Date: Tuesday , August 21, 2018
Most interesting is that these cards are "overclocked," unlike other FE cards before it. The 2080 jumps clocks to 1800Mhz from 1710Mhz, and the 2080 Ti FE clocks in at 1635MHz from 1545MHz. (You can see all the specifications here.) While NVIDIA still will not talk to HardOCP because of us telling you the truth about GPP, we did catch Tom Petersen on the live stream yesterday saying that the 2080 cards would be very overclockable, so we think there is a good bit of headroom left. I have known Tom for a long time, and he is not one to be full of BS in regards to things like these. The cooler on the FE is a full length vapor chamber, first used in retail cards by AMD. This cooler has more fins on it than any card we have likely ever seen. While we do not have pictures of the cold plate surface of the cooler, we do expect it to be a full contact card meaning that it has taken VRM and RAM heat into account.
The general layout of these cards is fairly traditional, but if you look at the picture below, which is huge in resolution, you will see that the PCB itself looks sparsely populated. NVIDIA notes that it uses an "all-new 13-phase iMON DrMOS power supply." NVIDIA notes that it has the ability to turn off power phases at lower workloads which of course is going to mean less heat produced while your card is not being pushed hard by your applications or games. It is notable that there is room for four more power phases down on the board in its current design. NVIDIA is boasting a VRM controller that works on a "sub-millisecond basis" providing for less variable current draw which will lead to better overclocking compared to Pascal cards. What exactly "better" means is of course a mystery at this point.
The RTX 2080 is noted at having a TDP of 225 watts and is noted as being able to use 280 watts when overclocked. This is the exact 280 watt number we saw recently on PNY's leak from last week. The FE has a one 6-pin and one 8-pin power connector, but many of the AIB cards we have already seen have two 8-pin connectors.
Of course we have the inclusion of 8GB or 11GB of GDDR6 providing 448GB/s and 616GB/s of bandwidth respectively.
The IO panel has a new feature as well. In its FE configuration we have 3 each DisplayPort connectors that support v1.4 (this means you can drive an 8K 60Hz display off a single port), a single HDMI 2.0b with HDCP 2.2 support, and the new VirtualLink USB Type-C connector for "next-gen" VR HMD support. This gives you single port connectivity for VR, but hopefully wired VR HMDs will be a thing of the past very soon. Still even with a wireless HMD setup, this will cut down on the at least one of the needed cables going to your breakout box. Not exactly earth-shattering, but this is good to see overall.
Where this new 2080 FE truly deviates from its predecessors is that it has dropped the blower style card coolers that are famous for making noise and not cooling as well as dual and triple fan coolers. Two 13-blade 3-phase motor fans are now in place on the new FE. NVIDIA says its new configuration is good for 10C better temperatures and is one fifth the noise produced when overclocked compared to previous FE cards. Of course anyone that has owned a blower card knows the value of the noise reduction.
NVIDIA made sure to make the back of the card look as good as the front, and even opted to supply us a matching cover for the NVLINK connector, so you don't have the ugly "golden fingers" sticking up in the air should you not need access. The backplate is stamped aluminum and literally helps round out the the no-edges look of the card's enclosure.
While we are still not sure about what all NVLINK will be about on the 2080 cards, the high-speed GPU interconnect is discussed as it pertains to Quadro here.
NVIDIA NVLink is the world's first high-speed GPU interconnect offering a significantly faster alternative for multi-GPU systems than traditional PCIe-based solutions. Connecting two NVIDIA Quadro graphics cards with NVLink enables scaling of memory and performance to meet the demands of your largest visual computing workloads.
What should pop in that sentence is "scaling of memory." Will we see this pertain to desktop cards and also give us mGPU? We are not sure on all the details about that yet exactly, but things are looking good in the memory scaling department, and with 4K monitors and high resolution textures somewhat becoming the norm, this could give NVIDIA a huge advantage in the gaming department, outside of simply owning the performance category.
One the the biggest issues we have seen with forward-looking GPU technologies over the years has been the fact that there was not gaming content that actually used the technology. Of course for game devs to spend the resources to use the tech, they need to know that there will be hardware on the ground for consumers before they will spend the money to implement it. NVIDIA seems to fully understand the value of this and has already made a huge push with game devs to make sure that we see this NVIDIA RTX technology very soon. Below are some of the promotional videos that NVIDIA has pushed out as well as game trailers that showcase RTX in action. A full listing of upcoming RTX Games can be found here.
GeForce RTX - Graphics Reinvented
Graphics Reinvented: Changing The Game With GeForce RTX
Battlefield V: Official GeForce RTX Trailer
Metro Exodus: Official GeForce RTX Video
Shadow of the Tomb Raider: Exclusive Ray Tracing Video
Control: Official GeForce RTX Trailer
Atomic Heart: Official GeForce RTX Video
If you like our content, please support HardOCP on Patreon.