Low profile graphics cards have always been something of a niche market, but they've found homes in many HTPC builds over the years, especially passively cooled cards. Now Gigabyte has launched a rather odd looking low profile NVIDIA GeForce RTX 4060 based graphics card that simply goes under the name of GeForce RTX 4060 OC Low Profile 8G. The 182 x 69 x 40 (D x H x W) mm card still manages to pack in three fans and requires an 8-pin power connector to work, which somewhat unfortunately plugs in at the rear of the card, although there wouldn't have been too many other locations to place it. This means that this card might not work in some Mini-ITX builds, due to the card being too long once the power connector is plugged in.
The card width should've given away that we're looking at a dual slot card, which is also pretty obvious from the pictures. Gigabyte provides a full height and a low profile bracket with the card, which not all low profile cards have shipped with in the past, so this is a plus. Connectivity wise, the GeForce RTX 4060 OC Low Profile 8G comes with two DP 1.4a and two HDMI 2.1a ports, making this a card well suited for HTPC usage. Gigabyte has even overclocked the GPU from 2460 MHz to 2475 MHz, which seems quite pointless and is unlikely to bring any huge performance advantages over stock clock speeds.
Update Aug 24th: During a recent meeting with Gigabyte we had a chance to go hands-on with their new RTX 4060 Low Profile card, here's some photos.
138 Comments on Gigabyte Quietly Launches Low Profile GeForce RTX 4060 Graphics Card
The Minisforum UM773 Lite is just 359 $ barebone and you get DDR5 + Radeon 680M. Pretty good deal!
Yes, the 4060 is faster, and Gigabyte's effort is appreciated, but it really doesn't make for a good LP card where LP is actually needed:
- The length isn't too much longer than the A2000, but the 6-pin adds to it. All in all now you're looking at length similar to 2060FE. Generally not a great deal of room to play with in these cases.
- Not having a blower = 100% exhaust dumped inside already a tiny or narrow case. No go considering the extra TDP, and the downdraft CPU coolers you will be using are also doing the same thing, all in a case that usually lacks any form of active exhaust whatsoever.
- Needing a 6-pin is just a major headache for picoPSUs all around, even HDPlex DC-ATX that has provisions for one. A2000 runs off slot power.
For better or worse, these cards will forever be niche. It's been the trend for years now, in the ~5L space to sacrifice CPU cooler height in favour of a sandwich layout that takes short, full-height dual slot GPUs. Case is no bigger, the 6-pin is no longer a problem, sandwich is easier to organize, and GPU heat also can be separated from everything else somewhat. Everyone follows the same formula now. People would much prefer getting any short single-fan 4060 for MSRP rather than have to hunt for one of these LP 4060s at an inflated price.www.notebookcheck.net/Radeon-680M-vs-GeForce-GT-1030-Desktop_11124_7996.247598.0.html
So that could work. For HTPC use, I'd prefer the balance of power shifted a lot more heavily into the GPU, but this is cheaper than many other options and it is a very small package. There is also a 780M option from Minisforum which is also tempting. This is a common thought. Traditionally, each generation, every model is at least as performant as the next tier up from the previous generation. But the 4060 Ti can't match the 3070; in fact in many cases it barely beats the 3060 Ti. And it's a smaller chip with a smaller memory interface and a lower power budget. Compare this to the 4090, which is like 40% faster than the 3090 Ti, has comparable resources, and uses more power. I believe the 4060 and 3060 follow the same pattern, since the 4060 Ti landed in the performance class I expected of the 4060.
And now, the 4060 has made it into a form factor that traditionally doesn't have room for anything bigger than a 50-series GPU, which is consistent with the idea that Nvidia named the 4060 one class up from what tradition would've called it.
My main tv is still a dumb tv. So I pair it with an old pc (Xeon X79) running Android TV x86 dual boot with Windows for other streaming needs.
At $300, the 4060 offers a larger 8GB framebuffer and a higher class of performance.
The 12GB models are north of $450.
So yeah. The length shouldnt be an issue. The majority of LP builds are either using the few DIY LP cases out there, which have 10+ inches for expansion length internally, or are using dell/HP chassis, which offer the same. Truly tiny LP cases with 2 slots are disappointingly rare, something I've ranted about before.
The blower isnt a big ordeal. The LP radeon 7850 had an open frame cooler, it was fine. Open frame LP 560x and 1650 cards work fine. So long as you have any sort of side vent by the GPU it will dump heat out of the chassis.
There's no issue for PICOs. A simple molex to 6 pin will handle the power needs of a 4060. Although I'd argue that most PICOs are too small for driving a card like this. HDplex has 6/8 pin connectors on board.
my post was more addressed towards those who believe that this is like, the second coming of christ and prior to this (somewhat) performant LP, SFF gpu's didn't exist - which is simply not true. the A2000's been readily available at a price not (too) off the charts for a good long while now.
(there's also the AD4000, but that's truly a moonshot-priced card, so there's that.)
Also, to my knowledge no x50 card post-Fermi required more than 100W until Ampere. They're also traditionally meant to address the sub-$150 space. The 4060 conclusively outperforms everything available at the $200-250 price point, which, incidentally, is the traditional stronghold of the 60-series. People can keep wringing their hands over the memory bus or the number on the chip, but it does exactly what a 60-class chip is meant to do. It just does it for too much money.
Like I said, HDPlex have provisions for 6-pin power thus it is not a compatibility issue by itself, but it adds to the already problematic/dealbreaking length and makes routing hell when space is at a premium.
The previous generations *50 tier cards had similar relative specifications.
RTX 4060 AD107 159 mm^2
RTX 3050 GA106 276 mm^2
GTX 1650 TU117 200 mm^2
GTX 1050 GP107 132 mm^2
GTX 950 GM206 228 mm^2
GTX 750 GM107 148 mm^2
GTX 650 GK107 118 mm^2
Do you see where it really belongs?
Now, look at the *70 tier.
RTX 4070 AD104 294 mm^2
RTX 3060 GA106 276 mm^2
RTX 2060 TU106 445 mm^2
GTX 1660 TU116 284 mm^2
GTX 1060 GP106 200 mm^2
GTX 960 GM206 228 mm^2
GTX 760 GK104 294 mm^2
GTX 660 GK106 221 mm^2
What you get is that nvidia rebranded the chips with one or two performance labels up, and charges three or four price tag levels up.
2 - I don't think cellphones support 4k and Dolby Atmos as passthrough.
What you want will be problematic with a custom x86 android machine, but you could also use the native tv apps just for that or a different media stick like a nvidia shield, google or fire tv for example. I think the use case you're specifying isn't very important for people running custom htpc's because they rely more on their own unrestricted sources if you catch my drift