System Specs and Performance
Test Setup Specification
Test System Setup | |
Motherboard | ASUS Strix Z270E Gamine |
CPU | Intel Core i7-7700K Stock – 4.2GHz CPU and Ring Bus, 100MHz Base Clock Overclocked – 5.0GHz CPU, 4.7GHz Ring Bus, 100MHz Base Clock |
Memory | Corsair Vengeance LPX 16GB (2 x 8GB) DDR4-2666 modules Stock – 2400MHz, 16-18-18-35-1T, 1.225V Overclocked – 2667MHz, 16-18-18-35-2T, 1.225V |
Hard Drive | Intel 730 240GB SSD Western Digital Caviar Black 1TB SATA III HD |
Sound Card | On-board sound |
Video Card | NVIDIA GTX 780 3GB |
CPU Cooling | Koolance CPU-360i water block |
Video Drivers | NVIDIA 376.33 |
Power Supply | Corsair HX750 |
Operating System | Windows 10 Pro x64 Anniversary Release (build 1607) |
Thermal Paste | MG Chemicals 860-60G Silicone Heat Transfer Compound |
CPU Temperature Testing Methods
To best gage the effectiveness of the relidding procedure, system CPU temperature were taken with the CPU at idle and under load. To replicate CPU idle conditions, the system was rebooted and allowed to sit idle for 10 minutes. To replicate a stress system load, AIDA64 System Stability Test was used in conjunction with EVGA OC Scanner X for 30 minutes per run. After each run, the system was shut down and allowed to rest for 10 minutes to cool down. Then the CPU cooler was removed, cleaned, and remounted to the CPU with fresh thermal paste applied. This procedure was repeated a total of twelve times – three times each for the stock and overclocking speed runs before and after the re-lid procedure.
Temperature measurements were taken directly from the CPU thermistors using CoreTemp v1.5. Because of the volatile nature of the Kaby Lake thermistor readings, the system temperatures were measured as follows. For idle temperatures, the lowest recorded value was used for the run. For load temperatures, a series of three values were notated: the average (high and low) across all cores, the average (high and low) across the single hottest core, and the high temperature of the hottest core.
Note that the temperature values are reported as deltas rather than absolute temperatures with the delta value reported calculated as CPU temperature – ambient temperature. For all tests, room ambient temperature was maintained between 23-27C.
Intel Z270-based Kaby Lake System Testing
CPU Stock Speed Testing
The CPU stock speed testing was conducted with the BIOS defaults set for the CPU (including enabling of the CPU-integrated graphics processor) and Turbo Mode disabled, equating to a 4.2GHz CPU speed, 2400MHz memory speed, 4.2GHz ring bus speed, and 100MHz base clock. The Intel Speedstep functionality remained enabled for the duration of the testing to get realistic CPU idle performance conditions.
At stock processor speed, the processor temp fell an average of 10C across all measured areas, a significant decrease in temperature. If you assume a room temperature of 25C, that's the difference between a maximum core value of 69C in the stock configuration and 59C after the processor was relidded.
CPU Overclocked Speed Testing
The CPU overclocked speed testing was conducted with known stable settings from a previous board review, equating to a 5.0GHz CPU speed, 2667MHz memory speed, 4.7GHz ring bus speed, and 100MHz base clock. Also, the CPU-integrated graphics processor was disabled to reduce the processor heat generation and increase system stability. The Intel Speedstep functionality remained enabled for the duration of the testing to get realistic CPU idle performance conditions.
Board voltage settings were configured as follows:
- CPU Core Voltage – 1.31V
- DRAM Voltage – 1.23V
- All other settings set to Auto or stock settings
At overclocked speeds, the temperature differences between the factory configuration and the relidded processor were more significant. The relidded processor saw a temperature drop of 14C in its measured average temperature across all cores and a 20C drop in it maximum core temperature. To put those numbers in perspective, that's a drop from 94C to 74C on the maximum core temperature and 72C to 58C on the average temperature over all cores with a room temperature of 25C. That's quite a significant temperature drop not to mention the real probability of increased processor stability at those overclocked speeds.
