Alienware Aurora R7 (8700K) Overclocking Questions

[hugin7]

The Alienware Aurora R7 PC (Intel Core i7 8700K – Coffee Lake, liquid cooling) comes with software to configure overclocking via presets or custom settings. I experimented with these settings and it appears that my PC gives the best results with the “Stage 2” preset and Windows power management set to performance mode:

4.8GHz
Voltage – 1.12 (static) in system BIOS, though CPU-Z and Core Temp VID fluctuates between 1.2 and 1.32 v
Memory – XMP 1 (2800MHz, 1.3V)

I tested via IntelBurnTest and Prime95 noticed that even at extended 100% CPU load and temperatures below 77, the core speed is usually around 4.5GHz, not 4.8. Every now and then, it spikes to 4.8GHz. Is this normal behavior or am I experiencing instability? Could it be that Intel Turbo Boost is taken into account and 4.8GHz really means that the system normally runs at 4.5 and is occasionally boosted to 4.8?

My other concern is with appropriate temperatures for this CPU. Short-term tests with IntelBurnTest and Prime95 (under 40 min) kept the temps at 79 or below for all cores. Normal gaming usage had the temps peak at 76 after several hours. However, after I ran Prime95 for 11 hours, temps peaked at 87. I suspect that the highest temps were reached shortly after I started the test after gaming, so the GPU was radiating some heat.

Should I keep my current settings or do they appear too aggressive for safe long-term use? Thanks in advance!

 

[Darkbreeze]

Few questions. What VERSION of Prime95 are you using to test?

What CPU cooler are you using?

What, exactly, are the rest of your hardware model numbers including motherboard, power supply, memory modules, graphics card and case (Including how many case fans and in what orientation, intake or exhaust, they are configured for)?

87 degrees is getting past where you want to see things at and I suspect this is probably due to less than adequate CPU and case cooling.

Is your GPU card also overclocked?

Overclocking via presets is a poor way to overclock anything except possibly graphics cards. CPU and memory overclocking should always be done in the bios and I'd be happy to help you with that if you are still trying to overclock.

Since the 8700k already has a 4.7Ghz peak single core boost clock, it's probably not even necessary to OC this chip to begin with, but if you'd still like to then I'd suggest removing any automatic utility related preset settings and configuring your overclock in the BIOS where you have much better granular control of things like the voltage, uncore, system agent, memory and other settings. Automatic overclocks pretty much ALWAYS overvolt any overclock in order to err on the side of caution when it comes to stability. Often a completely stable overclock can be achieved using far less voltage than what is used when overclocking via utility or presets.

 

[hugin7]

I’m using version 29.4 of Prime95.

Unfortunately, Alienware doesn’t publish full specifications of the components that go into their build and some of the main components are proprietary, including the motherboard and cooler. As a result, rich BIOS settings available in most motherboards are inaccessible, which limits manual tweaking options for OC. I was aware of the limitations when I purchased the PC, so I know that I’ll have to settle for a modest OC compared to a full custom build with equal specs.

My PSU is 850 watts. The GPU is a GTX 1070 Ti and is not overclocked. Memory is HyperX FURY DDR4 XMP at 2933 MHz (16GB). From what I’ve heard of Alienware cases, they tend to be cramped and have less than ideal air flow.

When I ran the machine at 4.8GHz, I was getting random crashes a couple of times per day. For the past couple of days, I scaled back to 4.6GHz and it hasn’t crashed. Temps went down, too. I still have fluctuations in my core speed, except now it’s usually around 4.3GHz with spikes up to 4.6 – is this normal? The voltage also fluctuates as described above. Oddly enough, it seems that Alienware went against convention and decided to under-volt the machine.

Few questions. What VERSION of Prime95 are you using to test?

What CPU cooler are you using?

What, exactly, are the rest of your hardware model numbers including motherboard, power supply, memory modules, graphics card and case (Including how many case fans and in what orientation, intake or exhaust, they are configured for)?

87 degrees is getting past where you want to see things at and I suspect this is probably due to less than adequate CPU and case cooling.

Is your GPU card also overclocked?

Overclocking via presets is a poor way to overclock anything except possibly graphics cards. CPU and memory overclocking should always be done in the bios and I'd be happy to help you with that if you are still trying to overclock.

