On May 14, Nvidia will make up for an announcement that is important for the graphics card world.
In March, Nvidia boss Jensen Huang should have climbed the stage at the in-house GPU Technology Conference (GTC) and talked about Nvidia’s next chip architecture for new RTX graphics cards . Then Corona happened.
His speech is announced in pithy words: One should be curious about “the latest platform breakthroughs in AI, deep learning, autonomous cars and professional graphics”.
Nvidia uses the phrase “Get Amped” as a voltage generator – this can be understood as a reference to the upcoming GPU architecture Ampere .
What is new about amps?
It is almost certain that new ampere chips will be produced in the more powerful 7nm manufacturing process .
More transistors on the chip are said to bring up to 50 percent more power than the current RTX cards. Professional GPUs for AI training should even offer up to 75 percent more performance in this area thanks to special deep learning optimizations.
VR users can look forward to the future with the special connection “Virtual Link” for VR glasses of the next generation. Nvidia developed it together with partners such as Oculus, HTC and Valve .
After Valve gave up the cable standard for Valve Index (test) due to technical problems and by no means all of the RTX cards support Virtual Link , Virtual Link could be logged out again. We are seeing Ampere ; Price: RTX 3080 likely to touch around $690-$750.
When will Nvidia show the RTX-3000 series?
At GTC, Nvidia actually shows new architectures and fundamental technological innovations rather than new products. The cards will then be unveiled at special gaming events or AI conferences. However, they will probably fall largely flat this year.
Recently there were rumors that Nvidia could wait to present the graphics card – supposedly also to wait and see how AMD moves in the market. Since Corona is currently slowing the world down, one should not hope for a timely presentation or even publication of the next RTX cards.
Amp: Nvidia’s Next-Gen Graphics card is said to be up to 75 % faster
“Big Red 200” is a powerful supercomputer that the University of Indiana wants to use for complex scientific calculations, especially in AI training . It supports both high-performance computing and AI-specific computing processes .
The supercomputer is built in two phases : The first phase is already installed and runs on the basis of 672 AMD processors (“Rome” Epyc 7742).
The more interesting phase is the second, which should start in the summer: Next-gen graphics cards from Nvidia are then added to the system. Brad Wheeler, chief technology officer at Indiana University, confirms this to NextPlatform .
First Nvidia Ampere cards for supercomputers available in summer?
The actual plan was, according to Wheeler, to equip the computer with already available Nvidia V100 Tesla GPUs. The system would then have achieved a total output of around 5.9 petaflops.
At the last moment, according to Wheeler, “luckily” the opportunity arose to wait a little longer and instead use Nvidia’s next generation of graphics cards in Big Red 200.
The university decided to wait: The new cards should offer up to 75 percent more power than the current generation. The finished supercomputer is said to offer a total output of around eight petaflops.
Wheeler doesn’t say which Nvidia graphics card his university bought. A new Tesla GPU based on the ampere architecture would make sense .
In the official announcements for Big Red 200 from the University of Indiana and from Nvidia , 256 “Tensor Core GPUs” are mentioned, which are built into the supercomputer. The computing units developed for AI are installed in particularly large numbers on Tesla GPUs.
According to Wheeler, Big Red 200 will be “the fastest AI supercomputer owned by a university” . The performance of the computer should exceed that of its predecessor, Big Red 2, by a factor of eight. According to Wheeler, the leap in performance should be “much bigger” especially for artificial intelligence .
Nvidia Ampere: March performance?
March could bring certainty to Nvidia’s ampere strategy: Then the company is supposed to reveal details about the new graphics chip at the in-house event “GPU Technology Conference” (GTC 2020) in San José (California, USA).
However, this does not necessarily mean that Nvidia is introducing new gaming cards, for example an RTX-3000 series .
It is known that Nvidia will change the manufacturing process for amps to 7nm , which enables more transistors on the chip and thus more power with lower power consumption.
Almost 30 percent better: AMD beats NVIDIA in VR benchmark
In the OpenVR benchmark, an unknown AMD GPU clearly outperforms NVIDIA’s RTX 2080 Ti. Is it a Big Navi?
NVIDIA or AMD: This is still a controversial issue, on which religious wars are kindled in comment columns with regularity. The days when NVIDIA was able to shine through better performance and stability are at least partially over.
While in the past the first question in the case of PC problems or game crashes was mostly “AMD graphics card?”, The manufacturer has caught up considerably in recent years and, at least in the medium price-performance sector, provides serious alternatives to the NVIDIA GPUs with its hardware.
Ampere or Big Navi: Who will win the race in 2022?
Both companies are of course eagerly awaiting new hardware. Most recently, NVIDIA delivered current flagships with the RTX graphics cards (based on Turing GPUs). The next generation of GPUs, called amperes, is to come in 2022.
