Massive winter storms providing perfect excuse to stay indoors and tinker? Check.
Internet access to upload pictures? DOH.
So, the good word is that we will have broadband access restored tomorrow. That is, if the service guy doesn't bail due to the two feet of snow we've gotten since Friday. Hopefully not.
In the mean time, I have been busy exhausting the possible tests to run on the system with stock air cooling. Most importantly, I wanted to answer the obvious question: Just how fast will a retail 920 run on the stock air cooler?
The line is made somewhat blurry due to Intel's Turbo Boost, and the way it works. You see, Turbo Boost increases the CPU multiplier from 20.0x to 20.5x or 21.0x if the CPU temperature is not rising too quickly. The sweet spot for this feature seems to be 70C in-core temps and/or roughly 150W TDP when using the stock cooler.
So this is where it gets interesting. With Turbo Boost enabled, my 920 quite competently hits 3.4Ghz at a measily 1.29VCore, only 0.03V above stock max voltage. The absolute peak VCore recorded was 1.3V. At this VCore, the processor's dissipated power @ 3.4Ghz is roughly 153W. Yeah, just over the sweet spot. What this means is that after about 10 to 15 seconds of running at 153W with the 21.0x multiplier, the processor drops its TDP to 130W and the multiplier drops to 20.5x. After a few minutes of OCCT @ max load, the multiplier drops to 20.0x briefly before returning to 20.5x. It never returns to 21.0x.
Now, the system simply will not post at lower VCores with the BCLK so high. The VCore requirements for full load look like this:
Stock 1.26V: 150-155BCLK
+0.02V = 1.28V: 156-160BCLK
+0.03V = 1.29V: 161-??? (This loses relevancy as the Turbo multiplier is lost too quickly at higher BCLKs, will explain)
So as you can see, there is a balancing act between BCLK and VCore in order to achieve a maximal overclock with the stock cooler without completely disabling Turbo Mode. The increased multiplier is essential to getting a peak overclock -- you don't want your TDP, and thus your core temperatures, to be so high that Turbo Mode is disabled.
The other factor is how often the processor will be at maximum load. The Core i7 is EXTREMELY good at partitioning power usage, so it is quite possible to run a high BCLK with the stock cooler as long as your usage patterns are normal. Even gaming presents no problems, because the times spent under 100% load are very short. In other words, you will never be at maximum load long enough to disable full 21.0x Turbo Mode.
Unfortunately, applications like FAH negate the Core i7's masterful power management due to the 100% constant load, so this must be taken into consideration when overclocking and it is a good justification for absurd cooling systems. (My watercooling plans seem rather justified!) Better cooling will allow Turbo Mode to work at much higher TDP's.
The thing to take away from this wall of text is this: It is definitely possible to actually have a slower system running at a higher BCLK because the system's TDP will be such that Turbo Mode is continually disabled.
Anyway, here's my current overclock running on the stock air cooler, which handily trumps all the reference benchmarks from the 3.2Ghz i7-965... I had just started OCCT here, so the core temperatures are reading low. They top out around 70C under this overclock.
Absolute bottom line:
You can buy a Core i7 920 today and overclock it to be faster than an i7 965 with the stock air cooler.
Oh, and it's just as quiet. The stock air cooler doesn't even know how to make noise.
Next stop: 1300Watts of dissipation and some Swiftech Sweetness.
Woa.. 70c That would freak me out but sounds like it's normal with these chips to hit that high and be stable. I bet you can't wait to really open her up under water.
Woa.. 70c That would freak me out but sounds like it's normal with these chips to hit that high and be stable. I bet you can't wait to really open her up under water.
Well, it's worth noting that the 70C reading is the actual in-core temperature. The motherboard's CPU temperature reading is "TMPIN0" which reads 41C in my screenshot and the maximum recorded was 59C even after the CPU had been at 100% load for hours. This is the temperature that most accurately compares to historical CPU temperature readings before the days of in-core temperature sensors.
The in-core temperatures are generally considered safe if they're under 80C to 90C when it comes to processing units like traditional CPU's and GPU's.
It all boils down to performance and stability. 70C in-core seems to be perfectly stable and does not present any problems with performance, so I'm callin' it good!
Is it higher than I'd like? Sure, but it's stock air...
So I've started work on the external 360mm radiator housing.
<hr> The best laid plans...
I want to make sure I get it all right, so before I cut anything I whipped up some snazzy CAD models and cut lists.