Since the 8700k already has a 4.7Ghz peak single core boost clock, it's probably not even necessary to OC this chip to begin with, but if you'd still like to then I'd suggest removing any automatic utility related preset settings and configuring your overclock in the BIOS where you have much better granular control of things like the voltage, uncore, system agent, memory and other settings. Automatic overclocks pretty much ALWAYS overvolt any overclock in order to err on the side of caution when it comes to stability. Often a completely stable overclock can be achieved using far less voltage than what is used when overclocking via utility or presets.

 

[Darkbreeze]

Unfortunately, with a proprietary board and custom bios that is probably defeating of your attempts to overclock by way of utility due to their specific implementation of EIST it might not be possible to get a solid grasp on your overclock with this system.

There's a reason overclocking by way of BIOS is recommended over the use of utilities and when the OEM works against you, it makes it even worse. What you might try doing to help limit or eliminate the fluctuations at least is going into the control panel power settings and change the default profile to performance. That's going to keep the CPU at full clock speed at all times though with no ability for it to reduce clock speed or power during idle or conditions where you don't really need full clock speed on all cores. If you simply want to stop the fluctuations being as dramatic as they are now, that will work.

Also, you can go into the advanced power profile CPU settings and change the minimum value to anything between 5-15%, so you could allow the CPU to drop to 50% when idle or if both min and max are set to 100%, which is what the performance profile does, it should not drop at all.

With my overclocks, I generally set values to 100/100 while configuring and testing the overclock and then change the minimum value back to 5% afterwards to reduce power consumption and help with cooling after I've determined the overclock is stable. A core that is always at 100% is much harder to keep cool than one that is only at 100% when it needs to be. Cores change from zero to full clock rates in hundredths of milliseconds so this is not something that will normally have any noticeably negative affect on performance.

IF you can find an option to turn OFF the turbo boost option in the utility you are using or in the BIOS, that would also go a long way in helping to reduce fluctuations and to also assure that the assigned clock/voltage values do not go beyond your designated values which can affect stability and cooling. Honestly I think anything over 4.7Ghz is too aggressive on the 8700k because of the additional two physical cores that add considerable heat to the package temps over what you could expect or might be used to with a 4/8 CPU.

Also, what is the model of the liquid cooler that came with your system?

 

[hugin7]

The test results I posted are with performance profile on, which makes the fluctuations all the more surprising. Oddly enough, my CPU seems to stay cooler during stress tests with the performance profile, though I should probably disable it for the long haul for better durability.

The cooler is one of those “mystery” proprietary components. I suspect that it’s a fairly low-end liquid cooler.

Considering that going from 4.8 to 4.6GHz greatly improved the temps and stability of my system, perhaps it’s best to stick with this configuration. If I wasn’t stuck with a proprietary motherboard and so-so cooling, I could probably push out a bit more, but due to the machine’s limitations, it’s probably not worth the risk.

Unfortunately, with a proprietary board and custom bios that is probably defeating of your attempts to overclock by way of utility due to their specific implementation of EIST it might not be possible to get a solid grasp on your overclock with this system.

There's a reason overclocking by way of BIOS is recommended over the use of utilities and when the OEM works against you, it makes it even worse. What you might try doing to help limit or eliminate the fluctuations at least is going into the control panel power settings and change the default profile to performance. That's going to keep the CPU at full clock speed at all times though with no ability for it to reduce clock speed or power during idle or conditions where you don't really need full clock speed on all cores. If you simply want to stop the fluctuations being as dramatic as they are now, that will work.

Also, you can go into the advanced power profile CPU settings and change the minimum value to anything between 5-15%, so you could allow the CPU to drop to 50% when idle or if both min and max are set to 100%, which is what the performance profile does, it should not drop at all.

With my overclocks, I generally set values to 100/100 while configuring and testing the overclock and then change the minimum value back to 5% afterwards to reduce power consumption and help with cooling after I've determined the overclock is stable. A core that is always at 100% is much harder to keep cool than one that is only at 100% when it needs to be. Cores change from zero to full clock rates in hundredths of milliseconds so this is not something that will normally have any noticeably negative affect on performance.

IF you can find an option to turn OFF the turbo boost option in the utility you are using or in the BIOS, that would also go a long way in helping to reduce fluctuations and to also assure that the assigned clock/voltage values do not go beyond your designated values which can affect stability and cooling. Honestly I think anything over 4.7Ghz is too aggressive on the 8700k because of the additional two physical cores that add considerable heat to the package temps over what you could expect or might be used to with a 4/8 CPU.

