• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

NVIDIA SFF-Ready System Build & Benchmark

"SFF" and a 300mm long GPU is a combo which just doesn't come up first to my mind.

Shouldn't a SFF have something like an ITX board with an "ITX" version of a card?

edit: Like the legendary Zotac GTX 1080 Ti Mini.


I know a guy that had one of these. He shunt modded and killed it :banghead:

Maybe if it's 300mm and single slot, like Galax 1070 Katana


The Katana was an unusual card, but I don't think any of these are ever coming back. Instead, they're changing the design of the cases so you can vertically mount the cards and make better use of the volume within. TPU has reviewed many if not most of the "certified" designs and they're not necessarily ITX either, seems like NVIDIA was targeting sub-35 liter volume for the program.

Look at the ASUS case for example, TPU review from 2022:


The even more compact, ITX-only Fractal Design Terra (which is my favorite of the bunch) was also reviewed this year:


If you see something in common between these micro-ATX and mini-ITX formats is that they all have a vertical GPU mount that maximizes the utilization of the compact space, in a very very similar way
 
I suppose that using a CPU such as a 14900K is impractical, but it's a good way to saturate thermals in a SFF context. 7800X3D is the go-to CPU when it comes to power and thermally constrained setups like SFF machines.
Judging from my 9800x 3d, I'd much rather have the 12900k in an ITX. There is a 30c delta between the two at the same workload using the same power and the same cooler.
 
I know a guy that had one of these. He shunt modded and killed it :banghead:



The Katana was an unusual card, but I don't think any of these are ever coming back. Instead, they're changing the design of the cases so you can vertically mount the cards and make better use of the volume within. TPU has reviewed many if not most of the "certified" designs and they're not necessarily ITX either, seems like NVIDIA was targeting sub-35 liter volume for the program.

The Katana is too niche sure, and perhaps even the 1080Ti mini

But surely asking for, as mentioned, single fan GPU (literally just put a 4060Ti's cooler on a 4070. or like the thread mentions, reuse designs for current cards) that takes 180-200W as we've had for 7/8 years is not asking for much?

The reddit thread within has examples I've found over the years. There are some good options for the 4070Ti Super, so besides Nvidia's segmentation/pricing/markup it's not as bad as it was.
 

Attachments

  • SilverStone_ZRNRCgG.jpg
    SilverStone_ZRNRCgG.jpg
    331.9 KB · Views: 44
  • lLnSsyM.jpg
    lLnSsyM.jpg
    312 KB · Views: 44
Judging from my 9800x 3d, I'd much rather have the 12900k in an ITX. There is a 30c delta between the two at the same workload using the same power and the same cooler.

Larger chips dissipate heat a lot better. Haswell-E Xeons are some of the best examples, they're super large, with low to moderate wattage (my 18-core chip sipped 80W at best full load), temperatures get very low on the CPU side but they'll easily overwhelm the cooling solution simply because the die area is so large and it can transfer heat so much more efficiently than newer, smaller chips. This is good if you have lots of real estate to have the heat dissipate over time (or a very fast fan to blow the heat away), but in an SFF context, it's going to warm the entire system up very fast.

Ryzen's chiplets have exceptionally high heat density and are very difficult to cool due to the small contact surface, meaning that the CPU itself will run hotter and much of the cooling potential of the heatsink assembly or liquid cooling system goes unused. This usually reflects on cooler reviews, where the Intel chips usually go more than 100 W farther on the same cooling system than the AMD chip, using a high-end 360mm AIO as an example:

amd-max-rpm.png


intel-max-rpm.png


 
This SFF-Ready certification is a bit of damage control for Nvidia partners making too many of their GPUs needlessly large. To some extent, there are thermal and noise-level improvements to using a larger cooler on a GPU, but with many of the reviews these days I'm seeing vendors compete to have the lowest operating temperature. Why? The Silicon boosts all the way to 83C and is power-gated and voltage-gated strictly by Nvidia, so there's no point having all this headroom for overclocking since Nvidia has that locked down so tightly. Instead, we're paying for these enormous coolers that have so much unusable headroom it's silly.

