Potato upgrade, Part 2 (CPU)

David Mann · Apr 13, 2020

After asking about how many CPU cores I need here and posting the results of my GPU upgrade here, I've finally upgraded my CPU. The upgrade is from a dual-core i3-4130 to a quad-core i7-4790 in an old Dell Alienware X51 R2. The computer is now officially upgraded from Potato status to Sweet Potato. Here are the before and after system summaries:

Before getting into the test results, I'd like to mention a few words about fans. First, this computer has trouble keeping the new CPU cool. I'll get into this in another thread (here), but I had to use SpeedFan to force the system fans to run fast enough to keep the CPU below its maximum temperature of 73 °C. During testing, the three system fans (CPU fan, a motherboard fan, and an intake fan) were all running at 100%. I also used MSI Afterburner to keep the GPU below it's 80 °C limit (GPU fan curve below). The computer gets pretty loud when running a game.

I tested frame rates, temperatures, and power usage before and after the upgrade with Dirt Rally 2.0, ACC, and Rise of the Tomb Raider. The CPU upgrade did not affect the frame rate results with DR2.0 and Rise of the Tomb Raider, so I won't go into detail about that here. I will mention a couple things about DR2.0 that I observed. First, temporal anti-aliasing (TAA) in DR2.0 is pretty good, but the normal spatial anti-aliasing methods don't seem to do much, so I'm running only TAA now. Also, there is a big frame rate jump between Medium and High Reflections. By setting Reflections to Medium, I can run pretty much everything else on High and get about 60 FPS.

The ACC test results tell an interesting story. As expected, the CPU upgrade provided a big boost. I tested by driving a formation lap and one race lap around the Nurburgring at night in stormy weather, with 15 AI opponents. This was NOT a replay test, but actual driving to test real conditions. Graphics were set to Low, Medium, and High presets. Screen resolution is 2560x1080. This table contains the old CPU results from the previous post, as well as the new results in the last three columns. I measured the FPS with Fraps, the power usage at the wall with a Kill-A-Watt meter, the temperatures with CPUID HWMonitor, and the CPU/GPU usage and GPU memory with Task Manager. Note that 50% CPU usage is fully utilizing all the physical cores (both CPUs have Hyperthreading).

	i3-4130 / GTX 645 (Low)	i3-4130 / GTX 970 (Low)	i3-4130 / GTX 970 (Medium)	i3-4130 / GTX 970 (High)	i7-4790 / GTX 970 (Low)	i7-4790 / GTX 970 (Medium)	i7-4790 / GTX 970 (High)
Minimum FPS	18	30	25	29	60	46	33
Mean FPS	26.1	46.8	39.6	35.7	80.2	66.3	47.4
Maximum FPS	40	58	70	47	100	92	66
Power	100 W	175 W	190 W	190 W	260 W *	260 W *	275 W
CPU Usage	90%	100%	100%	100%	65%	65%	65%
CPU Temperature	68 °C	71 °C	71 °C	71 °C	72 °C *	77 °C *	62 °C
GPU Usage	100%	30 %	50%	80%	80%	90%	100%
GPU Temperature	77 °C	80 °C **	80 °C **	80 °C **	77 °C	77 °C	75 °C
GPU Memory	1.0 GB (max)	3.6 GB	3.4 GB	3.0 GB	3.6 GB	3.6 GB	3.4 GB

* I think something went wrong with SpeedFan for these runs, so the CPU got hotter and power draw was slightly lower.
** The previous round of testing was done before I used MSI Afterburner to control the GPU fan, so the GPU was slightly throttling back at its temperature limit.

Looking at this now, I can see that the test conditions were not as well controlled as I'd like. First, between the previous test and this test, I started using MSI Afterburner to keep the GPU off its temperature limit. Second, I had some issues with SpeedFan that I think threw off some of the CPU temperature results in the new CPU test. Third, the room temperature was higher for the new CPU (69 °F / 21 °C for the old CPU and 76 °F / 24 °C for the new CPU), but the different fan controls make that a moot point.

Despite all that, there are clear trends in the results. The GPU and CPU upgrades both provided performance increases. With the new hardware, playing ACC is quite enjoyable, whereas it was marginal at best before. It looks like one or two more CPU cores would help, at least with this old generation of CPUs. ACC is definitely high on my list of games to play now due to the frame rate and graphical quality. I played around with the settings after doing these tests, and got 60-70 FPS depending on the track with great visual quality (settings below). In ACC, TAA produces lots of distracting artifacts, especially in the rain. FXAA doesn't do much but doesn't cost much either. I found that increasing the Resolution Scale above 100% does a better job of anti-aliasing, so I may play around with that more and keep it at 110% or 120%. I can't go higher due to FPS loss and GPU memory limitations, but I would suggest cranking up Resolution Scale as high as you can.

Thanks for reading. As I stated in the previous thread, I hope this will help some folks who can't afford the latest and greatest computer hardware. It's possible to get good results with decent second-hand hardware.

RasmusP · Apr 13, 2020

Awesome thread, enjoyed reading it all!
Nicely tested, written etc!

My only comment for now:
Where did you find 73°c for the CPU being maximum?
I agree that it would be nice to keep the temperature below that value but honestly I've never seen or heard about an Intel CPU dying at all since a decade.
Maybe you were finding the wrong temperature?
I forgot how they are called but there are different temperatures and you can only measure the higher one while the lower one is listed by Intel as maximum.

My 2 cents without guarantee:
All i3/5/7/9 CPUs have a temperature protection before they die. First they throttle, then they shut off.
There's a test from der8auer were he takes off the cooler of an i7 8700k while benchmarking it and it clocks down to 800 mhz and goes up to 99°.
It just won't shut down, they are really wondering about it in the video...
Fun fact: it drops to 97°c at some point and clocks itself back up, lol.

