is merely a bonus. However, that same consensus neglects the fact that the i5—the very basis this thread was initiated to compare with—has been shown to sustain ~500MHz above the base frequency. Most notable, is that it sustains ~100MHz less than the all-core-active-turbo-frequency rating Intel spec'd it at.
The 2018 MBP throttling was a case of the customer getting screwed beyond a promised minimum level. This also served to interject a misnomer on what exactly thermal throttling is. This was indeed (extreme) thermal throttling. However, processor performance / frequency is still governed by heat (thermals) above the base frequency, and is still very much subject to thermal throttling within its turbo range, as is demonstrated with the i9 iMac. What concerns me, is that Apple is barely getting by the "accepted" definition of thermal throttling, and not enough concern is present. I am shocked there are not more concerns about long-term implications (though I have read a few are). I've gone in and cleaned my 2014 MBP heatsink a few times, and recently replaced the thermal paste. These all made substantial differences. Correct me if I am wrong, but, is this as easy on iMac?
Regarding turbo boost, the customer is promised an "up to" turbo boost frequency. With the i9, that is up to 5.0GHz. Although not explicitly provided to the customer, with a little hunting, you will find that translates into a 4.7GHz all-core boost, as per Intel. This condition will be met, and sustained if the thermal conditions allow. If not, it will throttle to a level that is suitable for conditions. Though, as AnandTech found, this is more like 3.6 GHz when power (thermally) restricted to 95 watts. In the case of iMac, it seems it is thermally restricted to ~85 watts. Nevertheless, it also seems to be the consensus that other "business class" machines are not allowed to run "unrestricted," yet, they STILL all run much faster. Why? Because of thermals!
What it seems to come down to is: "Hey, how much are you turbo boosting?" Turbo boost is largely dependent on the implementation, and cooling solution. Turbo boost can, and does sustain at certain levels, provided there are no thermal constraints. This is well known, and shown with both iMac 27" processor options. I will add also, what about running in high ambient temperatures!?
In summary, the i9 iMac is up to 20% slower than other 9900K i9 equipped machines. Does this mean it's bad? No, not necessarily. The display is fantastic, you aren't plagued with a potentially problematic T2 chip, the fit and finish is leagues above the competition, it will calculate like no iMac before it, etc... Though restrained, the i9 is still a screamer and should always, but maybe not necessarily outpace the i5. It seems Apple is pushing the thermal envelop more than I think they ever have. This is not a good trend to applaud. The similar trend is what brought the not well-received butterfly keyboard.
As more benchmarks trickle in, we'll see how much more or less changes. It'll also be interesting to see a teardown, to see if the thermals have improved over the previous generation. What hasn't been discussed enough, are any long-term ramifications, i.e., extended processor loading, decreased thermal capabilities with age and use, etc. While the i9 may not throttle as per the majority definition today, I wouldn't be surprised if it does down the road, especially as it's now only a few hundred megahertz away. What will the consensus be then? Buy a new one, or complain?
What I find most interesting, is that the general consensus seems to be that the turbo function will only be a short bust, and that it is merely a bonus. However, that same consensus neglects the fact that the i5—the very basis this thread was initiated to compare with—has been shown to sustain ~500MHz above the base frequency. Most notable, is that it sustains ~100MHz less than the all-core-active-turbo-frequency rating Intel spec'd it at.
The 2018 MBP throttling was a case of the customer getting screwed beyond a promised minimum level. This also served to interject a misnomer on what exactly thermal throttling is. This was indeed (extreme) thermal throttling. However, processor performance / frequency is still governed by heat (thermals) above the base frequency, and is still very much subject to thermal throttling within its turbo range, as is demonstrated with the i9 iMac. What concerns me, is that Apple is barely getting by the "accepted" definition of thermal throttling, and not enough concern is present. I am shocked there are not more concerns about long-term implications (though I have read a few are). I've gone in and cleaned my 2014 MBP heatsink a few times, and recently replaced the thermal paste. These all made substantial differences. Correct me if I am wrong, but, is this as easy on iMac?
Regarding turbo boost, the customer is promised an "up to" turbo boost frequency. With the i9, that is up to 5.0GHz. Although not explicitly provided to the customer, with a little hunting, you will find that translates into a 4.7GHz all-core boost, as per Intel. This condition will be met, and sustained if the thermal conditions allow. If not, it will throttle to a level that is suitable for conditions. Though, as AnandTech found, this is more like 3.6 GHz when power (thermally) restricted to 95 watts. In the case of iMac, it seems it is thermally restricted to ~85 watts. Nevertheless, it also seems to be the consensus that other "business class" machines are not allowed to run "unrestricted," yet, they STILL all run much faster. Why? Because of thermals!
