Fact or myth? Vertically mounted graphics cards perform less
21.3.2019
Translation: machine translated
"Vertical GPU Brackets - Why NOT To Use Them!" headlines HardwareCanucks in a YouTube video. The conclusion: Vertically mounted graphics cards run hotter and therefore clock a little slower. I want to find out for sure and put it to the test.
A vertically mounted GPU looks cool. It has long been a mystery to me why graphics card manufacturers design the fan side of the cards so elaborately, even though you can't see it at all during normal installation. Only the PCB is usually visible on the "back". Many manufacturers don't even bother to include a backplate.
The premise
Why should a vertically installed graphics card run slower? The reason given makes sense: Because it's closer to the side window, the fans can draw less air to cool the card. Above a certain temperature - this varies depending on the graphics card, but 80 degrees Celsius is a guideline for desktop cards - the graphics card throttles itself. This is known as thermal throttling.
I want to try this out for myself and get two brackets for vertical mounting. The first one I choose is the Fractal Flex VCR-25. The advantage of this bracket is that I can easily install it in my Fractal Define R6 case, as I have vertical outlets on the back and holes are already made at the bottom of the case for mounting.
The second one I use is the CableMod Verti PCI-e Bracket. This bracket makes it possible to mount vertical GPUs in cases that do not have vertical outlets on the back. The card is mounted with the bracket in front of the horizontal outputs, thus blocking them. The part is therefore not for people who need more than one PCI slot. The advantage of the cable mod bracket is that the GPU is mounted further back in the case and the fans can therefore draw air better. However, the bracket makes the card 3.1 centimetres longer.
As I realise when measuring, the Cable Mod bracket doesn't fit in my case because of my water cooling. I therefore decide to give it to my colleague Raphael Knecht for testing. Unfortunately, it doesn't fit him either. Philipp Rüegg is also unsuccessful, as his sound card blocks the space. Shit, what now? Right, we still have the streaming PC that I put together for Phil. Oh wait, it's a mini-ITX case. I definitely can't attach the bracket to it. Anyway, I'll just do the test with the Fractal Flex VCR-25. I already have an idea what I can do with the Cable Mod bracket. If you're considering the Cable Mod mount, you should be aware of the tight space constraints.
But now I want to mount the Fractal bracket.
Small dimensions
The Fractal Flex VCR-25 is only a PCI riser cable with a bracket for the bottom. Your case must already have vertical cut-outs on the back, otherwise you will not be able to mount the card.
Mounting should be very easy. With the emphasis on should, because in my case it's not quite so easy. In addition to the cable, two spacers and two screws are included. I already have two holes in my housing for the spacers. I fix these and mount the Fractal Flex VCR-25 on the spacers with the screws. Now I just have to plug in my GPU. It doesn't work. It just won't fit into the PCI slot.
Graphics cards should always fit easily into the slot. If I have to mess around, something is wrong. On closer inspection, I realise that I can't push the card far enough to the left, the back of the case is in the way. What do I do? I loosen the holder a little and try to pull it more to the right. Unfortunately, that doesn't help either.
I don't understand why Fractal Design doesn't make wider holes so that the bracket can be moved further. Not every graphics card is exactly the same and the metal end of mine is slightly wider than Fractal intended. Fortunately, I like to tinker. Armed with a locksmith's file, I set about enlarging the two mounting holes in the bracket. After five minutes, the hole is wide enough and I can finally plug in the GPU.
The card is at a slight angle to the housing. I have about three centimetres to the glass at the back and one centimetre at the front. Not exactly a lot of room for the GPU fans to draw air. I can well imagine that this has a negative effect on the cooling performance.