Also, what is the model of the liquid cooler that came with your system?

 

[Darkbreeze]

I'd have to probably agree, especially until or unless you're able to swap that generic cooling solution out for a better one. Is it a single or double fan width (120/140 vs 240/280) liquid cooler?

Still, as I said, you can likely limit or almost entirely remove the fluctuations by changing the power profile if it bothers you or affects the normal function but sticking it at 100% full time will definitely cause an increase in overall temps which is why I don't generally do that except while doing stability/thermal testing.

 

[hugin7]

I searched the PC specs for fan info and came up empty, so it’s likely proprietary. With the latest settings, my temps peak at 68.

I'd have to probably agree, especially until or unless you're able to swap that generic cooling solution out for a better one. Is it a single or double fan width (120/140 vs 240/280) liquid cooler?

Still, as I said, you can likely limit or almost entirely remove the fluctuations by changing the power profile if it bothers you or affects the normal function but sticking it at 100% full time will definitely cause an increase in overall temps which is why I don't generally do that except while doing stability/thermal testing.

 

[Darkbreeze]

Is that temp under normal conditions or is it with all cores at 100% load?

 

[hugin7]

It’s the maximum temp recorded after a full day’s use with a lot of gaming.

Is that temp under normal conditions or is it with all cores at 100% load?

 

[Darkbreeze]

So, normal use, not under stress. I guess if that's good enough for you, then it's good enough, but anytime there is any kind of overclock the system SHOULD be stress/stability tested under 100% load on all cores, otherwise you have no way of knowing if there is silent data corruption happening. If this is a gaming only system, then probably not too critical. If you run any kind of applications and your overclock is not stable based on full stress 15+ hours of testing then I'd be concerned about it, plus, the operating system is still a concern and a great many systems that seem fine at the beginning but later begin having mysterious crashes or unsolvable errors can usually be traced to an overclock that was never proven to be stable using a standardized method.

Considering that somewhere around 68-73, and certainly no higher EVER than 80, under a full 15 minute thermal stress test, I think that for normal operation 68 degrees is probably a bit higher than you'd really like to see as a temp that high under normal conditions, no matter how long it's been running for, likely means you would not remain within the recommended thermal envelope during stress testing. Possibly, but it would likely be cutting it awfully close. So, that's up to you though. If you're good with it, not much else to say.

 

[hugin7]

I ran a Prime95 test overnight for 10 hours with the new settings and reached a peak temp of 75. All tests passed. Does this appear reasonable and sustainable for an 8700k or still a bit aggressive?

So, normal use, not under stress. I guess if that's good enough for you, then it's good enough, but anytime there is any kind of overclock the system SHOULD be stress/stability tested under 100% load on all cores, otherwise you have no way of knowing if there is silent data corruption happening. If this is a gaming only system, then probably not too critical. If you run any kind of applications and your overclock is not stable based on full stress 15+ hours of testing then I'd be concerned about it, plus, the operating system is still a concern and a great many systems that seem fine at the beginning but later begin having mysterious crashes or unsolvable errors can usually be traced to an overclock that was never proven to be stable using a standardized method.

Considering that somewhere around 68-73, and certainly no higher EVER than 80, under a full 15 minute thermal stress test, I think that for normal operation 68 degrees is probably a bit higher than you'd really like to see as a temp that high under normal conditions, no matter how long it's been running for, likely means you would not remain within the recommended thermal envelope during stress testing. Possibly, but it would likely be cutting it awfully close. So, that's up to you though. If you're good with it, not much else to say.

 

[Darkbreeze]

No, that seems pretty good. As I said, there can still be stability issues that are prominent enough to actually affect precision at anything below 24 hours, 15 at the very least, so when you have more time and it won't affect you to let it run longer I would do that just to be sure, but for now I'd say ten hours offers pretty fair insurance for the short term. Thermals are fine at 75. I like my rigs to stay below 70, so might want to work on fine tuning the cooling some more at your leisure. Maybe higher airflow or higher static pressure fans at some point, but you are still about a good five degrees within spec so technically, it's ok.