What it seems to come down to is: "Hey, how much are you turbo boosting?" Turbo boost is largely dependent on the implementation, and cooling solution. Turbo boost can, and does sustain at certain levels, provided there are no thermal constraints. This is well known, and shown with both iMac 27" processor options. I will add also, what about running in high ambient temperatures!?
In summary, the i9 iMac is up to 20% slower than other 9900K i9 equipped machines. Does this mean it's bad? No, not necessarily. The display is fantastic, you aren't plagued with a potentially problematic T2 chip, the fit and finish is leagues above the competition, it will calculate like no iMac before it, etc... Though restrained, the i9 is still a screamer and should always, but maybe not necessarily outpace the i5. It seems Apple is pushing the thermal envelop more than I think they ever have. This is not a good trend to applaud. The similar trend is what brought the not well-received butterfly keyboard.
As more benchmarks trickle in, we'll see how much more or less changes. It'll also be interesting to see a teardown, to see if the thermals have improved over the previous generation. What hasn't been discussed enough, are any long-term ramifications, i.e., extended processor loading, decreased thermal capabilities with age and use, etc. While the i9 may not throttle as per the majority definition today, I wouldn't be surprised if it does down the road, especially as it's now only a few hundred megahertz away. What will the consensus be then? Buy a new one, or complain?
Last edited:
Isn’t 3.8 GHz (what most extended tests show) greater than 3.6 GHz (base clock speed)?
Yes.
[i9] 3.8 falls between 3.6–4.7 (~18% of max potential) — Unacceptable processor implementation.
[i5] 4.2 also falls between 3.7–4.3 (~83% of max potential) — Acceptable processor implementation.
I really don't care if this behaviour fits the definition of "thermal throttling" or not. All I care about is how fast my iMac will be with typical tasks I throw at it - photo processing, some encoding, some instances of software synths etc. And how loud it will be while doing it, and how long I'll be able to use that machine.
So, if I decide to get the i9 over the i5 iMac, it will be based solely on what can be gained performance-wise, not what could be gained putting the i9 into a water-cooled, large PC enclosure. And so far, it looks like there's a lot of gain by choosing the i9 for processor-hungry tasks. (Although me personally, I would've liked to have the option of the i7-8700 in the 27" iMac (fast CPU and low TDP), as I've written elsewhere.)
Now this is an interesting question. Do you have any source where it is shown that CPUs "age" under sustained load? I was under the impression that while you can thermally destroy a CPU, there should be no damage to the circuits just by running it at top clock speeds.
This thread feels like such a waste of space now. My Core i9 + Vega 2019 iMac Pro is processing RAW images off my 5D and transcoding / analyzing 4K HEVC footage off my GoPro (about 2 hours of it) and the fans are significantly quieter than before. I'd hear nothing but a jet engine on my late-2015 4Ghz / M395X iMac. This is quieter and way faster.
VERY happy so far.
VERY happy so far.
This is why the term thermal throttling is useless now. So ANY processor without water cooling is thermal throttling?
Agreed! A 6C/12T would have been much more suitable than the same core / thread count as their entry Pro model.
To add to this, it is most frustrating that Apple is six months late to the 9900K party. They waited, they saw it was largely well-received and a beast of a chip, albeit power hungry. They could have set themselves ahead of the curve, and said, "hey, your huge tower performs this well, well, so does our AIO." This would have been HUGE for Apple, and, they could have done it with the iMac Pro thermal solution. But, no, they took the easy way out. They just slapped the chip in, said, "hey, look, we have the i9, too. Click "Buy"" but, what we will not tell you is that it behaves more like the latest i7.
I wish I could cite concrete information, but, perhaps others can chime in. It's always been my understanding that the hotter a processor runs, the more current will "leak." This leakage causes the need to feed it yet more current to achieve a given frequency. This just adds more heat, and reduces frequency, and accelerates degradation. Additionally, excessive throttling causes the core voltage to hunt (and overshoot) unnecessarily, causing further degradation via electromigration. Also, over time, processors will inherently need slightly more voltage (heat). Quantifying these theories I cannot.
[doublepost=1554325254][/doublepost]
Thermal throttling is not a binary term. Not all systems will throttle, including air. Some systems will throttle more, some will throttle less. As noted by many, there are a lot of factors that contribute to throttling.
Bboble, you need to show us a direct iMac competitor that has a 9900K doing all-cores @ 4.7GHz.
You've been all fluff with no substance.
You've been all fluff with no substance.
Seems it would be easier to show me the ones that do not. I don't know what others consider as a direct competitor. AIO, business class, etc...?
I see Puget Systems has been referenced in this thread. Their Cinebench R15 9900K score is much inline with my built 9900k system (and many others online) on an air cooler running all-core 4.7GHz. This same Cinebench score is 15–20% higher (as seen in testing variability of iMac) than the iMac 9900K.
I can hear the fan but it's not very loud..maybe 25% as loud as my 2015 iMac doing the same task:
You've got to be joking...any AIO with a 9900K inside would be a direct competitor.