The test scenario
I am testing with the following hardware:
- Case: Fractal Design R6
- Graphics card: Gainward GeForce RTX 2070 Phoenix GS
- CPU: AMD Ryzen 5 2600
- Cooling: Two Corsair LL140 fans at the front, one Corsair LL140 fan at the bottom, one Corsair LL140 fan at the rear, two Thermaltake Riing 120 mm fans at the top in push configuration with Alphacool NexXxoS ST30 Full Copper 240mm. The CPU is cooled with an Alphacool Eisblock XPX
- Mount: Fractal Flex VCR-25
I use two programmes: GPU-Z and Heaven Benchmark. I run the test twice so that I have a comparative value. Once with the GPU mounted horizontally and once with it mounted vertically. First, I open GPU-Z and log the data. Then I start the Heaven benchmark and let it run in full screen for two hours. I select Ultra for the quality, set Tessellation to Extreme and Anti-aliasing to x8. I leave Stereo 3D and Multi-Monitor deactivated. GPU-Z records the temperature and clock frequency of the GPU in the background. Of course, I do the test with the side panel attached.
The comparison test
Now it's finally time to get down to business. I first do the test with the graphics card mounted horizontally. The front door of my case can be opened. This improves the air supply. I do the test for the first hour with the front door open and close it for the second hour.
After a short warm-up phase of five minutes, the temperature of the GPU stabilises at 62 degrees Celsius. This temperature remains more or less constant for the entire first hour. GPU-Z logs the temperature every second. There are very few outliers, and when there are, the temperature only falls or rises by one degree for one second. With these differences, outliers is a bit of an exaggeration. They are merely slight fluctuations.
After an hour, I close the front door. After two minutes, the temperature stabilises at 63 degrees Celsius. As before, the temperature remains constant with the exception of a few one-second fluctuations.
I didn't notice any decrease in performance either. My GPU achieves around 1900 MHz during the two hours.
Looks cool, but what about the performance?
Optically, a vertically mounted graphics card has a lot to offer. Firstly, you can finally see the more beautifully designed side of the card and secondly, you cover the lower half of the mainboard, which doesn't look very good. Unfortunately, vertical mounting means you lose a few PCI slots. The riser cable is not long enough to run over several cards. If you need several PCI slots, you will have to stick with the horizontal variant for better or worse.
Now it's getting exciting. I start the Heaven benchmark again, log the temperature with GPU-Z and leave the front door open. After five minutes, the temperature stabilises at 62 degrees Celsius. As with the test with horizontal mounting, I only have a few, one-second fluctuations of one degree during the first hour of testing.
After an hour, I close the front door. The temperature rises within a few seconds and stabilises at 64 degrees Celsius after two minutes. Even so, I have one-second fluctuations of one degree more or less. There is also no loss of performance: The GPU always clocks around 1900 MHz.
With the front door open, I don't notice any difference in temperature compared to the horizontally mounted graphics card. Even with the door closed, the temperature only rises one degree higher than in the first test. I don't notice any loss of performance with these small temperature differences.
What might be the reason for this? Thanks to the two 140 mm fans at the front and the 140 mm fan at the bottom, plenty of cool air gets into the case. The GPU fans are well supplied. The fact that my RTX 2070 is equipped with three fans certainly helps too. The fact that the front fans provide a decent article shows that when the front door is closed, the temperature of the graphics card is higher when mounted vertically than when mounted horizontally.
The air circulation in my case seems to be very good and the fact that my CPU is water-cooled certainly helps. The CPU doesn't run as hot and therefore produces less waste heat that could additionally heat up the GPU.
Conclusion
In my case, mounting the GPU vertically did not result in any loss of performance. However, this does not mean that this is always the case. This is shown by the HardwareCanucks test. Depending on how you cool your PC, you may experience a drop in performance.
However, if you ensure sufficient air circulation, nothing should stand in the way of a vertically mounted graphics card. And let's be honest: people who want to mount their GPU vertically value optimum cooling performance anyway. <p
Kevin Hofer
Senior Editor
kevin.hofer@digitecgalaxus.chFrom big data to big brother, Cyborgs to Sci-Fi. All aspects of technology and society fascinate me.
24 comments
later