NVIDIA released its interesting, almost concept-style GeForce GTX 295 graphics card 15 years ago now... with sandwiched PCBs with dual GPUs enabling SLI on a single card back in 2009.
The NVIDIA GeForce GTX 295 with dual GPUs cost just $499 back in the day, which isn't bad for a ridiculous dual-GPU monster card. It isn't a monster today at all, with its GT200 GPU made on the older 55nm process node. Each GT200 GPU featured 240 unified processors for a total of 480 unified processors through the dual-GPU design, twice that of the GeForce GTX 280 and its single GPU.
At the time, the GTX 280 had its benefits -- from being a single-GPU card -- but the GTX 295 dual-GPU graphics card wasn't just two GTX 280s together. NVIDIA had a monster dual GT200 GPU sitting on the GeForce GTX 295, but the rest of the specs were knocked down between the GTX 260 and GTX 280 single-GPU graphics cards.
This was in the form of a lowered GPU clock of 576MHz, and a 448-bit memory bus on the GTX 295 instead of the 512-bit memory bus on the GTX 280. NVIDIA had 1792MB (1.8GB) of GDDR3 memory on the GeForce GTX 295, while there was 1GB of GDDR3 memory on the GTX 280.
- Read more: GIGABYTE GeForce GTX 295 Graphics Card Review
- Read more: Unleashing NVIDIA's GTX 295 Graphics Card Review
- Read more: ASUS GeForce ENGTX295 Graphics Card Review
- Read more: MSI GeForce GTX 295 Graphics Card Review
NVIDIA required one 8-pin and one 6-pin power connector, with a total of 289W of power consumption. The cooling design was interesting through the use of two separate PCBs and a single fan, and still maintained its tighter dual-slot design. At the time, PCIe 2.0 x16 was all the rage, so NVIDIA used it all up with a dual-GPU design, while it featured old-school dual DVI connectors and a single HDMI port.
NVIDIA's crazy GeForce GTX 295 dual-GPU graphics card featured SLI support, using Alternate Frame Rendering (AFR). This means it rendered frames interchangeably between the two GPUs, but it was far from perfect. GTX 295 owners and other dual-GPU users at the time had to go through micro-stuttering, even when you had high FPS on the screen.
More issues included both GPUs needing to access identical data, which meant using all of that 1792MB of GDDR3 memory was accessed for the same assets. Because of the dual GPU used on the GTX 295, there's really only 896MB of VRAM on each PCB.
This was such a cool time in PC gaming history, as I actually had multiple dual-GPU cards in my system from both sides with NVIDIA and AMD back in the day. AMD had its Radeon HD 4780X2 graphics card, which was also made on the 55nm process node.
NVIDIA hit back with other dual-GPU card designs like the GeForce GTX 590 and TITAN Z graphics cards... this entire post makes me miss those days. Sure, we've got the likes of the GeForce RTX 4090, but it would be nice to see some monster 1000W graphics card with dual GPUs that work in pure harmony with 48GB of GDDR6X memory in total.
I miss those days SO much.
Because NVIDIA opened the door to dual-GPU graphics cards with releases like the GTX 295, people like myself and many others, used two GTX 295s for a monster quad-SLI setup. It wasn't perfect, using SLI in dual-GPU and quad-GPU form was never perfect, but it was a wild, wild time for PC enthusiasts that wanted the bleeding edge. That's something we reviewed back in the day, running the GeForce GTX 295 in quad-SLI at the time.