Crowdfunded prebuilt SFF Gaming PC (custom case)

Valantar

Shrink Ray Wielder
Jan 20, 2018
2,201
2,225
yeah if you compare 1080s with 2080s then off course. Considering the RTX 2070 will performing between a GTX 1080 and a GTX 1080Ti, I think that's good. BTW the 30W higher power draw is due the USB-C Port I guess.
That seems doubtful - TDP is supposed to denominate the necessary cooling of a board ("Thermal Design Power"), and (even with some minor heat output from the necessary voltage conversion) a USB port capable of outputting 25W doesn't increase the need for cooling on the board. Any heat from the power passing through the USB port will be generated in the HMD or whatever unit it's connected to, not on the GPU board. As such, I doubt it affects TDP at all.

The reason for the increased TDP is far more likely due to the massive increase in die size and the smaller increase in cores. Even with the process refinement going from 16nm to 12nm (which is really just a minor improvement on the same node), you can't increase die area like they've done here without a power penalty. While the RT and Tensor cores are rumored to take up ~50% of the die, and won't be used for the majority of games, you still have a 21%(Ti)/15%(2080)/20%(2070) increase in CUDA cores with no major process improvement to alleviate the increase. The lowering of base and boost clocks would seem to corroborate this (though those clocks might be lowered to account for throttling while running hybrid RT/Tensor workloads). In other words: we don't know, but increased TDPs definitely don't speak to lower power draw.

Right. The case I built right now will be definitly not good to sell as a case. When I would want to sell a case I've another project. Maybe I will work on this sometime later. I feel very confident about what I want to do, I understand what it involves. I want people to be able to purchase complete builds from my site anyways at some time later in my life. Why not do it now. I will keep you all updated.
I don't suppose you'd care to show us this case? I for one am quite curious.
 

Freeman1

Chassis Packer
Original poster
Sep 12, 2018
16
4
That seems doubtful - TDP is supposed to denominate the necessary cooling of a board ("Thermal Design Power"), and (even with some minor heat output from the necessary voltage conversion) a USB port capable of outputting 25W doesn't increase the need for cooling on the board. Any heat from the power passing through the USB port will be generated in the HMD or whatever unit it's connected to, not on the GPU board. As such, I doubt it affects TDP at all.

The reason for the increased TDP is far more likely due to the massive increase in die size and the smaller increase in cores. Even with the process refinement going from 16nm to 12nm (which is really just a minor improvement on the same node), you can't increase die area like they've done here without a power penalty. While the RT and Tensor cores are rumored to take up ~50% of the die, and won't be used for the majority of games, you still have a 21%(Ti)/15%(2080)/20%(2070) increase in CUDA cores with no major process improvement to alleviate the increase. The lowering of base and boost clocks would seem to corroborate this (though those clocks might be lowered to account for throttling while running hybrid RT/Tensor workloads). In other words: we don't know, but increased TDPs definitely don't speak to lower power draw.


I don't suppose you'd care to show us this case? I for one am quite curious.

Yeah you are right with the TDP. I just read it somewhere and took it as fact without thinking about it if it makes sense. Obviously it doesn't as you explained. The jump in efficiency isn't big, that's true but there's one and that being a fact the heat generated per frame is less so will the power draw.
I heard that one game company will use the RT cores for Global Illumination and DLSS is present in quite a bit more games which uses the Tensor Cores. (even if I wouldn't use it because I want to play at 4K without AA but let's see how DLSS will look)

I will reveal the case closer to production. There's a lot of things (for example collaborations with other companies) that will have to go right that the case will even makes sense. That sentence doesn't make sense to you now but it will after the reveal.
 

Damascus

Master of Cramming
Feb 27, 2018
550
387
I have a couple big questions

1.
What's the appeal of your system? Is it a drop dead gorgeous, tiny and low compromise but higher complexity sub 5L system like the S4m or ZS-A4M? Is it a more consumer friendly build in the vein of the dancase (+ clones), or the modivo lineup? Does it offer the extreme modularity and top notch production that the loque ghost or streacom cases offer? Is it a larger but better cooling case like the Ncase M1?

Basically, is there a niche where your systems can beat the competition? Given your use of a reference PCB GPU and custom PSU (hopefully internal, maybe a special run of flex atx?) we can assume your case is at least 6L, with a more likely estimate of 8-14L if you want to quietly cool i9's.

2.
Beyond that, while there is clearly a consumer market for prebuilts, why would someone choose your rig over one built in an S4m, by certified professionals with years of experience? Does someone with enough money to buy an i9 rig worry about the margin they'd need to pay such a system integrator? Probably not, it would be a fraction of a fraction of what they're already paying.
 
  • Like
Reactions: el01