7700k temps vs. 6700k

[frankla 44]

I have read that 7700k runs much hotter than 6700k which gives me a pause on getting it in order to get very little benefit.   What do you guys think?

[martin-w 3,234]

Are you saying that you're upgrading from a 6700K? If so, no point, the IPC is more or less the same. RAM is faster and other advantages like Optane support but other than that a pointless exercise from such a recent platform.

If from an older platform, which one?

Re temps, 7700K overclocks further than 6700K and temps are thus higher. But at the same frequency, temps are the same from what I've seen. Silicone lottery applies of course so all chips vary. Some hotter than others. And of course how well Intel did re that particular chips die/IHS interface is a factor, your cooling, ambient temp etc.

[frankla 44]

6 hours ago, martin-w said:

Are you saying that you're upgrading from a 6700K? If so, no point, the IPC is more or less the same. RAM is faster and other advantages like Optane support but other than that a pointless exercise from such a recent platform.

If from an older platform, which one?

Re temps, 7700K overclocks further than 6700K and temps are thus higher. But at the same frequency, temps are the same from what I've seen. Silicone lottery applies of course so all chips vary. Some hotter than others. And of course how well Intel did re that particular chips die/IHS interface is a factor, your cooling, ambient temp etc.

I have an old bloomfiled I7.     The article that gave me a pause was this:  http://www.tweaktown.com/news/55219/7700k-vs-6700k-much-hotter-barely-faster/index.html

"But the temperatures of Intel's new Kaby Lake architecture are worrying, with Tom's Hardware noting that the new 7700K was hitting 54C under gaming loads, while the 6700K runs relatively cool in comparison at just 34C. When overclocked, we're looking at 61C for the 7700K while the 6700K hits just 48C."

To me running a system 25% hotter for 1-2 percent frame rate increase seems like not a good idea.  Am I wrong?  Or is Tom's Hardware test flawed?

[lownslo 592]

2 hours ago, frankla said:

 Or is Tom's Hardware test flawed?

Not saying the test was flawed, but you should perhaps look to other tests and sites for more info.

Greg

[Tabs 2,790]

I've seen mine spike to over 80C at stock speed in FSX - I have an NZXT Kraken x62 closed loop water cooler too, one of the best ones out there. It's definitely a really hot chip. Thinking about delidding it.

[OneWhoKnocks53 51]

2 hours ago, Tabs said:

I've seen mine spike to over 80C at stock speed in FSX - I have an NZXT Kraken x62 closed loop water cooler too, one of the best ones out there. It's definitely a really hot chip. Thinking about delidding it.

Delliding is a godsend for 6700k and 7700k.  I do not exaggerate when I say I saw a 25c decrease in temps after delliding.  Use a gallium-based paste between the die and IHS.  The thermal headroom gained is phenomenal.  I used the Delid Die-Mate, made the job painless and quick.  Use some electrical grade firm RTV to re-secure the IHS.

[angeli662 43]

Intel tells Core i7-7700K owners to stop overclocking to avoid high temps

http://www.pcgamer.com/intels-tells-core-i7-7700k-owners-to-stop-overclocking-to-avoid-high-temps/

[martin-w 3,234]

10 hours ago, Tabs said:

I've seen mine spike to over 80C at stock speed in FSX - I have an NZXT Kraken x62 closed loop water cooler too, one of the best ones out there. It's definitely a really hot chip. Thinking about delidding it.

 

Yep, it will do Ryan. Brief spikes are common with Kaby and to a somewhat lesser degree with Skylake. That's how the CPU is supposed to work. The spikes are brief though so not an issue. Spikes like that occur whether you are overclocking or at stock. 

You'll notice the same with Skylake, although not as obvious. It's been suggested by our own Techguymax, in the Intel thread in question, that it's relates to how Turbo is supposed to work. Perhaps he'll comment when he sees this. Others have pointed out that Intel "Speed Shift" has been changed for Kaby Lake hence temp spikes being more noticeable. 

[martin-w 3,234]

7 hours ago, angeli662 said:

Intel tells Core i7-7700K owners to stop overclocking to avoid high temps

http://www.pcgamer.com/intels-tells-core-i7-7700k-owners-to-stop-overclocking-to-avoid-high-temps/

 

No, this has been taken way out of context. Intel haven't told anyone to stop overclocking. What they said in the original Intel forum thread where this kicked off was, "Intel don't recommend running outside of the processor specifications". This comment has been misinterpreted and misquoted.

Intel have never "recommended" overclocking for any platform. If they did, they would be encouraging all manner of inexperienced individuals to dive in and overclock, they would then have to cover all manner of warranty claims for all manner of associated components. Intel's legal department make sure the company doesn't "encourage" overclocking.

