The 12GB version of the GeForce RTX 4080 looks like the renamed GeForce RTX 4070

  • AD102 – GeForce RTX 4090
  • AD103 – GeForce RTX 4080
  • AD104 – GeForce RTX 4070
  • AD106 – GeForce RTX 4060

The GPU-card binding will not be as simple as expected so far (above), as this time Nvidia has also done (similarly to before the release Ampere) some last minute changes. In addition to the expected 16GB GeForce RTX 4080, there will also be a 12GB version (which we already reported), but this is far from being the only model with a lower memory capacity.

GeForce RTX 4080 12GB with smaller AD104 GPU

The GeForce RTX 4080 12GB is not built on the AD103 GPU (~10,000 stream processors, 256-bit bus), but on the lower GPU AD104 (<8,000 stream processors, 192-bit bus) and its parameters are exactly what was expected from GeForce RTX 4070 / RTX 4070 Ti. Some leakers see this renaming as an attempt to give the "GeForce RTX 4080" an official price tag that won't be higher than the current generation. Of course, it's not just a difference in price, lower parameters will also mean lower performance. Looking at the specs, it looks like the full GeForce RTX 4080 (16GB) will have ~22% more arithmetic and texturing performance and even 46% more memory throughput. It can therefore be estimated that ~30% higher game performance compared to the 12GB version.

It is somewhat reminiscent of the 3GB / 6GB GeForce GTX 1060, which also differed not only in memory capacity, but also in the number of active functional units (stream processors, texturing units) and thus also in performance. Opinions that defended such a solution with the argument “everyone expects lower performance anyway with a card with a lower memory capacity, so it’s OK” probably did not realize that if such a step passes once, sooner or later it will come again, but on an even larger scale. The fact is that a card that looks like a lower memory variant is supposed to be a lower memory card and not a model with completely different specifications.

GeForce
RTX 4070
GeForce
RTX 4080
12GB
GeForce
RTX 4080
16GB
GeForce
RTX 4080 Ti
GeForce
RTX 4090
(?) Titan /
RTX 4090 Ti
GPUAD104
~30 billion tr.
[7680 SP]
AD103
[10752 SP]
AD102
>75 mld. tr.
[18432 SP]
flat??~600 mm²
trial4nm TSMC
rate?2310 MHz??2235 MHz?
boost?2610 MHz2505 MHz?2520 MHz>2750 MHz(?)
stream
proc.
7168?768097281484816128
16384
18432
18176
TMU224240304464512568
ROP??????
FP32
TFLOPS
?~40~49>7082~100
bus160bit?192bit256bit320bit384bit384bit
memory

?

21 GHz
GDDR6X

23 GHz
GDDR6X

21 GHz
GDDR6X

21 GHz
GDDR6X

24 GHz
GDDR6X

capacity10 GB12 GB16 GB20 GB24 GB48 GB
data?504 GB/s736 GB/s840 GB/s1,01 TB/s1,15 TB/s
power supply?16pin16pin16pin16pin2× 16pin(?)
TDP?285-366W340-516W~400W?450-660W600-800W?
datum
edition
?listop. ’22listop. ’22?October ’22?

How it will look with the GeForce RTX 4070 remains unclear. It will most likely be based on a cut-down AD104 core and will not differ much from the GeForce RTX 4080 12GB (7680→7168 stream-processors, 192→160bit bus). After a generation when the base model GeForce RTX 3080 was created on the largest core of the given generation (GA102) is the next generation, when the basic GeForce RTX 4080 model will be equipped with up to the third largest core (AD104) of the given generation.

Two TDP values

Surely you remember alternating information leaks about TDP values. There was always some fairly extreme value followed by a somewhat milder number, after some time a high number again and a milder reading. The reason for this duality seems to be clarified. The lower value will be the default TDP of the card, the goal of which is to maintain compatibility (in short, so that, for example, the owner of a set with a GeForce RTX 2080 Ti can exchange the card for a GeForce RTX 4080 12GB and not have to change the power supply and case for this), and the higher value is activated by a switch on the card. It is not yet clear whether all cards will be designed this way, or whether there will also be models supporting only the basic TDP. There are known values:

  • GeForce RTX 4080 12GB: 285→366 watts
  • GeForce RTX 4080 16GB: 340→516 watts
  • GeForce RTX 4090: 450→600 watts.

The information that the GeForce RTX 4090 is supposed to have a PCB rated at 600 watts, which some editors and some users considered nonsense, is thus confirmed.

The GeForce RTX 4090 is currently on track for the first half of October, (both) GeForce RTX 4080 should arrive a month later, in the first half of November.

It can be expected that the parameters of these three cards are definitive and will no longer change. Which was also hinted at by leaker kopite7kimi, who stated that his work is basically done (in other words, that he won’t have anything more to say about these cards).


Source: Diit.cz by diit.cz.

*The article has been translated based on the content of Diit.cz by diit.cz. If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!

*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.

*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!