According to a well-known leak source, more details about the technical specifications of the NVIDIA GeForce model of RTX 4090, GeForce RTX model of 4080, and GeForce RTX model of 4070 VGA cards have leaked online. According to sources from the recent some time there are no such rumors regards to the series of 4000. Now, surprisingly a new leak from user kopite7kimi suggests that the latest model of the RTX 4090/4080/4070 specification of further reports has added some details compared to the last reports for your last models.
The series of these cards according to leakers is expected to use the upgrade math technology of Ada Lovelace GPUs based on the newly updated architecture and built on the process node of the TSMC 4N, a slightly optimized variant process of the 5nm node. The changes that happen to every each product regarding specs have brief descriptive new features which are going to be discussed are as follows:
NVIDIA GeForce RTX 4090 new specifications will be
According to the source of the leakers this model of RTX 4090 is expected to come with data from insider kopite7kimi, which suggests that the NVIDIA GeForce RTX 4090 which much have the frequency range of 2235 MHz which can be a boost of 2520 MHz with actual game clocks of 2750 MHz. The power of this graphic card is expected high-end with AD102-300 GPU, but it only happens in the “Ti” variant that will feature the chip of filled. When we talk about the latest CUDA 16384, it simply means that at the stated clock frequency of 2750 MHz, GPU processing power can reach 90 TFLOPS for single-precision workloads. This performance must say that it is twice as much as the GeForce RTX model of 3090. In line with earlier rumors, kopite7kimi is also dropping an RTX 4090 with a total board power of 450W. This corresponds to the maximum amount of power the 12VHPWR connector can carry without having four additional sensor pins.
NVIDIA GeForce RTX 4080 new specifications will be
RTX 4080 will be expected that the memory up-gradation is the target for an increase in efficiency of the card, according to leaks the bus of 256 bits and 16GB of GDDR6X clocked memory at the same speed as the RTX 4090 while comparing to the recent models. The RTX 4080 will be a shortened configuration of AD103-300 GPU with cores of 10,240 or 80 SM enabled out of a total of 84 units. The GeForce RTX 4080 will reportedly come with a completely different processor compared to the top-end card, known as the AD103-300. We’ve already seen the “103” GPU; there’s the GA103, which is used in Ampere’s top-of-the-line mobile GPUs.
The Ada Lovelace version will reportedly have 10,240 CUDA cores, giving it 80 SM out of a possible 84. After this there is an expected updated memory in this series specifications, the leak reveals the rock 16 GB of GDDR6X capacities, which are said to be tuned to 18 Gbps speeds across a 256-bit bus interface, which leads to the bandwidth of 576GB/s
It’s a little strange that the RTX 4080 will have a lower bandwidth than its RTX 30-series predecessor, but we may see a GDDR6X variant as an update. The leak seems less confident in its performance estimate for the GeForce RTX 4080, listing it with a question mark. The estimate is 420W, which is not that far from the GeForce RTX 4090. Due to the huge difference in the allocation of functional units, the GeForce RTX 4080 may be clocked much higher than its larger sibling, so that they are not too far outside the final performance.
NVIDIA GeForce RTX 4070 new specifications will be
The model of this card is no longer expected with 12GB of memory, but 10GB. This surely indicates the memory with the narrower bus from 192-bit to 160-bit. The RTX 4070 is rumored to feature 7168 CUDA cores, which would mean 512 disabled CUDA cores on the AD104 GPU. It is expected to be a shortened configuration of the AD104-275 GPU with a slightly higher core count than the RTX 3070 TI (6144 cores) with 7168 cores or 56 SM. The GPU will come with 48 MB of L2 cache and up to 160 ROPs. Taking rumored memory specs, there’s a chance that the GeForce RTX 4070 is expected to rock 10GB of GDDR6 capacity, which is said to be clocked at 18Gbps over a 160-bit bus interface. The card can reach a TBP of 300W and the leak also mentions that the price of this card would not be lower than RTX 3070 or 3070 Ti graphics cards. Collectively, the rumor mill predicts that 10GB of GDDR6 memory at 18Gbps on a 160-bit bus for the RTX 4070 is another short-memory bandwidth allocation that puts it on par with the RTX 3060. That seems doubtful to us, but again, Ada Lovelace is a drastic increase in GPU cache compared to Ampere. Maybe that makes the difference.