GT1030 shunt mod

[Aranwe]

I had an old 1030 laying at home that I never use, so I used it to experiment with shunt-modding. Even though this card doesn't make any sense price to performance wise, it is a great opportunity to experiment without the risk of bricking an expensive peace of hardware.

 

The particular card I used is an MSI GT1030 Aero OC. It is the GDDR5 model and has active cooling whereas some use passive cooling. This makes it a great contender for a shunt mod because it's basically always at 40°C and caps out at around 25~30W. So without further ado, here goes :

 

Baseline figures:

Stock, it boosts to 1782 MHz on the core and 3003 MHz on the memory. This is already substantially higher than the reference boost clock of 1468 MHz. Sadly enough, because all of the power is pulled through the PCIe slot I had no way of measuring the power it pulled, and there are no sensors reporting power.

 

Results:

I was able to push the card to 2164 MHz on the core and 3804 MHz on the memory in Heaven, though these clockes weren't stable in Superposition where it did 1924 Mhz / 3803 MHz. Geekbench is not the best benchmark out there but I ran it anyway just for the fun of it.

 

Maybe I'll look into soldering the components of the missing phase they didn't bother putting on the board, or maybe design a "zombie" VRM for this card (I'm an electrical engineer so I'm quite liking that idea). I'm once again reminding you this is mainly a learning opportunity I'm sharing and has little use on a card like this.

 

Anyway let me know if you have any suggestions for future mods.

 

Tschüss

 

 

 

 

[manikyath]

love the learning opportunity.

 

besides that you essentially proved that this exercise is purely academic, it would be  an interesting next step to disconnect +12v from pcie, and bring it in on a propper connector.

[Somerandomtechyboi]

id suggest trying to put a tec cooler on it, maybe subzero temps will help with ocing, though you will prob have to do abit of janky wiring to power the damn thing, though atleast it doesnt look too hard since the gpu and vram are pretty close to eachother (if you subzero gpu + vram), i think a 72w tec would suffice here

 

Though you do have to figure out how to mount it + wiring + insulating (art eraser or closed cell foam would work) + cooling the tec (aio would prob be the most convenient solution here)

 

 

Im looking to do something similar with my ram cause 1250mhz only = 625fsb and 1250 isnt even stable, max i think is 1150 = ~580fsb, im on ancient lga 775 so bad ram = bad ocing results

[Aranwe]

2 minutes ago, Somerandomtechyboi said:

id suggest trying to put a tec cooler on it, maybe subzero temps will help with ocing, though you will prob have to do abit of janky wiring to power the damn thing, though atleast it doesnt look too hard since the gpu and vram are pretty close to eachother (if you subzero gpu + vram), i think a 72w tec would suffice here

 

Though you do have to figure out how to mount it + wiring + insulating (art eraser or closed cell foam would work) + cooling the tec (aio would prob be the most convenient solution here)

 

 

Im looking to do something similar with my ram cause 1250mhz only = 625fsb and 1250 isnt even stable, max i think is 1150 = ~580fsb, im on ancient lga 775 so bad ram = bad ocing results

A peltier module might indeed be an option because of the low power output of the chip. I can just terminate a tec to a molex connector. What would be more concerning is how to manage condensation. A tec can probably pull the core to around -20°C at most with a 70W module.

Thanks for the idea, I'll look into it

[Somerandomtechyboi]

2 minutes ago, Aranwe said:

A tec can probably pull the core to around -20°C at most with a 70W module.

I think youll need a good cooler for that cause i think tecs will create a temp diff/delta between the hot and cold side, the colder the hot side the colder the cold sidw and vice versa

 

I dont have the proper equipment to measure my tec temps so until i can get something like that then im prob not gonna f around with overvolting the tec cause after all i can only cool it with a cheap chinese 4 heatpipe tower cooler, ill have to find what volt i can get my tec to be at its coldest and pretty sure hand feeling is not gonna cut it, currently it just freezes over anything that i put on its cold side, even without tim itll freeze and if i wait long enough the object that i put on the cold side becomes welded to the tec and i have to turn the thing off to get the object off

[manikyath]

1 minute ago, Somerandomtechyboi said:

I think youll need a good cooler for that cause i think tecs will create a temp diff/delta between the hot and cold side, the colder the hot side the colder the cold sidw and vice versa

 

I dont have the proper equipment to measure my tec temps so until i can get something like that then im prob not gonna f around with overvolting the tec cause after all i can only cool it with a cheap chinese 4 heatpipe tower cooler, ill have to find what volt i can get my tec to be at its coldest and pretty sure hand feeling is not gonna cut it, currently it just freezes over anything that i put on its cold side, even without tim itll freeze and if i wait long enough the object that i put on the cold side becomes welded to the tec and i have to turn the thing off to get the object off

i like the idea of a tec, but that's a lot of work to get right, and a heatsink capable of sinking the 40W of the GPU, and 80-something W of the tec combined.

[Somerandomtechyboi]

32 minutes ago, manikyath said:

i like the idea of a tec, but that's a lot of work to get right, and a heatsink capable of sinking the 40W of the GPU, and 80-something W of the tec combined.

Yea they are super inefficient though easier to build then a full on phase change cooler so more accsesible

 

Maybe ill look into phase change ram + cpu + gpu but its prob gonna be really hard to build, maybe later in the future but for now i just need to subzero ram, once ram bottleneck is gone then i might move on to subzeroing the nb, currently the bottlenecks from immedeate to lesser is cooling - mb - ram - nb

 

Spoiler

The aforementioned chinese cooler is the cooler that i use so thats a bottleneck, then the mb i use is a p5q (modded bios to p5q pro turbo) and fsb caps ~530, still gets its ass whooped by my g31m s2c that does 542 (maybe higher if i voltmod nb and sb), then ram which is team extreem dark 1066 ddr2 6-6-6-18, im running ~1150 6-6-6-12 atm and it caps at around 1250, once ive gotten my hands on a gigabyte ep45 im gonna have to subzero the nb if i want to venture into 700fsb territory

 

Im kinda torn on whether i should work on voltmodding my g31 and making that my main rig or just using the p5q cause p5q sucks ass fsb wise compared to the g31 but has alot more ocing settings than the g31 (non voltmodded), though for voltmodding ill have to ask another forum cause its some pretty niche info nowadays

 

Btw for 14nm absolute max safe volt seems to be around 1.6v and deathzone is around 1.9v so you got quite abit of volt headroom though you just need to have adequate cooling for the card

 

For the tec idea i think youll need to cover the gpu and vram with a 5mm thick peice of metal (copper prob works best but aluminum is cheaper), i think its to stabilize the temp or something or thats what i heard when i asked w9 forums about nb tec cooling

[Aranwe]

I just stumbled on this forum thread with someone saying they managed a voltage mod on the 1030. Might be worth looking into because we might be voltage limited. The thread is pretty much dead so I'll try it for myself and share if I get any results

 

https://community.hwbot.org/topic/171990-my-vmod-experiences-with-galax-gt-1030/?do=findComment&comment=489480