<hr> COMMENCE DREMELITION!
First order of business: Corner bracing. I determined that I needed to make my enclosure 40cm tall (interior dimension) in order to house the overall 390mm height of the radiator and provide 5mm of clearance top and bottom which will be gapped with neoprene to provide an airtight seal.
<hr>
I'm constructing the sides of the housing out of 22ga weld steel. This is fairly heavy stuff. My steel snips did not like 22ga weld steel. I was hoping to save time. Instead I had to straighten and grind my first cut. The end portion of the cut was done using the Dremel. Quite a dramatic difference in cut quality.
<hr>
These magic discs are just fricken' awesome. Komete -- I really am not that great with the Dremel. I owe all my clean cuts to these discs. They're brilliant.
<hr>
Despite being awesome, this thick steel did cause me to burn through four of them. Still, only four discs for over 2 meters of total cuts through steel and aluminum with NO breaks and butter-smooth performance. Amazing.
<hr>
The second side plate is about to come off here. The mini-clamps and angle bar brace keep the sheet steady when there isn't enough metal left to hold it all together.
This is also the cleanest shot of the room. Trust me.
<hr>
The corner braces clamped to one side sheet and shot from the back. The 22ga steel plating fits flush into the inside corners of the angle bar lengths.
These joints will be cold welded with my buddy JAY BEE! Of course, since I am 100% sure exactly how this is all going together, I haven't welded anything yet. Cause you know, I'm 100% sure.
I haven't seen those cutting discs before. That locking mechanism is different than my deimel but I imagine with a different bit I could use them with mine.
Will you be using two pumps? One for the internal rad and one for the external?
I haven't seen those cutting discs before. That locking mechanism is different than my deimel but I imagine with a different bit I could use them with mine.
There is a special EZ-Lock bit which is sold as part of a kit that includes 5 cutting discs for $14. Then the discs can be purchased in 5-packs for around $5. As someone who has spent a lot of time cutting with the old-style discs, I say it's $20 well spent. The EZ-Lock Metal discs work wonders and the EZ-Lock bit makes swapping discs in and out really fast.
Will you be using two pumps? One for the internal rad and one for the external?
Lookin good bro., keep it going
It's all one loop right now with one pump. There is some question about whether the single Laing D5-B pump is going to be enough flow, but I think it will be.
The loop looks like this... PUMP -> 240mm RAD -> HD4870 -> RES -> 360mm RAD -> CPU
The 360mm rad is going to have 20F air blowing through it... I think it should be sufficient.
I like the EZ-Lock discs. Garg and I tried using the old style ones in various projects and were always frustrated with the constant need to take everything apart when changing discs. I'm sure the new ones make it much easier.
Dude, not only is it 2000% faster to swap discs in and out, but the discs lock into a floating, spring-loaded tip so as you hit snags in the material or tilt the Dremel, the thing automatically adjusts its angle of attack ever so slightly. The sides of it are also a grinding surface, so your cuts are self-cleaning. The net effect is cleaner, straighter cuts and discs that never break apart (!!!). Highly recommend to any of my fellow dremolishers out there.
The loop looks like this... PUMP -> 240mm RAD -> HD4870 -> RES -> 360mm RAD -> CPU
The 360mm rad is going to have 20F air blowing through it... I think it should be sufficient.
Thank you.
Woa, that's some cool air. Where will you be getting the air from? Also if the pump is underpowered and the flow is slow getting to the cpu, will there be any risk of condensation? If the water is in the 360mm rad long enough then there could be a risk of the water hitting the inside of the pc case below ambient temps. Then again, I know that's a problem we would all like to have.
My last water cooling adventure was a disaster, so I'm no pro. I was shooting for the famed water cooling for under $100.00 challenge a few years back. Somehow I had the bright idea I could build my own radiator. It was pretty cool for about 20 seconds until the leaks got too bad. Then I tried the bucket of water and a submergeble pump. Worked great until my first born started crawling. The things we do to overclock.
Woa, that's some cool air. Where will you be getting the air from?
Living inside the Arctic Circle is not all bad.
Seriously though, I live in northern NH and we do not heat the room that abuts the room the computer is in... So my plan is to build a baffle for the door to the cold room and install some plastic dryer duct so that the fans for the 360mm rad are sucking the frigid air from that room and exhausting it back into that room. The computer sits right in front of the door, so it all works out.
