Currently, most of the leaks surrounding Nvidia graphics cards are primarily concerned with the RTX 5090 and perhaps the underlying RTX 5080, but amidst the mountains of information, the RTX 5070 also keeps popping up.
So that you can keep an overview, we summarize belowall the rumors and information about the RTX 5070 together and classify them.
Quick selection
RTX 5070 price: (Almost) everyone is in the dark
Nothing is known for sure, but currently you have to expect at least 600 euros, in the worst case 700 euros.
As of now, only one leaker has onecalled: the YouTuber “Moore's Law is Dead” (MLID). However, this is blessed with an average hit rate for its preliminary information, so you should treat the following details with caution.
- As a result, the RTX 5070 will come onto the market in the USA with an MSRP between $599 and $699. In the case of a blunt conversion, the euro price should be between 550 and 650 euros at the current exchange rate.
- As a rule, however, Nvidia does not carry out such direct conversions in Euro countries - simply because of the VAT, which is included in the purchase price in this country. If the MLID leak actually turns out to be true, you have to expect at least 600 euros, in the worst case scenario 700 euros.
- The RTX 4070 was also within this price range when it was introduced: At that time, the Nvidia GPU went on sale for 659 euros.
RTX 5070 Release: Unveiling at CES 2025?
There have been rumors for months that Nvidia will present new graphics cards at CES 2025 - including the RTX 5070. A release could follow in February 2025.
- The Consumer Electronics Show traditionally marks the starting point of the technology year and is used by numerous manufacturers and companies to get in the mood for upcoming products.
- CES 2025 will take place from January 7th to 10th, 2025 in Las Vegas (USA). the day before, which takes place at 3:30 a.m. German time.
However, an unveiling of the RTX 5070 would not necessarily result in a subsequent release. Instead, Nvidia will probably stick to the usual cycle and launch the RTX 5090 first and then the RTX 5080.
A somewhat more concrete schedule was leaked in mid-October. According to the rumor, Nvidia should leave significantly less time between individual GPU releases, so that.
RTX 5070 Specs: The chip as the driver
The RTX 5070 chip has been known for a few months, at least according to leaks. While the RTX 4070, nominally its direct predecessor, was based on the AD104 chip, Nvidia is probably going one level lower for the upcoming mid-range GPU and using itGB205-Chip.
According to the leaker kopite7kimi, who is known for his excellent hit rate, it is already clear howwill look like.Five GPCs(Graphics Processing Cluster) with eachfive TPCs(Texture Processing Clusters) should be. Thanks to the hierarchical structure of a GPU chip, the number of CUDA cores can also be traced:
- From the above information it follows that with the full expansion of the GB205 chip25 TPCsare installed.
- Two streaming multiprocessors (SM) are included per TPC - that's it50 SMs.
- Included in an SM 128 CUDA cores, giving the GB205 chip 6,400 shader units.
However, the last line of this calculation assumes that Nvidia will stick with 128 CUDA cores per SM, which has not yet been confirmed.
This is also a full expansion of the GB205 chip. It also remains to be seen whether this will also be used in the RTX 5070 or whether Nvidia will partially deactivate it. The big leak at the beginning of October at least did not specify a core number for the RTX 5070 –.
RTX 5070 Specs: Probably 12 GB of GDDR7 memory
However, the huge leak from the wccftech portal revealed other information that didn't make you happy either. That's how it should beset, which is of the GDDR7 type with 28 Gbps, which gives us a bandwidth of 672 GB/s.
At least in view of the expected memory interface, this information makes sense: the GB205 chip has been available since the summer192-bit wide interfaceassigned, which stubbornly maintains itself as "confirmed" in every leak.
- Since the data bus of a single memory module is 32 bits, only a maximum of six modules can be installed in such an interface - unlikely imponderables like one»Clamshell-Modus«sometimes left out.
- Since the GDDR7 chips themselves are still only available in a size of 16 GB (or 2 GB), this follows with six modulesMaximum possible storage amount of 12 GB.
Diewill most likely come too late for the RTX 5070. With these, up to 18 GB of VRAM would have been possible with the same interface.
RTX 5070 Specs: TDP, Ports and More
Power connection
What is certain is that the RTX 5070 is on the12V 2x6 connectorfor the power supply. This was done with theNvidia has made it mandatory for upcoming GPUs, so a sudden change is unlikely to occur.
Compared to the 12VHPWR connector, which made its debut with the release of the RTX 4090 and made the headlines due to melting parts, only one aspect changes here on a technical level.
- The sensor pins are shortened here, while the power pins are lengthened. This ensures that the 12V lines are in front of the sensor pins.
- The advantage: If the plug is not connected correctly - which is probably the casewas – the sense pins are also not connected to the socket. The permitted power supply is accordingly denied so that possible overheating problems cannot occur.
Power consumption
The expected power consumption is quickly told and classified:250 WattAccording to the big wccftech leak, it should be with the RTX 5070. kopite7kimi also mentions this number.
- This means that the TDP is slightly higher than the RTX 4070 (200 watts); However, gigantic jumps like the RTX 5090, which is supposed to go from 450 to an absurd 600 watts, are not expected.
- However, this increase in required performance is unlikely to come from the memory: GDDR7 promises improved energy efficiency by up to 33 percent.
All that remains are the remaining technical specifications, but information about clock speeds or number of fans or other possible “causes” of the increased TDP are still completely missing.