Anyway this listed maximum temperature I mentioned earlier is somewhere in the 70's for that i7 too but it's not the temperature we can measure!

So my thoughts: if your cpu temp, your measured one, isn't going into the 90's your cpu should be totally fine and won't die for at least 5 years. Maybe it can run 15 years at 90 degrees too but I won't bet on that hehe.

For your graphics card: the 970 can take a lot too. What are you using for the fan curve?
I would suggest you raise the "temperature target" or "limit" from 80 to 87 an it won't throttle anymore.
Heat isn't really a linear function so 2-3 degrees higher temperature need a lot less cooling power.

David Mann · Apr 13, 2020

RasmusP said:
Awesome thread, enjoyed reading it all!
Nicely tested, written etc!

My only comment for now:
Where did you find 73°c for the CPU being maximum?
I agree that it would be nice to keep the temperature below that value but honestly I've never seen or heard about an Intel CPU dying at all since a decade.
Maybe you were finding the wrong temperature?
I forgot how they are called but there are different temperatures and you can only measure the higher one while the lower one is listed by Intel as maximum.

My 2 cents without guarantee:
All i3/5/7/9 CPUs have a temperature protection before they die. First they throttle, then they shut off.
There's a test from der8auer were he takes off the cooler of an i7 8700k while benchmarking it and it clocks down to 800 mhz and goes up to 99°.
It just won't shut down, they are really wondering about it in the video...
Fun fact: it drops to 97°c at some point and clocks itself back up, lol.

Anyway this listed maximum temperature I mentioned earlier is somewhere in the 70's for that i7 too but it's not the temperature we can measure!

So my thoughts: if your cpu temp, your measured one, isn't going into the 90's your cpu should be totally fine and won't die for at least 5 years. Maybe it can run 15 years at 90 degrees too but I won't bet on that hehe.

For your graphics card: the 970 can take a lot too. What are you using for the fan curve?
I would suggest you raise the "temperature target" or "limit" from 80 to 87 an it won't throttle anymore.
Heat isn't really a linear function so 2-3 degrees higher temperature need a lot less cooling power.

Thanks Rasmus!

For the CPU, the temperature I found is on the Intel spec page for that CPU, T_CASE = 72.72 °C. But I guess that doesn't correspond to the core temperature. I did a bit of research based on your comments and came to the same conclusion. I will increase my CPU temperature to keep noise down and preserve the life of the CPU fan.

For the GPU, the fan curve is in the first post here. The fan hits 100% at about 78 °C, and GPU temperatures are generally in the high 70's for most games. I'd like to keep the limit at 80 °C, at least for now. With the fan properly running to keep the GPU cool, the GPU is usually sitting on the power limit. The power limiter in MSI Afterburner seems to work below 100%, but setting it greater than 100% doesn't actually let the GPU go over 100% (makes sense I guess, if it's a hardware limit of the power electronics on the GPU). Honestly, there seems to be very little performance difference when the GPU hits the temperature limit.

RasmusP · Apr 14, 2020

David Mann said:
For the CPU, the temperature I found is on the Intel spec page for that CPU, T_CASE = 72.72 °C. But I guess that doesn't correspond to the core temperature.

Yep, exactly. It's the same temperature for my old i7 2600k. Last week I tried Fortnite for the first time and somehow that game kills speedfan.
I use speedfan for controlling my case and cpu fans linked to the cpu temp.
I saw the temperature rising mid game but thought "screw it, I wanna play this round to the end, let's see if it shuts down or not".
I reached 94°C and wasn't really comfortable I can tell you... But all went fine! No throttling, no shut down, no stability issues!

maelstrom · Apr 15, 2020

It may just be my poor memory, but have the TCASE temps always been that low? For my 6700k it says 64C. I would have put money on it being around the 90 mark. Have they adjusted the temps on the Intel site down? I often reach 64 when gaming. According to Intel, TCASE is ”the maximum temperature allowed at the processor Integrated Heat Spreader (IHS)”. I’ve always understood that throttling to protect the CPU doesn’t start until about 100C.

RasmusP · Apr 15, 2020

maelstrom said:
It may just be my poor memory, but have the TCASE temps always been that low? For my 6700k it says 64C. I would have put money on it being around the 90 mark. Have they adjusted the temps on the Intel site down? I often reach 64 when gaming. According to Intel, TCASE is ”the maximum temperature allowed at the processor Integrated Heat Spreader (IHS)”. I’ve always understood that throttling to protect the CPU doesn’t start until about 100C.

Tcase isn't the Temperatur the sensors read. It's the temperature of the heatspreader after the hotter cores nicely spreaded the heat into all parts and all surfaces available and that is directly cooled by the cpu cooler.

I don't really get why Intel just doesn't state the temperature where it throttles and where the cpu will shut off.
Very confusing...

You can imagine it like having fever. Your body shouldn't go above 41°c but your skin will never have that temperature!
Intel for some reason decided that stating the maximum skin temperature would be the correct thing, lol.

John-Eric Saxén · Apr 15, 2020

I did some quick research and it looks like Intel changed their official spec sheet from Skylake where it still reported the Tcase value (typically around 73) to reporting Tjunction instead from Kaby Lake onwards (typically 100 degrees). These aren't exactly one-to-one measurements, but it's hardly going to be a 27 degrees difference. Given that most gaming laptops these days run with a CPU 95 degrees hot during boosting, I don't see any concern with 80-90 degree temperatures.