Also if the pump is underpowered and the flow is slow getting to the cpu, will there be any risk of condensation? If the water is in the 360mm rad long enough then there could be a risk of the water hitting the inside of the pc case below ambient temps. Then again, I know that's a problem we would all like to have.
The thought has crossed my mind. I don't think that the liquid will drop far enough to cause condensation, but I have some contingency plans for that possibility. The air is extremely dry up here around this time of year. Roughly 30% humidity and the ambient air in the computer room is around 60F.
My last water cooling adventure was a disaster, so I'm no pro. I was shooting for the famed water cooling for under $100.00 challenge a few years back. Somehow I had the bright idea I could build my own radiator. It was pretty cool for about 20 seconds until the leaks got too bad. Then I tried the bucket of water and a submergeble pump. Worked great until my first born started crawling. The things we do to overclock.
The disasters are half the fun, right? Props for building your own radiator, though. That is quite the ambitious undertaking.
This is the first time I've ever replied to this thread from the new Core i7 rig!
It won't be the last.
The system is now complete with the exception of the enclosure and fan arrays for the 360mm radiator. I am running it right now at stock speeds with only the 240mm radiator actively cooled.
I'm in the process of sorting through all of the 711 pictures that I have taken so far of this build and uploading the non-junk ones to Flickr. There are tons of pics of the MSI X58 Platinum, some random fan mods that I did for intake tunnels, the Sapphire HD4870 getting torn to pieces to install teh blockz, etc. etc.
I am definitely satisfied with the current results. CPU is idling at 18C and making even less noise than it was before. The GPU temperature is reading 30C, down from around 50C with the stock cooler. The CPU idled around 30C with the stock cooler.
I am hesitant to do any load tests as it is only running with the one smaller radiator right now, although I have a feeling that at stock speeds it is more than sufficient. The 360mm is actually in its cardboard box to protect it at the moment. I cut holes for the hoses, but I don't want any fins getting bent or paint getting chipped so I am leaving it in the box. So it is not contributing at all to the cooling of the system. Right now it's just the 240mm guy handling the 920 and the 4870.
<hr>
MORE:
Added pics from the Sapphire HD4870 1GB waterblock install and created a Flickr Collection which now organizes all the sets from the build. Check out the collection for all the latest pics.
All around very clean. With low temperatures like that I'm sure you could do some overclocking and be fine. But I can understand your hesitation. I looked at the video card water block. Is the memory for the video card below that block too? It was hard to make out the memory modules.
Very nice. Stayed clean and some freaking sick hardware, more vid card!?!.....get it more cow bell......more vid card. Especially like the time spent on the mods and how professional things look. I am a sucker for green and bling. The only thing that could of made it better would be green liquid and you built a ROM to spit out ninja turtle action figures.
All around very clean. With low temperatures like that I'm sure you could do some overclocking and be fine. But I can understand your hesitation. I looked at the video card water block. Is the memory for the video card below that block too? It was hard to make out the memory modules.
Probably could at least run the same overclocks that I ran with the stock air cooler, but I'm not worried about it. Next weekend my order from McMaster-Carr will get here and I'll have the mats I need to finish my 360mm enclosure and the baffle for the door, and then I will REALLY be in business.
The video card memory is under the block. These pictures from the Wasserkuhlen ein Sapphire HD4870 1GB Flickr Set should illustrate it nicely...
Very nice. Stayed clean and some freaking sick hardware, more vid card!?!.....get it more cow bell......more vid card. Especially like the time spent on the mods and how professional things look. I am a sucker for green and bling. The only thing that could of made it better would be green liquid and you built a ROM to spit out ninja turtle action figures.
Thanks for the props man. ^^ Always good when my fellow ghetto modders dig this shit.
I like the mix of green and blue myself, although I wish that the green and blue LED's on the motherboard were swapped. Then it would almost have a yin-yang look to it with a dot of each color inside the other.
Overall though, I am happy with the colors; especially the fact that the motherboard came stock in dark gray / black PCB with blue and green LED's. As Thrax would say, I would have bought it if it was magenta with dick-shaped PCI slots because I think it's the best X58 motherboard for the money, but it's cool that MSI made it look fscking pimp!
Great work, man. Nice documentation as well. Bring on the overclocking!
Thanks! Overclockage is forthcoming.
