It looks like Apple may have solved a multi-GPU problem that has been troubling AMD and Nvidia for years thanks to the UltraFusion interface it announced earlier this week, which allows two M1 Max chips to come together to produce a single M1 Ultra chip.
When Apple launched the M1 Max last year, it did it underhandedly, not revealing that the processor had a hidden secret. This secret was exposed at Apple’s March Event earlier this week when the firm revealed that it can combine two M1 Max chips to create an M1 Ultra with double the power — including twice as many GPU cores.
Apple isn’t the first corporation to use the combined power of two GPUs. Nvidia graphics cards have supported SLI (Scalable Link Interface) for many years, which is a high-speed link that allows you to connect multiple GPUs at once for greater performance. AMD had a comparable technology called CrossFire that accomplished the same thing. Whether you’re a casual gamer who plays rainbow riches or a hardcore gamer this is exciting news.
Resolving the issue
SLI and CrossFire have the problem of not being able to double the power of the PC by the number of GPUs installed. So, if you had two Nvidia GPUs in SLI, you didn’t receive twice the performance.
The best-case situation would be a 90% improvement, but in many cases, it was more like a 50% improvement — and performance degraded as you added additional GPUs. For many people, the return on investment was simply not worth it, especially if you paid for three GPUs but only got around the power of 2.5 GPUs.
SLI and CrossFire both introduced overheads to the PC, particularly with the CPU, affecting performance. The amount of space required inside a PC chassis and motherboard to handle multiple GPUs increased, as did power consumption, and hence running expenses and cooling considerations.
Finally, there was also an issue with apps, and especially games, giving poor – or non-existent – support for SLI and CrossFire. In some circumstances, a game would only use a single GPU, no matter how many you had installed.
As a result, it’s no wonder that CrossFire and SLI have become extremely niche features, with neither AMD nor Nvidia mentioning them. In fact, Nvidia effectively killed SLI for its RTX GPUs a few years ago in favor of NVLink. However, because of the difficulty of purchasing a single Nvidia GPU, let alone numerous GPUs, this has remained unpopular.
Apple looks to have addressed many of these concerns with the M1 Ultra. For starters, the UltraFusion link between the two M1 Max chips has been designed to have exceptionally low latency (basically a delay in the transfer of data, the lower the better) – with a bandwidth of 2.5TB/s. This is insanely fast, far faster than SLI or CrossFire ever could.
The speed of UltraFusion has multiple advantages, the most obvious of which is performance, which is why Apple is confident in asserting that the M1 Ultra will perform twice as well as the M1 Max.
Apple has also avoided the performance overheads associated with multi-GPU configurations by combining two M1 Max chips into a single M1 Ultra chip. Despite the fact that the M1 Ultra is now Apple’s largest chip, it is still smaller than physical discrete graphics cards, allowing it to be used within the compact Mac Studio, which was also revealed at Tuesday’s event.
The M1 family of chips has also been praised for their performance per watt, which is substantially higher than that of competitors. This has allowed MacBooks with M1 CPUs to last longer on a single charge while remaining cool even when working hard, and the M1 Ultra maintains this trend. The M1 Ultra’s GPU, according to Apple, outperforms Nvidia’s formidable RTX 3090 while consuming 200W less power.
That’s a bold statement, but in these days of escalating energy expenditures, the power efficiency of our components will become increasingly crucial. According to rumors, Nvidia’s next generation of GPUs, such as the RTX 4080, would be even more power-hungry, making Apple’s approach much more enticing.
Finally, Apple’s strategy appears to be paying off in terms of software support as well. Apps will recognize the M1 Ultra as a single chip, requiring no additional code — support will be available right out of the box, allowing applications to take advantage of the increased processing capability. This is a significant distinction that could prove to be game-changing.
Speaking of which, while Macs aren’t generally thought of as games machines, games should also recognize the M1 Ultra as a single chip, allowing them to utilize the extra power. Could Apple have produced a superb gaming GPU with the M1 Ultra, which appears to outperform the RTX 3090?
We’ll have to wait till Apple’s claims are put to the test, but it sounds good. Apple’s victories may encourage Nvidia and AMD to reconsider their multi-GPU technology, resulting in even better performance for both gaming and professional use. We can’t wait to get started.