 

[hugin7]

My system switched to Windows power saver mode on its own and I checked with Alienware support on this issue. They installed a driver and said that it should fix the problem, but while they were at it, they recommended leaving the system in high performance mode permanently. They claim that it works more efficiently with overclocking than balanced mode.

Seems that there are two conflicting schools of thought on this… Many say that balanced mode is smart enough to properly utilize your CPU and GPU when pushed to the limit and helps your system’s longevity at rest, while others agree with the manufacturer.

No, that seems pretty good. As I said, there can still be stability issues that are prominent enough to actually affect precision at anything below 24 hours, 15 at the very least, so when you have more time and it won't affect you to let it run longer I would do that just to be sure, but for now I'd say ten hours offers pretty fair insurance for the short term. Thermals are fine at 75. I like my rigs to stay below 70, so might want to work on fine tuning the cooling some more at your leisure. Maybe higher airflow or higher static pressure fans at some point, but you are still about a good five degrees within spec so technically, it's ok.

 

[Darkbreeze]

They are stupid. And, I never said anything about using balanced mode. What you want to do, and I've run fifty or more, likely a lot more, overclocked systems this way with zero problems, is set the system to performance, then go into the advanced settings for the performance profile and set the minimum processor power state to 5 or 10% and leave the maximum power state at 100%. I generally use 5% and have never had an issue on any system configured that way.

Also, in the bios, make sure that EIST (Enhanced Intel speed step) is enabled, if there are any user configurable settings for that which usually there are even on OEM boards. Also, did we ever discuss the power supply? I don't think we did. There could be issues related to the power supply being group regulated, especially on a system like this that almost certainly uses an OEM power supply.

The 8700k uses the C7 low power core state which can cause crossloading issues on the power supply if it is a group regulated platform. Many OEM's pay no attention to this and I've seen plenty of them that came with incompatible power supplies for the CPU platform even though you'd think this would be something that they'd have already considered and factored into the build. It can result in a number of problems such as instabiilty, blue screens or just shutting right the hell down.

Take a look at the back of your machine and tell me what the model number listed on the specifications sticker says.

 

[hugin7]

Thank you, I adjusted “Minimum Processor State” for performance mode from 100% to 5%.

The guide at http://topics-cdn.dell.com/pdf/alienware-aurora-r7-desktop_service%20manual_en-us.pdf shows all of the settings I have in my BIOS. Unfortunately, I couldn’t find anything for EIST.

I also couldn’t locate the specific PSU model, but I did find the info below. I pulled it from the manual at http://topics-cdn.dell.com/pdf/alienware-aurora-r7-desktop_specifications_en-us.pdf.

Power ratings
Table 14. Power ratings specifications
Input voltage
100 VAC–240 VAC
Input frequency
50 Hz–60 Hz
Temperature range
Operating
5°C to 50°C (41°F to 122°F)
Storage
–40°C to 70°C (–40°F to 158°F)

Type
850 W

Input current (maximum):
10 A

Output current:
3.3V/20A, 5V/20A,
12VA/32A, 12VB/48A,
12VD/16A, -12V/0.5A,
5Vaux/4A

Rated output voltage:
3.3V, 5V, 12VA, 12VB,
12VD, -12V, 5Vaux

They are stupid. And, I never said anything about using balanced mode. What you want to do, and I've run fifty or more, likely a lot more, overclocked systems this way with zero problems, is set the system to performance, then go into the advanced settings for the performance profile and set the minimum processor power state to 5 or 10% and leave the maximum power state at 100%. I generally use 5% and have never had an issue on any system configured that way.

Also, in the bios, make sure that EIST (Enhanced Intel speed step) is enabled, if there are any user configurable settings for that which usually there are even on OEM boards. Also, did we ever discuss the power supply? I don't think we did. There could be issues related to the power supply being group regulated, especially on a system like this that almost certainly uses an OEM power supply.

The 8700k uses the C7 low power core state which can cause crossloading issues on the power supply if it is a group regulated platform. Many OEM's pay no attention to this and I've seen plenty of them that came with incompatible power supplies for the CPU platform even though you'd think this would be something that they'd have already considered and factored into the build. It can result in a number of problems such as instabiilty, blue screens or just shutting right the hell down.

Take a look at the back of your machine and tell me what the model number listed on the specifications sticker says.