However... Intel do recognise that we enthusiasts like to overclock, so they provide "K" series CPU's with that in mind. In addition, Intel provide the Performance Tuning Plan that actually provides a warranty for overclocking. Ruin your precious chip overclocking and Intel will send you a brand new one.

From the Intel Performance tuning Plan...

 

Quote

Does this mean that Intel is supporting or encouraging overclocking?
No. While we will, under the Plan, replace an eligible processor that fails while running outside of Intel’s specifications, we will not provide any assistance with configuration, data recovery, failure of associated parts, or any other activities or issues associated with the processor or system resulting from overclocking or otherwise running outside of Intel’s published specifications.

 

 So to sum up, the temp spikes that kicked all of this off on the Intel forum are normal behaviour, brief and not an issue. It's how Turbo/Speed Shift etc is supposed to work. And Intel have not at all suddenly decided we shouldn't overclock Kaby.

As Intel confirmed in the thread, they have tested the CPU and all is as it should be.

To be honest it's fuss about nothing. If fans ramping up and down are an issue due to this normal behaviour, then simply set a fan spin up and fan spin down delay in the bios. I recall mine is set to 4 seconds. 

 

 

[martin-w 3,234]

On 5/7/2017 at 2:50 PM, frankla said:

I have an old bloomfiled I7.     The article that gave me a pause was this:  http://www.tweaktown.com/news/55219/7700k-vs-6700k-much-hotter-barely-faster/index.html

"But the temperatures of Intel's new Kaby Lake architecture are worrying, with Tom's Hardware noting that the new 7700K was hitting 54C under gaming loads, while the 6700K runs relatively cool in comparison at just 34C. When overclocked, we're looking at 61C for the 7700K while the 6700K hits just 48C."

To me running a system 25% hotter for 1-2 percent frame rate increase seems like not a good idea.  Am I wrong?  Or is Tom's Hardware test flawed?

 

I'd agree with Greg... don't just listen to Toms Hardware.

A 7700K hitting a meagre 54 degrees when gaming is not at all "worrying" as Tom's Hardware put it. That claim is nonsense. And NO, I've never come across a 6700K that runs as cool as 34 degrees when heavily gaming. 34 degrees is close to an IDLE temperature. You may achieve such a low temp with a  custom loop and no overclocking, but other than that, no.

 

61 degrees while gaming, as pointed out is a perfectly acceptable temp. Nobody should be concerned about 61 degrees. 

There's a multitude of individuals on this forum that have purchased Kaby Lake, many of them overlooking up to 5 GHz. Some delidded, beyond that frequency. 

 

[TechguyMaxC 516]

There's a large thread about this on Intel's own forums.  I've made a few posts there recently in an attempt to explain things.  Please be aware, this is a very technical subject.  If you don't understand something I'm happy to answer questions.  

Intel's CPUs have a collection of technologies designed to allow them to quickly alter their clock speed (you will recognize this as the "MHz" or "GHz" rating of your particular processor) in order to respond to ever-changing demands of software.  These technologies are known as Turbo and Speedshift.  Turbo refers to the ability to raise the clock speed of a processor above the rated or "base" clock speed.  e.g. an i7 7700k has a base clock speed of 4.2GHz and a maximum Turbo clock speed of 4.5GHz.  The mechanism for altering clock speed falls under the category of the Speedshift technology, which allows clock speeds to change in either direction very rapidly (in milliseconds).  

The behavior some are observing which is now being reported upon by the tech press sounds alarming.  The claim is that Intel CPUs are running very hot, much hotter than they should, and that this is causing adverse affects and may eventually cause processor failure.  These claims are false, as they are based on misinformation and lack of understanding.  I will attempt to clear up the matter for anyone who wishes to read the following:

 

 

There is speculation among the community and by tech press that the reported high temperatures are damaging/could damage CPUs.  This is false.  The observed temperatures peaks are transient spikes, i.e. very short.  There is an implication in these claims that Intel CPUs have a high overall temperature but facts show otherwise.  The reported temperature spikes last only a second or less, and there are very few reports that these maximum temperatures have reached the point at which technologies designed to protect the processor will kick in and do their job.  Most reports have spikes to somewhere in the range of 15-25C below this threshold.  Again, it is of critical importance to understanding this issue that we acknowledge the duration of these spikes.  Here is a graph of temperature over time as well as a corresponding graph of CPU utilization over time posted by a user on the Intel forums:

The top graph shows temperature spikes on individual cores.  You can make several observations from these graphs, and draw several conclusions based on those observations.  First, there are upward spikes in temperature, but they quickly fall back to preceding values.  Second, these temperature spikes only reach a maximum observed temperature of just below 70C, which is at least 35C distance from the maximum allowable temperature AKA Tj Max (Thermal Junction Maximum).  If a chip reaches this value, the protection mechanisms kick in to prevent the processor from operating at this dangerous temperature for too long and potentially incurring damage.  A third observation is that there is CPU activity, which does not persist (i.e. is not constant).