The Ampere graphics cards are loud rumors first bring up to 50 percent better performance than RTX graphics cards . At the same time, electricity consumption is to be halved . These are pretty good guidelines that AMD has to imitate first.
The term “Big Navi” has been around AMD for a long time. The new graphics chip is said to bring significant progress over the Navi 10 chip currently installed in the Radeon RX 5700 XT, for example. AMDs CEO Dr. Lisa Su teased “Big Navi” with RDNA 2 architecture , ray tracing support at hardware level, improved energy efficiency and more.
But who will win the race?
Unknown AMD GPU leaves NVIDIA far behind
Naturally there is always wild speculation on the Internet. They are currently fed by a rather blatant VR benchmark , in which an unknown AMD GPU beats a possibly overclocked RTX 2080 Ti from NVIDIA by a whopping 17 percent .
Unknown AMD GPU beats NVIDIA’s RTX 2080 Ti by a whopping 17%. PICTURE: Reddit / Imgur
The OpenVR benchmark used (Stonebrick Studios) was only published a few weeks ago and combines performance comparisons for graphics cards with the connected VR glasses . The OpenVR benchmark tool is available for free on Steam . There is an entry in the leaderboard by purchasing additional content.
A very interesting entry now appears in this list. An untitled “AMD Radeon ™ Graphics” GPU, in conjunction with an HTC Vive (@ 90Hz) and a resolution of 1,512 x 1,680 pixels, delivers a whopping 103.32 points. This is an increase of 17 percent compared to the comparable entry of an RTX 2080 Ti (in combination with a strong i9-9900K CPU).
Buggy benchmark or is the hype justified?
The CPU used in the AMD benchmark in question, however, puzzles. Apparently it is sample hardware from AMD (Engineerring Sample). A DxDiag entry in the Turkish forum technopat.net shows that the code given in the benchmark refers to a CPU in a gaming notebook from ASUS (Asus Zephyrus GA401IV). It is obviously a Ryzen 7 4800HS.
But why should someone use a notebook for such a benchmark? Has a strange combination of AMD GPU and the NVIDIA RTX 2060 built into the Zephyrus GA14 series been used here? Criticism of the still young benchmark quickly arose . Maybe this value was only caused by bugs? Is it a fake picture?
OpenVR developer: Most likely a new AMD GPU
The developer of OpenVR, CyubeVR, commented on this benchmark in the subreddit and explained that this benchmark really does exist in the leaderboard . A Photoshop fake is therefore excluded. He also argues that in the meanwhile over 2,000 benchmarks that have so far been carried out with OpenVR Benchmark, any bugs would have been noticed and then also fixed.
He also clarifies that the benchmark is a flawless GPU test and that the CPU has hardly any influence on it: “The OpenVR benchmark is a GPU benchmark, so limitation by the CPU is practically impossible. The 4800H processor would definitely be fast enough for a graphics card that is 30 percent faster than a 2080 Ti. ”
He definitely considers the use of an RTX 2060 to be out of the question, the performance is too good for that . So is the benchmark the “Big Navi” GPU from AMD?
The values of the mysterious AMD GPU are almost 30% better than regular RTX 2080 Ti
OpenVR developer CyubeVR has to consider that the 17 percent beaten RTX 2080 Ti is probably a “really good 2080 Ti”: “The majority of the 2080 Ti actually achieves around 80 frames per second in this benchmark, so the unnamed GPU we’re talking about is about 29 percent faster than a regular 2080 Ti, while it’s 17 percent faster than the highest overclocked 2080 Ti. ”
A lead of almost 30 percent would mean golden times for AMD if NVIDIA is unable to counter with its own amp GPUs . In any case, a strong competitive situation is something good for the users, after all, this should have a positive effect on the sometimes extreme prices in the graphics card segment. And in turn, computing-intensive VR worlds could particularly benefit from this.
2022 could be a very interesting year for PC hardware. Maybe NVIDIA will even dig out the VR cable standard Virtual Link in the course of the big Half Alyx hype ? However, this is not particularly likely, after all Valve of all days gave a harsh rejection .
You should also consider: Benchmarks are a good indicator of the performance of a graphics card , but they only reflect part of the actual performance.
The drivers mentioned above and especially the VR implementation are also crucial. Nvidia can be used for an innovation here and there, for example recently introduced “Variable Rate Shading”, which can provide VR with significantly better image quality with the same computing effort . In comparison, AMD still has some catching up to do.
EVGA 08G-P4-3081-KR, GeForce RTX 2080 Super Black Gaming, 8GB GDDR6
Author Profile
- Amram is a technical analyst and partner at DFI Club Research, a high-tech research and advisory firm .He has over 10 years of technical and business experience with leading high-tech companies including Huawei,Nokia,Ericsson on ICT, Semiconductor, Microelectronics Systems and embedded systems.Amram focuses on the business critical points where new technologies drive innovations.
