Alleged NVIDIA GeForce RTX 4090 Ti Founders Edition graphics card cooler leaked at Chiphell Forums. The cooler may give us a first look at the updated Founders Edition design seen on the next generation of Ada Lovelace GPUs.
NVIDIA GeForce RTX 4090 Ti Founders Edition graphics card cooler from alleged leaks: massive heatsink, larger core board for both GPU/memory coverage
Recently, there were reports that NVIDIA’s GeForce RTX 40 series graphics card could be launched in mid-July. This may indicate that we are only 2-3 months away from the first reveal of the next generation lineup based on Ada Lovelace’s graphics architecture. The GeForce RTX 40 series is set to be powerful and power-hungry, and will require some solid cooling and it looks like NVIDIA is making a significant upgrade to the current Founders Edition design to meet the 450-600 Watt specs.
Pictures of the alleged “GeForce RTX 4090 Ti” show us three different parts of the FE cooler. First, we have a silver-tone metal chassis that circles the heatsink to give it a beautiful finish. Currently, the metal plate has a small card logo fitted to it. The leaked board indicates that it will likely run on the main NVIDIA GeForce RTX 4090 Ti graphics card. But, one thing to point out is that the position of this logo has been changed from right to left. This part will only be visible on the underside of the graphics card so it doesn’t mean much where to put the label but from a marketing point of view, it might look weird.
The most interesting part is the heatsink itself. It looks like NVIDIA will now use a larger core board that will not only cool the GPU but memory dies out as well. Currently, NVIDIA uses thermal pads to dissipate heat from the memory death to heatsink on the RTX 3090 and 3090 Ti Founders Edition graphics cards.
NVIDIA’s new heatsink design will allow to cool at least a dozen GDDR6X memory sticks. The same configuration as the RTX 3090 Ti PCB will likely be used here with 4 GDDR6 chips on the right/left side, three on the top, and 1 on the bottom. With NVIDIA expected to use the high-spec GDDR6X chipset on its next generation cards, additional cooling will really be necessary for key AD102 parts. There is also a large aluminum heat spreader for the VRMs and heat pads for various components. The heatsinks on the sides also appear to have been upgraded from 23 fins to 24 fins on the leaked design.
We can also see the heat pipes leading from the main heatsink to the one in the back. Heat will be dissipated from the heatsink and from the card through a similar flow-cooling design as existing FE cards. With that said, the card looks like it’s going to be 3 or even 4 slots this time around. Now it remains to be seen if the cooler RTX 4090 Ti happened to leak too soon, or if it will be part of the same lineup.
So far, there are only reports that the RTX 4090 will be the first to be released on the market with the release of the RTX 4090 Ti later. This could be an early design of the “Ti” cooler or it’s possible that NVIDIA made quite a few prototype coolers and decided to use the “non-Ti” variant first. Again, this is a purported image of what might be the next-generation NVIDIA GeForce RTX 40 series cooling design, but we can’t say for certain, so treat it as a rumor for now.
NVIDIA GeForce RTX 3090 Ti Founders Edition Cooler (Image credits: Tech Power Up):
Here’s everything we know about the NVIDIA GeForce RTX 4090 series graphics cards!
The AD102 “Ada Lovelace” GPU is is expected It houses 144 SMs, a 71% increase over the current GA102 GPU and houses a whopping 18,432 CUDAs. The GeForce RTX 4090’s GPU is rumored to have 16,128 CUDA cores which is a 50% increase in core count over the RTX 3090 but 14% less than a full-fat AD102 configuration that may arrive later as the RTX 4090 Ti. It’s the same situation in this generation where the ‘Ti’ variant of the 3090 got a full-fat GA102 GPU and the non-Ti variant, although released earlier, came in a reduced configuration.
Alright, let’s do a fresh recap.
RTX 4090, AD102-300, 16128FP32, 21Gbps 24G GDDR6X, 450W, ~2×3090.
I am disappointed with RDNA3.
– kopite7kimi (@kopite7kimi) May 16 2022
In addition to the SM count, Ada Lovelace GPUs will also have increased L2 cache sizes. Starting with the AD102 GPU, the flagship will be equipped with up to 96MB of L2 cache, an insane 16x increase over the 6MB L2 cache found in the GA102. This cache is still not as large as the Infinity Cache found on AMD RDNA cards but it’s still a huge leap for NVIDIA’s graphics cards.
The NVIDIA GeForce RTX 4090 graphics card will also boost 24GB of 21Gbps GDDR6X memory via the 384-bit bus interface and deliver bandwidth in excess of 1.0TB/s. The latest details also bring the overall TDP numbers down to 450W which means the RTX 4090 might not go crazy with the hours the full-fat 600W variant is expected to fit.
As we’ve seen with the GeForce RTX 3090 Ti, there are certain models that go above 450W TBP (reference) and go up to 516W (custom). The same will happen with cards based on Ada Lovelace. And while this will require a lot of cooling, it looks like the AIBs are ready and ready. We’ll likely see more hybrid-style GPU coolers used instead of traditional air coolers once we cross the 600W TBP limit. This will result in a 33% increase in TBP on the RTX 3090 Ti. According to the latest rumors, NVIDIA is is expected To announce the next generation GeForce RTX 40 series by mid-July, which means we’ll get a new generation of cards early in the third quarter of 2022.
NVIDIA GeForce RTX 40 Series graphics cards (rumored):
|Photographers card||GPU||Variable PCB||SM modules/cores||memory / bus||Memory clock / bandwidth||TGP||power connectors|
|NVIDIA GeForce RTX 4090 Ti . Graphics Card||M 102 – 350?||to be announced later on||144/18432?||48 GB / 384 Bit||24 Gbps/1.15 terabytes/sec||~600 watts||1x 16 teeth|
|NVIDIA GeForce RTX 4090||M 102 – 300?||PG137 / 139 SKU330||126/16128?||24 GB / 384-bit||21 Gbps/1.00 TB/sec||~ 450 watts||1x 16 teeth|
|NVIDIA GeForce RTX 4080||M 103-300?||to be announced later on||> 84/10752?||16 GB / 256 bit||21 Gbps / 672 Gbps||~ 350 W||1x 16 teeth|
|NVIDIA GeForce RTX 4070||AD104-400?||PG141-310 SKU341||>60/7680||12 GB / 192 bits||18 Gbps / 432 Gbps||~ 300 watts||1x 16 teeth|