Puget System benchmark is irrelevant, data was not compiled from testing in AIO's. Stop deflecting to it.
Once again, show us a direct iMac competitor that has a 9900K doing all-cores @ 4.7GHz.
Don't give us the runaround until you do.
Bboble, show us where either Apple or Intel ever promised an all-core boost at 4.7 GHz. You’ve been banging on and on about this, but the only evidence we’ve seen (as far as I can recall) is this chart at AnandTech showing their test results. But read the paragraph immediately above that chart:
The performance that Intel guarantees is the one on the box: the base frequency at the sustained TDP.… when we set our power consumption limits for the Core i9-9900K, this is what we saw at full load: 95W gave 3.6 GHz at 7-8 core load.
The Core i9-9900KF in the 2019 iMac is outperforming AnandTech’s test system: a sustained 3.9 GHz on all cores in adamjackson’s testing above.
Moreover, AnandTech’s chart shows that when restricted to 95W, their test system achieved 4.7 GHz only when loading six of the eight cores. Quagmire already posted the results of his Cinebench test showing the 2019 iMac with Core i9 running at a sustained 4.5 GHz on six cores, not at 3.6 GHz. (True, 4.5 GHz is not 4.7 GHz, but again, when did either Apple or Intel ever promise 4.7 GHz?)
(Furthermore, in the 2019 iMac, the i9 on six cores is Turbo Boosting higher than the i5 on six cores: 4.5 GHz vs 4.2 GHz. So again, I don’t understand your complaint about the i9 vs the i5.)
Links, please. Show us evidence of other all-in-one systems equipped with the Core i9-9900K outperforming the i9-equipped 2019 iMacs.
Last edited:
Bottom line is the i9 is damn fast in the new iMac and is even faster if housed with better cooling. What anyone calls that difference doesn't really matter. A rose is a rose by any other name. I've certainly not seen anything suggesting that Apple has misrepresented the product. I'm trying to decide if it's worth it FOR ME to purchase the i9 rather than the i5. But I have no concerns about the i9 not meeting my needs because it might be a little faster if cooling was better.
AIO is a direct competitor. That puget is not an AIO. The echo which is the smallest is based on the intel/amd combo 8809 chip, no option for an i9.
With the next smallest echo Pro you can also see the size of the cooler fitted to the chip, and the number of fans to cool that system. AIO just don’t offer anything like that amount of cooling system.
Puget is also a custom builder known for building high performance systems and put the cooling in to do the job. They are good systems but not exactly mainstream.
Lenovo AIO doesn’t offer an i9 option.
Dell AIO doesn’t offer an i9 option.
Can’t find anyone using an AIO that offers the level of cooling that the iMac seems to be doing.
I am not seeing anything yet anywhere that shows the i9 is overheating and reducing it’s clock speed. Simply that it isn’t give the headroom to start drawing more power so that overheats and yet still manages to meet the published specs.
What about people who just need the power of an i9 and macOS, while not necessarily wants an AIO? Apple isn't offering that option, and in that context, this i9 iMac is limiting the potential that you can get out of the chip. This i9 iMac is a great *Mac*, especially among the lineup in recent years. But if it requires further leashing just so it can fit inside a power & thermals limited chassis designed years ago, then we are not getting the best we can get.
Okay so lets stick with AIOs. And in the case of the iMac, there is literally a better thermally designed iMac Pro right there. Why is that the i9 could reach as high as 140w consumption in order to be unleashed, but Apple felt it is ok to not utilize the iMac Pro cooling which incidentally is rated at 135w?
Okay so lets stick with AIOs. And in the case of the iMac, there is literally a better thermally designed iMac Pro right there. Why is that the i9 could reach as high as 140w consumption in order to be unleashed, but Apple felt it is ok to not utilize the iMac Pro cooling which incidentally is rated at 135w?
Look how small the iMac forum is compared to iPhone. Or the Macintosh in general. Apple has 3 iPhones if you don't count last year's models at a discount. They have 4 iPads and yet they have 7 different Macs. The number crunchers who run Apple now would say that they have a very diverse product line from $999 - $15000 USD suiting all kinds of needs. the tower you want is the Mac mini and no it doesn't have the i9 but the iMac Pro & MacPro will both be Xeon so if you want an i9, the iMac is your option or a MacBook Pro but I can tell you that I have the 2.9Ghz i9 MacBook Pro and now the i9 iMac and the MacBook Pro feels slow as dog-poo compared to the iMac. If I knew that two $4,000 computers released within once cycle of each other and the desktop would be this much faster like 2X faster to me, I would have done a mid-range MacBook Pro instead. the iMac is no slouch and if you don't want the 'free' 5K monitor, I guess you can just go Hackintosh. Apple definitely is going to release a consumer tower with an i9