Overclocking
With lowering the operating temperature of the processor comes the possibility of increased overclocking headroom. Prior to the relidding operation, the 7700K processor could reach a maximum speed of 5.0GHz with a 4.7GHz ring bus speed, operating at nearly chip crippling temperatures (read as 85-90C+). After the successfully relidding of the 7700k, the processor overclocked to a 5.1GHz core speed and a 4.8GHz ring bus speed with all cores enabled and only the graphics processor disabled for stability and heat reasons. The CPU voltage had to be bumped up a bit to 1.33V (from 1.31 voltage as 5.0GHz/4.7GHz operating speeds), but the temps did not increase dramatically as a result (with the maximium temperature showing a relatively cool 75C max core temperature in the screencap).
5.1GHz Core Speed with 4.8GHz Ring Bus Stats
Yay, Morry’s back!
Yay, Morry’s back!
Great article guys!
Great article guys!
I think the impressive thing
I think the impressive thing is that not only did you receive a 20 degree delta on hottest temps, but how much easier it is to do with these kits now.
Though the fact the Ryzen IHS and TIM is already pretty great means you dont need to do this etc etc.
Very true there. AMD took
Very true there. AMD took the solder route, similar to what Intel does with their LGA2011 processors. Makes heat transfer and dissipation much more efficient.
Intel TIM: Thermal Isolation
Intel TIM: Thermal Isolation Material, it works as advertised 😀
joking aside….
I think it is a nearly a crime that intel can sell $300+ dollar cpus without proper TIM. It seems that the material they used is even worse than ketchup.
@Morry, do you have any tips for us about the expected lifetime of the Coollaboratory Liquid Metal? Is it advised to re-apply after a year? And would you then recommend using the adhesive to glue back the heat spreader?
Liquid metal application
Liquid metal application should last for the life of the proc. This should be a "one and done" type procedure. You could do it more then once, but every time you crack the heatspreader off the PCB, you risk ruining the proc.
As far as re-adhering the heatspreader to the proc PCB, I used black RTV. relaively cheap, easily found as your local walmart or automotive store, and is electronics and plastic safe.
You *could* try naked cooling the CPU die, but I wouldn't recommend it, especially after my failed attempted with teh 4770K….
When HardOCP did their
When HardOCP did their de-lid/re-lid testing on the 7700k/7600k processors, they attempted naked die cooling and it sounded really risky. Basically, the head spreader also spread the clamping force necessary to adequately seat the processor into the socket. Without the heat spreader, all that downward force is is isolated to the bare die unless you make some kind of shim. I think their results with bare die cooling was that it was only a few degrees better than just re-lidding with the better TIM, but not worth the added risk to the die.
“Intel TIM: Thermal Isolation
“Intel TIM: Thermal Isolation Material, it works as advertised :D”
TIM is fine, it’s the separation between the IHS and die surfaces that is the issue. If you delid and then relid and don’t even touch the stock TIM, you’d get the same improvement in thermal performance.
https://forums.anandtech.com/threads/delidded-my-i7-3770k-loaded-temperatures-drop-by-20%C2%B0c-at-4-7ghz.2261855/page-23#post-34053183
While that forum post from a
While that forum post from a few years back remains relavent to the conversation, it discussed delidding of an entirely different processor. There may be an issue in the separation between IHS and die surface, but for the 7700 series, the TIM is most definately problematic.
Thanks…
@Morry, it feels like you’re
@Morry, it feels like you’re missing an image near the bottom of page 1.
We go from “Next, carefully snap the Re-Lid hold down..” to “At this point, you are almost done. Before applying the black RTV..” without showing the application of the RTV
Hopefully I’m not imagining things
yeah, I noticed that too.
yeah, I noticed that too. Unfortunately, I did not get a pic of the RTV applied to the heatspreader before putting humpty dumpty back together again. And I didn't want to press my luck after getting a good mating…
True, a good mate is hard to
True, a good mate is hard to find.
Maybe moving the picture with the headspeader down so it sits above “Screw the center screw into the torque plate and then carefully fix the torque plate in place over the…” would help?