 

[Darkbreeze]

There has to be a model number of the power supply somewhere. It's either on the back or on one of the sides. By law companies cannot sell power supplies, not even for use in OEM systems, without the power specifications and model number being on the unit.

It might be covered by part of the case if it's mounted on the back or it could be on one of the sides. I'd unplug it from the wall, remove the four screws holding it to the case and just remove it enough that you can look on each side/back until you can see the specifications sticker or model number. Has to be on there somewhere or it's in violation of law.

Either that, or somebody removed the specifications sticker, but even then you'd be able to probably see a square where it is cleaner than the rest of the unit where it was removed. I need the model to verify the internal platform. That other info is great, but not really useful for this purpose.

 

[hugin7]

I opened the case and found the following numbers on the PSU:

CN-01XCK3-PE200
-79I-0511-A00
DP/N 01XCK3

There has to be a model number of the power supply somewhere. It's either on the back or on one of the sides. By law companies cannot sell power supplies, not even for use in OEM systems, without the power specifications and model number being on the unit.

It might be covered by part of the case if it's mounted on the back or it could be on one of the sides. I'd unplug it from the wall, remove the four screws holding it to the case and just remove it enough that you can look on each side/back until you can see the specifications sticker or model number. Has to be on there somewhere or it's in violation of law.

Either that, or somebody removed the specifications sticker, but even then you'd be able to probably see a square where it is cleaner than the rest of the unit where it was removed. I need the model to verify the internal platform. That other info is great, but not really useful for this purpose.

 

[Darkbreeze]

Is there a brand name on the Power supply? Is this a Corsair unit in there?


Those numbers did not bring up anything AT ALL, on a Google search, so they are probably internal numbers only and not related to the model. Somewhere on that unit there is a label similar to this. It might not be exactly like this but there should be something at least similar.

So those voltage specifications that are listed? Those are mandatory by law. So is the model number. So someplace there is something that looks similar to this. It may be on one of the sides that is not visible to you, depending on how the PSU was installed, you might have to take it free and turn it in order to see it, but it's got to be there somewhere unless somebody removed it and then you would be able to tell that it was removed.

This one, which MIGHT even be the exact unit that is in your system, has it on top. Yours might have it on top, one of the side, on the bottom, but it is somewhere.

 

[hugin7]

I couldn’t find additional info and sent an e-mail to Alienware support to provide it.

Is there a brand name on the Power supply? Is this a Corsair unit in there?


Those numbers did not bring up anything AT ALL, on a Google search, so they are probably internal numbers only and not related to the model. Somewhere on that unit there is a label similar to this. It might not be exactly like this but there should be something at least similar.

So those voltage specifications that are listed? Those are mandatory by law. So is the model number. So someplace there is something that looks similar to this. It may be on one of the sides that is not visible to you, depending on how the PSU was installed, you might have to take it free and turn it in order to see it, but it's got to be there somewhere unless somebody removed it and then you would be able to tell that it was removed.

This one, which MIGHT even be the exact unit that is in your system, has it on top. Yours might have it on top, one of the side, on the bottom, but it is somewhere.

 

[Darkbreeze]

Cool. Let me know what they tell you. Very unusual for this information not to be on the unit, so I'd actually have wanted to do that myself if it was my unit. In the meantime, I'm sure it's probably ok, but knowing for sure is certainly a good thing.

 

[hugin7]

Alienware finally responded with the info below:

Power Supply, 850 Watts, 850 Watts, ATX, 240 Volts, Mini Tower, Active Power Factor Correction, Delta, E-Star, RoHS Compliant, (Bronze).

Please let me know if that helps or if I should fish for other specifics.

Cool. Let me know what they tell you. Very unusual for this information not to be on the unit, so I'd actually have wanted to do that myself if it was my unit. In the meantime, I'm sure it's probably ok, but knowing for sure is certainly a good thing.

 

[Darkbreeze]

Well, being a Delta built unit means it's PROBABLY halfway decent, but since it's likely built to Dell's specifications, one of the requirements is likely that it be inexpensive so they can make profits off of it. Too bad it wasn't the Corsair RM850i. That would have been an excellent unit. Still, this likely means it is at least not a fire starter.

I would ask them to specifically verify whether or not this unit has a group regulated design or is Haswell certified. Getting them to give you an exact model number would be the best resolution of all though. The rest, APC, E-star, RoHS are all fairly worthless.