I was going to wait to get the second radiator online, but that's an exercise in futility. My trigger finger is itchy. I'm going to see what I can do with just the 240mm radiator.
I will be working on a CPU-only overclock at the moment. I think if I overclock the 4870 and try to run 3D benchmarks I will definitely overload the radiator. After all, it only has 2x120mm fans on it.
The 360mm radiator will have 6x120mm fans on it and cold air, so that is going to be the key to super-high i7 clocks. I am aiming for 4Ghz. Anandtech did it!
Update:
I had my ~3.4Ghz overclock settings (162 x 21) saved in the BIOS, so I just loaded those up to see how they'd perform.
That was my max overclock on the stock air cooler. At that point, it was cutting down the multiplier to 20.5x and 20.0x in order to reduce the temperature, because the in-core temperatures were hitting 70C.
With just the 240mm radiator going, my cores are sitting at 50C at the same 162 x 21 overclock, a drop of over 20C without the big, external radiator in the loop.
I imagine the temps will equalize a bit higher after running 100% OCCT load for a while, but it's still impressive. With the air cooler, my temps shot to over 70C within 15-20 seconds. Here I'm been running it for almost 6 minutes with 8 threads churning away and then temps just cracked 54C.
Note these are the IN CORE temperatures we're talking about. The motherboard CPU reading -- the traditional "CPU temp" we usually see -- is reading 41C.
In core temperatures are always much, much higher. For example, even when the motherboard reports that the CPU is idling at 18C, the in core temperature is ~35C. That's because the in core temps are basically measured RIGHT AT the heat source. The processor will ALWAYS generate a given amount of heat in the core itself... It's about how well you can dissipate that heat.
Comments
New hardware? Check.
Christmas break? Check.
Massive winter storms providing perfect excuse to stay indoors and tinker? Check.
Internet access to upload pictures? DOH.
So, the good word is that we will have broadband access restored tomorrow. That is, if the service guy doesn't bail due to the two feet of snow we've gotten since Friday. Hopefully not.
In the mean time, I have been busy exhausting the possible tests to run on the system with stock air cooling. Most importantly, I wanted to answer the obvious question: Just how fast will a retail 920 run on the stock air cooler?
The line is made somewhat blurry due to Intel's Turbo Boost, and the way it works. You see, Turbo Boost increases the CPU multiplier from 20.0x to 20.5x or 21.0x if the CPU temperature is not rising too quickly. The sweet spot for this feature seems to be 70C in-core temps and/or roughly 150W TDP when using the stock cooler.
So this is where it gets interesting. With Turbo Boost enabled, my 920 quite competently hits 3.4Ghz at a measily 1.29VCore, only 0.03V above stock max voltage. The absolute peak VCore recorded was 1.3V. At this VCore, the processor's dissipated power @ 3.4Ghz is roughly 153W. Yeah, just over the sweet spot. What this means is that after about 10 to 15 seconds of running at 153W with the 21.0x multiplier, the processor drops its TDP to 130W and the multiplier drops to 20.5x. After a few minutes of OCCT @ max load, the multiplier drops to 20.0x briefly before returning to 20.5x. It never returns to 21.0x.
Now, the system simply will not post at lower VCores with the BCLK so high. The VCore requirements for full load look like this:
Stock 1.26V: 150-155BCLK
+0.02V = 1.28V: 156-160BCLK
+0.03V = 1.29V: 161-??? (This loses relevancy as the Turbo multiplier is lost too quickly at higher BCLKs, will explain)
So as you can see, there is a balancing act between BCLK and VCore in order to achieve a maximal overclock with the stock cooler without completely disabling Turbo Mode. The increased multiplier is essential to getting a peak overclock -- you don't want your TDP, and thus your core temperatures, to be so high that Turbo Mode is disabled.
The other factor is how often the processor will be at maximum load. The Core i7 is EXTREMELY good at partitioning power usage, so it is quite possible to run a high BCLK with the stock cooler as long as your usage patterns are normal. Even gaming presents no problems, because the times spent under 100% load are very short. In other words, you will never be at maximum load long enough to disable full 21.0x Turbo Mode.
Unfortunately, applications like FAH negate the Core i7's masterful power management due to the 100% constant load, so this must be taken into consideration when overclocking and it is a good justification for absurd cooling systems. (My watercooling plans seem rather justified!) Better cooling will allow Turbo Mode to work at much higher TDP's.