The Founders Editions are the GPUs I buy for SFFs, because they're a sensible size. If I want a quieter card and have room for a larger cooler there are plenty of options that aren't over 30cm long and using 4 slots. Those gigantic 'flagship' factory OC variants of the 4070Ti/S, 4080/S and 4090 are just willy-waving contests between manufacturers and they're needlessly impractical.

I kind of understand the rationale of having a tall PC that doesn't occupy that much desk space, since most people are not particularly low on vertical space. After all, that's the reason why skyscrapers exist.
SFFs are definitely more about footprint for most people. Desk surface is premium real estate but with most cases putting IO on the front and often low down, the height of the case is almost irrelevant. It's one of the reasons I think fishtank dual-chamber cases are dumb - they're display cases so they're going to go on top of a desk rather than underneath it, but the relocated PSU makes the footprint absolutely vast which eats up tons of vital desk space that is best reserved for important things that need to go in a specific place like your mouse pad, second screen, or one of your fancy speakers.

The issue with SFF is still the cost of Mini ITX MBs. You can pay 3 times what a Matx board would cost where I live and some of them are more expensive than ATX too. I also have issues with smaller cases in terms of install and cable management. If indeed Mini ITX boards were less expensive they would be more attractive as once you buy RAM for Mini ITX you are in the range of the ROG Ally. Some Mini ITx boards actually cost more than the Steam Deck when it is on sale.
If it's an SFF that's not using a riser cable to relocate the graphics card, mITX is utterly pointless anyway. When you're plugging a 2.5-slot GPU into a motherboard, that needs additional clearance to let the GPU fans breathe, you're at the same number of slots as an mATX board anyway. As for the length of the board from front to back, mITX is equally pointless because GPUs are typically longer than even an mATX board.

There are a few cases on the market that can squeeze a high-end gaming build into an mATX case, and I wish there were more options like them, because losing so many M.2 slots and being limited to 2 RAM slots is bad enough, but paying double or triple the price to be hit with such compromises is a hard pill to swallow.
 
This SFF-Ready certification is a bit of damage control for Nvidia partners making too many of their GPUs needlessly large. To some extent, there are thermal and noise-level improvements to using a larger cooler on a GPU, but with many of the reviews these days I'm seeing vendors compete to have the lowest operating temperature. Why? The Silicon boosts all the way to 83C and is power-gated and voltage-gated strictly by Nvidia, so there's no point having all this headroom for overclocking since Nvidia has that locked down so tightly. Instead, we're paying for these enormous coolers that have so much unusable headroom it's silly.

The Founders Editions are the GPUs I buy for SFFs, because they're a sensible size. If I want a quieter card and have room for a larger cooler there are plenty of options that aren't over 30cm long and using 4 slots. Those gigantic 'flagship' factory OC variants of the 4070Ti/S, 4080/S and 4090 are just willy-waving contests between manufacturers and they're needlessly impractical.


SFFs are definitely more about footprint for most people. Desk surface is premium real estate but with most cases putting IO on the front and often low down, the height of the case is almost irrelevant. It's one of the reasons I think fishtank dual-chamber cases are dumb - they're display cases so they're going to go on top of a desk rather than underneath it, but the relocated PSU makes the footprint absolutely vast which eats up tons of vital desk space that is best reserved for important things that need to go in a specific place like your mouse pad, second screen, or one of your fancy speakers.


If it's an SFF that's not using a riser cable to relocate the graphics card, mITX is utterly pointless anyway. When you're plugging a 2.5-slot GPU into a motherboard, that needs additional clearance to let the GPU fans breathe, you're at the same number of slots as an mATX board anyway. As for the length of the board from front to back, mITX is equally pointless because GPUs are typically longer than even an mATX board.

There are a few cases on the market that can squeeze a high-end gaming build into an mATX case, and I wish there were more options like them, because losing so many M.2 slots and being limited to 2 RAM slots is bad enough, but paying double or triple the price to be hit with such compromises is a hard pill to swallow.
I know exactly what you are saying. I have never envisioned a Mini ITX system with a DGPU. As I stated before they have only become viable with the advent of APUs but that was killed by the 8700G being over $400 where I live. If AM5 CPUs did not come with basic Graphics I would not have got this board in the first place. The thing though is that this will be much smaller in a suitcase.
 
Back
Top