Good idea, thanks for the
Good idea, thanks for the input…
Morry – Do you have any
Morry – Do you have any recommendations for a thermal paste that would last for years like the original TIM would? My only concern with de/re-lidding is you might need to reapply the paste in 2-3 years as it does dry out over time..
not entirely sure. Some
not entirely sure. Some reports say that liquid ultra does degrade within a year, but on most of those forum posts, people were using direct die cooling without an IHS installed. That introduces direct air contact as well as possible humidity affects. The IHS would better protect the die area and the TIM from "open air" type affects…
Good point. Thanks!
Good point. Thanks!
Morry, it looks like the
Morry, it looks like the original heatspreader adhesive application left a small air gap. The adhesive does not appear to make a closed loop all the way around the perimeter of the heatspreader. Did you replicate this when you re-adhered the HS to the PCB? If so, doesn’t that allow air under the HS and potentially cause the open air/humidity issues?
I’m not sure if I left the
I'm not sure if I left the gap when I resealed it. While it is true that the gap would leave a section for air exchange, the IHS is very securely fixed to the PCB surface, so the gap is minimal. So it still wouldn't be as greatly affected as a naked die would…
They leave an opening to let
They leave an opening to let out air pressure created as the die heats up. Old processor IHS’s used to have a small hole for this.
Sorry your onclusion is
Sorry your onclusion is wrong. Conclusion should be that Intel is doing shit this time. How come an experienced CPU manufacturer like Intel does not make it better than all what they did in the past? Someone at Intel is paid way too much and should be fired. Teenagers still at school can do better.
Scotch tape… for real?! You
Scotch tape… for real?! You never heard of ESD i take it.
From now on use Anti Static Tape. Kapton Tape is my choice.
Nevertheless, delidding IHS
Nevertheless, delidding IHS voids warranty, and may possibly damage CPU unless he/she uses proper delidding tool.
Be careful with that Gallium.
Be careful with that Gallium. It eats aluminum and corrodes copper.
http://www.overclock.net/t/1313179/official-delidded-club-guide/14340#post_19550310
Learn some stuff,
Learn some stuff, guys:
http://overclocking.guide/the-truth-about-cpu-soldering/
When you consider the thermal density of these small dies, it is much easier to have micro-cracks in the solder due to constant thermal cycles, which will increase the chance of killing the CPU. That’s the reason why Intel doesn’t solder their mainstream chips ever since Ivy Bridge. They could get away with it with sandy because it was on a larger node.
While paste doesn’t perform as well as solder, it eliminates this issue, which is why the Engineers went with it.
The reason why liquid metal wasn’t used is because it was not necessary. As you can see in most reviews, you’re only gaining a few degrees from your typical paste vs. liquid metal. Hardly meaningful from a temperature standpoint, and especially from a cost standpoint. Remember, Intel is making hundreds of thousands of these CPUs, and liquid metal isn’t cheap, nor is it as easy to apply to a product as paste is.
The reason why we are seeing increased temperatures isn’t primarily because of the TIM used, but because of the gap between the heatspreader and the die. The glue they use to stick the heatspreader to the die raises it away from the die. As you can see in PCPer’s picture, they took all of that glue off, so of course heat transfer is going to be more efficient due to it being closer. But doing so greatly increases the risk of damaging the die, but so does delidding.
4.8GHz is very easy to get to
4.8GHz is very easy to get to with a decent cooler and airflow. So I’d be looking at a 300MHz increase if all goes well, or a 6.25% increase, for about $50.
On a CPU limited game at 50fps that’s an extra 3.1fps. On a compiler run that’s CPU limited it’ll be 56.5 seconds instead of 60.
I love reading this stuff but I guess I’m old and jaded anymore, it’s cool for bragging rights though 😉 Thanks for the write-up and the pictures, I like that you kept it so concise and informative.
Yeah, exactly.
It’s why I
Yeah, exactly.
It’s why I don’t bother to overclock any more.
However, I have one of those Eurocom X9E3 laptops using a desktop i7-7700k processor and before I delidded and used some OCP I had around from an old overclock project I was throttling at 99C under full load in a 21C room.
After delidding and OCP application (using standard Arctic Silver 5 between the top of the HS and the heat spreader) it max’s out at 78C, no throttling and idles at 34C instead of 40C in similar conditions.
So if you are having cooling issues and can’t use a better CPU cooler (as I couldn’t, laptop and all) a delid and OCP does seriously help out.