- Joined
- Sep 24, 2008
- Messages
- 2,704 (0.45/day)
| System Name | Dire Wolf IV |
|---|---|
| Processor | Intel Core i9 14900K |
| Motherboard | Asus ROG STRIX Z790-I GAMING WIFI |
| Cooling | Arctic Liquid Freezer II 280 w/Thermalright Contact Frame |
| Memory | 2x24GB Corsair DDR5-6600 |
| Video Card(s) | NVIDIA RTX4080 FE |
| Storage | AORUS Gen4 7300 1TB + Western Digital SN750 500GB |
| Display(s) | Alienware AW3423DWF (QD-OLED, 3440x1440, 165hz) |
| Case | Corsair Airflow 2000D |
| Power Supply | Corsair SF1000L |
| Mouse | Razer Deathadder Essential |
| Keyboard | Chuangquan CQ84 |
| Software | Windows 11 Professional |
First, a public service announcement:
WARNING:
There is almost no room for error and you need a very steady hand with a Dremel (Or more professional tooling) to get this done. If someone attempts this and kills their card, I take no responsibility. Be careful with taking power tools to your video cards
Don't do this to a piece of hardware you'll miss if something goes wrong. Killing it is real easy.
And now we return to our scheduled report:
With the possible acquisition of a new TV and with the wish for a third monitor in any case, the need arose for a second video card in my system to drive that display. Since I did not want to take up the second PCI-E x16 slot in my system, and since PCI video cards are relatively expensive (and nearly impossible to find here locally), I decided to mod a Sapphire HD4350 (availiable everywhere here and el cheapo) to use a PCI-E x1 slot. I did not want to cut the slot on the motherboard (to make it open-ended), since the motherboard cost me plenty and is still under warranty, whereas the card was cheap.
In the process, I decided to also benchmark the effects of the HD4350 being run at full bandwidth, at PCI-E 2.0 x1 and PCI-E x1. Considering it is a weakling of a card, I thought the bandwidth might not matter, but as we shall see shortly, I was wrong. No issue for me, I only use it to drive two more monitors, but interesting to see nonetheless.
The testing system is my main rig in my system specs tab (The leftmost one in the list) and the benchmarks are ran were 3DMark06, 3DMark Vantage and Crysis at 1680x1050 on Low settings.
Here's the victim's box:
And here is the victim from the front:
And from the rear:
And how small it is next to my HD4870X2:
To check whether I am going to cut off the right number of connectors, I first tested the card with some of the golden fingers taped over. The card looked like this with most of the connectors covered:
And here's the back of it:
Dismemberment !
And here it is in the computer for testing:
And here it is sitting above my HD4870X2 (Temps on both are alright this way):
Success !
Okay, I promised some benchmarks, so here they are:
Crysis was benchmarked by running through the early game (Contact) until the GPS jammer on the beach (Including taking it out and watching the fireworks) and recording the FPS with FRAPS. 3DMV and 3DM06 were run on their defaults. The card was OC'ed to the maximum CCC Overdrive would allow it just for the heck of it.
First, the card running at half-full bandwidth. It comes up at PCI-E 2.0 x8 in GPU-Z since that's what the slots on the mATX DFI P45 are wired as, but this shouldn't have any effect on its performance - Even far more powerful cards do not mind the PCI-E 2.0 x8 bandwidth.
Crysis:
Min: 7
Avg: 16.559
Max: 24
3DMark06:
3DMark Vantage:
Now, the card running at the PCI-E 2.0 x16 slot in the pre-cut testing, taped over to reveal only one PCI-E lane. Resulting bandwidth is a single PCI-E 2.0 lane.
Crysis:
Min: 7
Avg: 15.009
Max: 23
3DMark06:
3DMark Vantage:
Finally, the chopped down card in the PCI-Ex1 slot.
Crysis:
Min: 6
Avg: 13.633
Max: 21
3DMark06:
3DMark Vantage:
As you can see, getting chopped down to a single PCI-E lane had an effect on the card's performance, with the biggest drop occurring in 3DMark06.
I hope this is informative and of use to someone. I had tons of fun doing it and was a little surprised I did not kill it in the process
WARNING:
There is almost no room for error and you need a very steady hand with a Dremel (Or more professional tooling) to get this done. If someone attempts this and kills their card, I take no responsibility. Be careful with taking power tools to your video cards
Don't do this to a piece of hardware you'll miss if something goes wrong. Killing it is real easy.
And now we return to our scheduled report:
With the possible acquisition of a new TV and with the wish for a third monitor in any case, the need arose for a second video card in my system to drive that display. Since I did not want to take up the second PCI-E x16 slot in my system, and since PCI video cards are relatively expensive (and nearly impossible to find here locally), I decided to mod a Sapphire HD4350 (availiable everywhere here and el cheapo) to use a PCI-E x1 slot. I did not want to cut the slot on the motherboard (to make it open-ended), since the motherboard cost me plenty and is still under warranty, whereas the card was cheap.
In the process, I decided to also benchmark the effects of the HD4350 being run at full bandwidth, at PCI-E 2.0 x1 and PCI-E x1. Considering it is a weakling of a card, I thought the bandwidth might not matter, but as we shall see shortly, I was wrong. No issue for me, I only use it to drive two more monitors, but interesting to see nonetheless.
The testing system is my main rig in my system specs tab (The leftmost one in the list) and the benchmarks are ran were 3DMark06, 3DMark Vantage and Crysis at 1680x1050 on Low settings.
Here's the victim's box:
And here is the victim from the front:
And from the rear:
And how small it is next to my HD4870X2:
To check whether I am going to cut off the right number of connectors, I first tested the card with some of the golden fingers taped over. The card looked like this with most of the connectors covered:
And here's the back of it:
Dismemberment !
And here it is in the computer for testing:
And here it is sitting above my HD4870X2 (Temps on both are alright this way):
Success !
Okay, I promised some benchmarks, so here they are:
Crysis was benchmarked by running through the early game (Contact) until the GPS jammer on the beach (Including taking it out and watching the fireworks) and recording the FPS with FRAPS. 3DMV and 3DM06 were run on their defaults. The card was OC'ed to the maximum CCC Overdrive would allow it just for the heck of it.
First, the card running at half-full bandwidth. It comes up at PCI-E 2.0 x8 in GPU-Z since that's what the slots on the mATX DFI P45 are wired as, but this shouldn't have any effect on its performance - Even far more powerful cards do not mind the PCI-E 2.0 x8 bandwidth.
Crysis:
Min: 7
Avg: 16.559
Max: 24
3DMark06:
3DMark Vantage:
Now, the card running at the PCI-E 2.0 x16 slot in the pre-cut testing, taped over to reveal only one PCI-E lane. Resulting bandwidth is a single PCI-E 2.0 lane.
Crysis:
Min: 7
Avg: 15.009
Max: 23
3DMark06:
3DMark Vantage:
Finally, the chopped down card in the PCI-Ex1 slot.
Crysis:
Min: 6
Avg: 13.633
Max: 21
3DMark06:
3DMark Vantage:
As you can see, getting chopped down to a single PCI-E lane had an effect on the card's performance, with the biggest drop occurring in 3DMark06.
I hope this is informative and of use to someone. I had tons of fun doing it and was a little surprised I did not kill it in the process
Last edited:
'
