A few months ago Cryorig was kind enough to ship me a review sample of the then upcoming C7Cu CPU heatsink/fan. I've had it installed in a few of my prototype S401 cases and finally got a chance in the last few weeks to really collect some data and figure out what role the C7Cu might have to play for the SFF community.
Obviously, the desired outcome would be much better cooling & heat dissipation than the original C7, even though the C7 is generally thought to be an already great performer in the SFF space. When the official announcement of the C7Cu was made, many were shocked to hear that it would carry a premium price of $50 compared to it's sibling the C7 original coming in at $30. So the real question is: What additional performance is there to be had for the extra $20?
I intended to find out, but quickly discovered that hardware reviews (like many other things) are always harder than they seem.
Although I have a lot of AMD hardware to test with, the test system had the following specifications:
ASRock B350 AM4 mITX mainboard
AMD AM4 Ryzen 1800X
16GB G.Skill DDR4 3200 RGB RAM
Toshiba 256GB m.2 SSD
AMD Vega 64 LE 8GB
FSP Dagger 500W SFX PSU
Salvo Studios S401 Revision 4 prototype case
I began by installing Windows 10 64-bit and applying all patches. I also ensured my BIOS was updated to the latest revision. I then looked at a few other recent reviews and decided to use HWInfo to collect data and Prime95 to stress-test the CPU. All tests were run for a minimum of 30 minutes and then I used the following HWInfo parameters to generate my graphs. Please correct me if I've made any glaring mistakes.
1. Tdie - Measured die temperature as recorded by on-board thermal diodes. From my research, this is the best value to use when attempting to measure an accurate value for the CPU core temp for Ryzen processors.
2. CPU1 Frequency - This is the measured frequency that CPU1 (in HWInfo this is actually "CPU0" but I think a non-zero based number is easier to read for the purposes of a review).
3. CPU Fan RPM - This is the measured RPM of the PWM (Pulse Width Modulation) controlled fan on top of the CPU itself. This will be important when I cover the additional data points I gathered when I modified the C7Cu to suppor the Noctua NF-A9 fan. I also took a Wraith Prism and used it to generate some data for comparison.
Below is a slideshow of the results I measured for Tdie in the various configurations. When you read "open air" this means the vented cover on the S401 was removed, "closed case" means with the vented cover installed. All tests were run using 16 worker threads in Prime95 with the exception of a few that were executed using 4 or 2 threads for the purpose of testing whether the overall TDP would make a difference in clock speed or heat dissipation of the C7Cu.
So for Prime95 at maximum load, the C7Cu appears no different than the C7 original for maximum temperature. Interestingly, the NF-A9 swap also made no measurable difference, other than the additional 5-10mm of clearance it requires (this combination will NOT fit inside the S401 with the vented panel installed). It appears there is headroom to be made however, as the Wraith Prism far outperforms the C7, C7Cu, or C7Cu w/ NF-A9.
Next up are the same test results but looking at the frequency of CPU1 (monitor of CPU0).
Here is where it gets interesting because no longer can we simply look at the temperature of the CPU die to determine how effective a HSF combination is. A few things to note here, the stock C7 settles in after several minutes at around 3500Mhz (mean) while the C7Cu has a mean value of around 3550Mhz. This is fairly insignificant, but I think it is enough empirical data to conclude that the C7Cu can dissipate heat at least a fraction better than the C7. The other thing to note here is again the NF-A9 doesn't seem to make a large difference when all threads are consumed by Prime95.
Lastly are the CPU Fan RPM numbers. This should help us get a gauge for how hard the fan is having to work to keep the CPU at the peak 74.8 degrees C as measured from Tdie.
Here we see what is expected. Both the C7 original and C7Cu fans are the same (other than color) so they perform nearly identical. The C7 original has a bit of a strange start on the closed case measurement but settles in around the 2375RPM mark after several minutes. The NF-A9 stays below 2000RPM which also implies it will run quieter. That said, I'm unsure how any of these fans operate from a noise perspective as my FSP PSU must have a bad thermal diode as the 80mm fan ran at full speed the entire time I ran these tests. With no fan speed selector switch on the unit, I'm going to have to RMA it back to FSP and see what they find. Regardless, I don't have a noise meter so I couldn't perform those tests even if the PSU wasn't overwhelmingly noisy.
Around Q1 2018, the C7Cu was rumored to be a fantastic upgrade over the already well-performing C7. Unfortunately, these rumors really hurt the C7Cu as expectations were unrealistic and drastically different from what Cryorig eventually released. Perhaps more testing reveal a scenario where the C7Cu drastically outperforms the C7, but I've found no obvious empirical evidence to suggest otherwise. Even so, it would be very niche in application and when factoring in the additional 66% cost of the C7Cu vs. the C7, it makes it really hard to justify purchasing. There are vain reasons to purchase it of course as copper is a cool color and I'm sure some will buy it simply to match their color-themed rigs. For that the C7Cu is perfect for an SFF build awaiting the right color accessory. However, I can't recommend the C7Cu over the C7 original even if some of the claims are true regarding potential temperature decreases of up to 4 degrees C. I think it's safe to say that for SFF users the purchase decision generally follows this order: size, frequency/performance, noise, cost (from most to least important), aesthetics. In all aspects, the C7 original equals the C7Cu except for the cost where the C7 bests the C7Cu and aesthetics (in the eye of the beholder).
Includes AM4 bracket
Equal performance to C7 original
$20 more than C7 ($15 more if you need the AM4 adapter)
For these reasons, I have to say stick with the C7 original as it still performs great and will deliver better value than the C7Cu.
Speculation: There are wildly varying results being reported of the C7 Cu's performance by users. I believe that this could be explained several ways. One could be a variance in manufacturing process as this is a new product for Cryorig so perhaps they have improved the manufacturing process to weed out what could have been some lesser quality units that made it into the retail channel. Another could be that the C7 Cu simply performs differently on different CPUs. I will be purchasing another C7 Cu through a retail channel to try and see if there are any differences between my review unit and the retail one I purchased. I will attempt to write a new blog once I have time to test a newly-purchased C7 Cu.
Buy the Cryorig C7 or C7Cu here!
Valley Precision has finished fabricating the 4th revision of the S401 and it should be in my hands tomorrow. I know I have not been updating this page as often as I'd like but rest assured that I'm working on both the S401 and the S700 in every spare minute of my time (which isn't much). A 4 year old and a newborn consume both my wife's and my time. I'm very close to complete with the S401, this prototype will be the first prototype to include all the parts for a completed case - so I'm also very excited. I'll try to post some pictures in the next few days of the S401.
As far as the rest of the supplies are concerned, I've purchased and received nearly everything I'll need to assemble and ship the cases once they are fabricated. I will be manufacturing 125 S401 units in the first batch. One hundred of these will be powder coated a fine texture black while the remaining 25 will be left unpainted. These I will offer to custom powder coat for a premium price.
I'm nearing completion of the 1st revision of the S700. I've managed to make the flex slot quite dynamic for such a small case. I'm still ironing out some clearance issues when water blocks are being used on the GPUs, but I should have this revision completed by Monday July 9th. Take a look at the video below to see some of the options for the flex slot.