Based on the first observation it is possible to draw the conclusion that these spikes are not very energetic.  This is due to the quick rise in temperature and immediate return to the prior state.  Based on the second observation we conclude that these spikes are not dangerous, as they remain a very safe distance from the Tj Max value at which "thermal throttling" occurs.  Intel themselves have defined the maximum allowable temperature of 105C thanks to testing which shows that a processor which operates at this temperature for too long can sustain damage and eventually become unusable.  This is why protection mechanisms exist, and only kick in at this value and not before.  The risk to the processor's lifespan is sufficiently low when operating below this value that thermal throttling need not be implemented.  Based on the third observation we can draw the conclusion that there is in fact work to be done by the CPU.  One of the central claims of the argument is that these temperature spikes can be observed in an "idle" state.  While it may be true that a given user might not be directly interacting with their computer (i.e. giving it commands issued via inputs such as the keyboard/mouse) that does not mean the computer is in an "idle" state in the sense of having no work to do.  Modern Operating Systems often use "idle" states (in this case idle means periods of no user activity) to perform background tasks such as checking for software updates, preparing to install them, or in the case of the user running the monitoring program, this program is itself generating activity.  

Now, having established that these spikes are occurring, but are not in fact dangerous, let us address why this is occurring.

The observed behavior may seem abnormal, at a glance, but once you understand the intent of the CPU design team everything becomes clear.  Modern processor designs feature both power saving as well as speed boosting technologies.  Thanks to the prevalence of mobile devices, there are competing needs for both speed and battery life.  The faster your processor design, the more power it will consume, and the shorter your battery life will be.  However, this is only true if a processor consumes high levels of power over long periods of time.  Most users of mobile devices do not use their devices in a way which requires high CPU performance (and thus high power consumption) over long periods of time.  The usage patterns we see require only momentary bursts of speed to accomplish tasks quickly.  Companies that design processors have created a design methodology which allows them to satisfy both demands (high performance and high battery life).  This is where the idea of Speedshift (and similar technologies) is born.

Speedshift is a term for a collection of technologies which allow a processor to quickly respond to a request for work to be performed.  The theory behind it states that the quicker a task can be completed, the faster the processor can return to an "idle" state and power saving techniques can be enabled.  The result is that the overall or total power consumption of the processor will be lower than a slower performing processor which must remain in a high power state for a longer period of time.  The difference in power consumption between low and high power states can be of an order of magnitude.  If for example a processor consumes 90W of power in its highest power state but only say 6W at idle, it is highly beneficial to return to that idle state as quickly as possible.

Let's return to the graph for a moment in order to see Speedshift in action.  The temperature spikes we observe can be attributed to the processor responding to a task, a request for work to be performed which has been made by some piece of software.  The graphs do not tell us which particular piece of software in this case, but it is not relevant since this behavior can be produced by any software, or combinations of software.  The most likely scenario here is that the user of the PC is not actively using the computer, not providing any input via keyboard or mouse, but has numerous background applications running (as well as the temperature/CPU use graph utility itself) which are in fact creating workload(s) i.e. tasks for the CPU.  Now if we combine our observations of both graphs, having established that there is work for the CPU to perform thanks to tasks being generated by background applications or the Operating System itself, we begin to see just what is occurring.  The processor, in a response to requests for work, is quickly ramping up from a low power state to a high power state in order to complete that work as quickly as possible.  This is evidenced by both the spikes in CPU use in the bottom graph as well as the spikes in temperature in the top graph.  Note: these lines do not perfectly overlap nor should they be expect to do so.  The graphs simply do not have the granularity to depict workload in real time.  After completing a given task (or portion of a workload) the processor temperature quickly drops back down as the processor itself returns to an "idle" state.  The next piece of work can then begin, causing another spike in temperature, and once it completes, another return to idle, ad infinitum.

In summation, the observed behavior is explainable thanks to the design of the CPUs in question, and it is in no way dangerous.  

 

[martin-w 3,234]

Fantastic explanation Max. Thanks for taking the time to post it. 

[brucets 94]

Thank God for M & M!! You guys make this forum!!

 

Bruce

[TechguyMaxC 516]

Don't forget Hasse (westman)! He's tested more CPUs than you can shake a stick at!

[martin-w 3,234]

11 hours ago, brucets said:

Thank God for M & M!! You guys make this forum!!

 

Bruce

 

M&M featuring Westman!

It's a brand new Rap/Rock band coming to a venue near you very soon!

Tickets on sale shortly.