The thing to take away from this wall of text is this: It is definitely possible to actually have a slower system running at a higher BCLK because the system's TDP will be such that Turbo Mode is continually disabled.
Anyway, here's my current overclock running on the stock air cooler, which handily trumps all the reference benchmarks from the 3.2Ghz i7-965... I had just started OCCT here, so the core temperatures are reading low. They top out around 70C under this overclock.
Absolute bottom line:
You can buy a Core i7 920 today and overclock it to be faster than an i7 965 with the stock air cooler.
Oh, and it's just as quiet. The stock air cooler doesn't even know how to make noise.
Next stop: 1300Watts of dissipation and some Swiftech Sweetness.
The in-core temperatures are generally considered safe if they're under 80C to 90C when it comes to processing units like traditional CPU's and GPU's.
It all boils down to performance and stability. 70C in-core seems to be perfectly stable and does not present any problems with performance, so I'm callin' it good!
Is it higher than I'd like? Sure, but it's stock air...
Flickr updated with a few random shots from the 600 (or so) pictures I have of this build thus far.
The pinned CPU socket I was talking about... Cool design...
A short series of pictures from the mounting of the top 240mm radiator and the painting of the frame...
And here's a shot of what I'm currently working with as of an hour ago...
So I've started work on the external 360mm radiator housing.
<hr>
The best laid plans...
I want to make sure I get it all right, so before I cut anything I whipped up some snazzy CAD models and cut lists.
<hr>
COMMENCE DREMELITION!
First order of business: Corner bracing. I determined that I needed to make my enclosure 40cm tall (interior dimension) in order to house the overall 390mm height of the radiator and provide 5mm of clearance top and bottom which will be gapped with neoprene to provide an airtight seal.
<hr>
I'm constructing the sides of the housing out of 22ga weld steel. This is fairly heavy stuff. My steel snips did not like 22ga weld steel. I was hoping to save time. Instead I had to straighten and grind my first cut. The end portion of the cut was done using the Dremel. Quite a dramatic difference in cut quality.
<hr>
These magic discs are just fricken' awesome. Komete -- I really am not that great with the Dremel. I owe all my clean cuts to these discs. They're brilliant.
<hr>
Despite being awesome, this thick steel did cause me to burn through four of them. Still, only four discs for over 2 meters of total cuts through steel and aluminum with NO breaks and butter-smooth performance. Amazing.
<hr>
The second side plate is about to come off here. The mini-clamps and angle bar brace keep the sheet steady when there isn't enough metal left to hold it all together.
This is also the cleanest shot of the room. Trust me.
<hr>
The corner braces clamped to one side sheet and shot from the back. The 22ga steel plating fits flush into the inside corners of the angle bar lengths.
These joints will be cold welded with my buddy JAY BEE! Of course, since I am 100% sure exactly how this is all going together, I haven't welded anything yet. Cause you know, I'm 100% sure.
<hr>
Money.
<hr>
More to come.
Top Mounted 240mm Radiator
Now with bracket action!
External 360mm Radiator Housing
See me score... With myself... Twice. Also, steel snips suck.
Will you be using two pumps? One for the internal rad and one for the external?
Lookin good bro., keep it going
It's all one loop right now with one pump. There is some question about whether the single Laing D5-B pump is going to be enough flow, but I think it will be.
The loop looks like this... PUMP -> 240mm RAD -> HD4870 -> RES -> 360mm RAD -> CPU
The 360mm rad is going to have 20F air blowing through it... I think it should be sufficient.
Thanks for keeping up with my little project man.
Thank you.
Woa, that's some cool air. Where will you be getting the air from? Also if the pump is underpowered and the flow is slow getting to the cpu, will there be any risk of condensation? If the water is in the 360mm rad long enough then there could be a risk of the water hitting the inside of the pc case below ambient temps. Then again, I know that's a problem we would all like to have.
My last water cooling adventure was a disaster, so I'm no pro. I was shooting for the famed water cooling for under $100.00 challenge a few years back. Somehow I had the bright idea I could build my own radiator. It was pretty cool for about 20 seconds until the leaks got too bad. Then I tried the bucket of water and a submergeble pump. Worked great until my first born started crawling. The things we do to overclock.
Living inside the Arctic Circle is not all bad.
Seriously though, I live in northern NH and we do not heat the room that abuts the room the computer is in... So my plan is to build a baffle for the door to the cold room and install some plastic dryer duct so that the fans for the 360mm rad are sucking the frigid air from that room and exhausting it back into that room. The computer sits right in front of the door, so it all works out.
