NVIDIA Will Show Off Next-Generation GeForce GPU At Hot Chips 30
NVIDIA has just listed a session detailing a ‘Next Generation Mainstream GPU’ for Hot Chips 30 (via Videocardz). This is the first official confirmation of the intended timeline for the next generation of NVIDIA GeForce GPUs. It also means that there is a possibility we might be seeeing sneak peaks of the cards before Hot Chips 30, which is almost 3 months away. The NVIDIA presentation in question will take place on August 20th.
NVIDIA’s Next-Generation GeForce GPU will be detailed at Hot Chips 30
This is exciting stuff because it means that the company is preparing for a brand new series of cards soon. We have been hearing leaks of everything from Volta, to Turing, to Ampere and it is unclear at this point what the final series will end up being called. Even nomenclature is up for grabs with rumors swirling of both 1100 series and 2000 series cards. Before we go any further, here is the schedule in question:
Stuart Oberman will be taking the stage to talk about NVIDIA’s next-gen mainstream GPU (whatever it ends up being called). Unfortunately, NVIDIA did not slip up this time and refrained from using the codename for scheudling purposes. Hot Chips GPU sessions usually happen after a soft reveal so it makes us feel that we are probably going to be seeing this GPU before the actual session. A reveal set for Hot Chips isn’t NVIDIA’s usual style.
NVIDIA also usually schedules launches to take place immediately before a financial quarter. So either of 2 things will happen: 1) we see a launch happen sooner rather than later before the 3rd quarter, or 2) we see a launch around the time of Hot Chips before the 4th quarter. Either way you slice it, looks like gamers will have a new series to play with before the year is over. The timeine for beginning of Q4 makes the most sense to us and its worth noting that this is also the time frame around which GDDR6 is expected to enter into mass production.
What we know about Ampere and Turing so far
According to Beyond3D, the company has been preparing the Ampere architecture mainstream champion, the GA104, as the perfect replacement for high end Pascal. The second quarter of 2018 would mark exactly 2 years since the GP104 was unveiled back in 2016. To keep its two-year cycle alive, the company has to upgrade to the next iteration during 1H 2018.
According to Reuters, A brand new NVIDIA Turing GPU is launching at GTC 2018. This means that either NVIDIA is undergoing a massive rebranding campaign or that Volta will not be making it to the consumer side of things. This in itself has interesting implications, because it could mean a couple of things: anywhere from NVIDIA optimizing the architecture further to yield concerns.
In any case, before we go further, a short intro to Turing. The codename Turing of course refers to Alan Turing who is considered the father of modern computing and was the chief architect of the British program to break Enigma machine encryption, used by Germany to encrypt communications during World War II. He is also the person behind the concept of the Turing machine, a hypothetical device that can simulate any algorithm regardless of how complex it is. Programming languages that can simulate this device are referred to as Turing complete languages.
Now there are two possibilities as I see them. 1) NVIDIA decided to ditch the Ampere codename because it’s very similar to another company – an ARM server maker called Ampere – or 2) it decided to fork the upcoming architecture into two distinct flavors, namely Ampere and Turing. Both possibilities are just as likely as NVIDIA has been known to do that in the past simply to discredit the leak scene (anyone remember the GeForce GTX 800 series?).
The second possibility however is much more exciting, it could mean that NVIDIA is preparing two distinct products in the consumer space: Ampere and Turing. Since there have been rumors floating of a cryptocurrency focused product from the company soon, then it could mean that the Turing GPU is actually the one NVIDIA will market as the mining variant. This makes a semblance of sense, because Alan Turing is well known for his work on cryptography. In fact he was part of the team that cracked the Enigma machine which resulted in the World War being won.
If Alan Turing’s cryptographic roots are indeed being honored in this new product then it could very well mean that NVIDIA is becoming very serious about cryptocurrencies. The company has already given its customers free reign to use its GeForce GPUs in data center environments as long as they are used for ‘blockchain mining’ purposes so this wouldn’t be entirely out of character. Of course there remains the remote possibility that Reuters is simply wrong about this and there is only one GPU which will in fact be called Ampere after all.
Rumors also suggest that the NVIDIA GTX 1180 graphics card will be manufactured on TSMC’s 12nm FinFET process which will introduce significant power efficiency upgrades on a core to core basis. The exact core count will be 3584 CUDA cores divided in 28 SMs. 64 ROPs and 224 TMUs make up the rest of the specifications. According to the same entry, the memory in question will be the GDDR6 variant with up to 16 GB worth of DRAM.
NVIDIA GeForce GTX 1180 Specifications
|Wccftech||GeForce GTX 1180||GTX 1080|
|Lithography||12nm FinFET||16nm FinFET|
|Core Clock||1405/1582 MHz||1607MHz|
|Boost Clock||~1800MHz (Expected)||1733MHz|
|Peak FP32 Performance||~13 TFLOPS||8.7 TFLOPS|
|Memory||16 GB GDDR6||8GB GDDR5X|
|Launch||Q3 (July) 2018||July 20 2016|