The thought has crossed my mind. I don't think that the liquid will drop far enough to cause condensation, but I have some contingency plans for that possibility. The air is extremely dry up here around this time of year. Roughly 30% humidity and the ambient air in the computer room is around 60F.
The disasters are half the fun, right?
Thanks!
I posted 12 new pictures to the Flickr account in a new Set...
Wasserkuhlen ein Core i7
Now with more RES-250!
Check out the Flickr Set for more!
This is the first time I've ever replied to this thread from the new Core i7 rig!
It won't be the last.
The system is now complete with the exception of the enclosure and fan arrays for the 360mm radiator. I am running it right now at stock speeds with only the 240mm radiator actively cooled.
I posted about 20 new pictures to the Wasserkuhlen ein Core i7 Flickr Set. Check it out! Let me know what you think!
I'm in the process of sorting through all of the 711 pictures that I have taken so far of this build and uploading the non-junk ones to Flickr. There are tons of pics of the MSI X58 Platinum, some random fan mods that I did for intake tunnels, the Sapphire HD4870 getting torn to pieces to install teh blockz, etc. etc.
I am definitely satisfied with the current results. CPU is idling at 18C and making even less noise than it was before. The GPU temperature is reading 30C, down from around 50C with the stock cooler. The CPU idled around 30C with the stock cooler.
I am hesitant to do any load tests as it is only running with the one smaller radiator right now, although I have a feeling that at stock speeds it is more than sufficient. The 360mm is actually in its cardboard box to protect it at the moment. I cut holes for the hoses, but I don't want any fins getting bent or paint getting chipped so I am leaving it in the box. So it is not contributing at all to the cooling of the system. Right now it's just the 240mm guy handling the 920 and the 4870.
<hr>
MORE:
Added pics from the Sapphire HD4870 1GB waterblock install and created a Flickr Collection which now organizes all the sets from the build. Check out the collection for all the latest pics.
The video card memory is under the block. These pictures from the Wasserkuhlen ein Sapphire HD4870 1GB Flickr Set should illustrate it nicely...
Thanks for the props man. ^^ Always good when my fellow ghetto modders dig this shit.
I like the mix of green and blue myself, although I wish that the green and blue LED's on the motherboard were swapped. Then it would almost have a yin-yang look to it with a dot of each color inside the other.
Overall though, I am happy with the colors; especially the fact that the motherboard came stock in dark gray / black PCB with blue and green LED's. As Thrax would say, I would have bought it if it was magenta with dick-shaped PCI slots because I think it's the best X58 motherboard for the money, but it's cool that MSI made it look fscking pimp!
I was going to wait to get the second radiator online, but that's an exercise in futility. My trigger finger is itchy. I'm going to see what I can do with just the 240mm radiator.
I will be working on a CPU-only overclock at the moment. I think if I overclock the 4870 and try to run 3D benchmarks I will definitely overload the radiator. After all, it only has 2x120mm fans on it.
The 360mm radiator will have 6x120mm fans on it and cold air, so that is going to be the key to super-high i7 clocks. I am aiming for 4Ghz. Anandtech did it!
Update:
I had my ~3.4Ghz overclock settings (162 x 21) saved in the BIOS, so I just loaded those up to see how they'd perform.
That was my max overclock on the stock air cooler. At that point, it was cutting down the multiplier to 20.5x and 20.0x in order to reduce the temperature, because the in-core temperatures were hitting 70C.
With just the 240mm radiator going, my cores are sitting at 50C at the same 162 x 21 overclock, a drop of over 20C without the big, external radiator in the loop.
I imagine the temps will equalize a bit higher after running 100% OCCT load for a while, but it's still impressive. With the air cooler, my temps shot to over 70C within 15-20 seconds. Here I'm been running it for almost 6 minutes with 8 threads churning away and then temps just cracked 54C.
Note these are the IN CORE temperatures we're talking about. The motherboard CPU reading -- the traditional "CPU temp" we usually see -- is reading 41C.
In core temperatures are always much, much higher. For example, even when the motherboard reports that the CPU is idling at 18C, the in core temperature is ~35C. That's because the in core temps are basically measured RIGHT AT the heat source. The processor will ALWAYS generate a given amount of heat in the core itself... It's about how well you can dissipate that heat.