Home / News / PC Gaming News / AMD Zen Performance: everything you need to know

AMD Zen Performance: everything you need to know

by

AMD Zen Lisa Su

AMD has finally run some benchmarks on its top-end Zen processor, and things are looking good.

Don’t call it a comeback. Actually, do. “AMD is back” is the internal catchphrase the company is using to describe its new CPUs and based on the trickling of information we’re getting about the new Zen architecture, it might just turn out to be accurate.

It’s hard to overstate how important Zen is for AMD. The new processor architecture has been in the works for several years and, if all goes to plan, should finally let AMD challenge Intel’s Core i line-up of desktop processors, which have enjoyed absolute dominance in the desktop PC market over the last half decade.

What is AMD Zen?

AMD has long lagged behind Intel when it comes to performance per clock. For years an AMD core would have slower performance at the same clock speed compared to an Intel core. As a result, in previous AMD generations you’ve needed higher clock speeds and huge power consumption to get performance on par with Intel. AMD hopes that will change with Zen.

Related: Best gaming PC specs to build yourself

AMD reckons the performance of its top-end chip will be 40% greater than its previous high-end processor, but this is hardly surprising considering how long it’s been since the company refreshed its line-up. AMD Zen Computex 2

This is partly down to the new 14nm process, which means more transistors can fit onto a given piece of silicon, resulting in improved performance without a big increase in power consumption.

Importantly, Zen chips will support “Simultaneous Multi-Threading”, similar to Intel’s Hyper-Threading tech. This allows for better distribution and handling of multiple tasks and means this particular chip will have 16 threads.

Things have been fairly slow going on the announcement front for AMD Zen. The company has started making some slightly more specific performance claims beyond its initial 40% figure, but we still don’t know exactly what sort of chip we’re looking at.

AMD Zen Performance

AMD’s latest claims surround Zen's raw instructions-per-clock (IPC) performance versus a comparable Intel product. The claim stemmed from a speed test conducted live on stage at a partner-only AMD event yesterday. The test pitted an 8-core, 16-thread, Zen-based “Summit Ridge” chip (which will likely equate to a new FX-branded processor in the future) against an Intel Broadwell-E-based enthusiast chip with the same number of cores that had been set it to the same clock speed.

AMD benchmarked the two chips in a Blender rendering task and, surprise surprise, the AMD chip won. You can watch this happen in the video below. It's a close-run thing, but the Zen-based chip just sneaked ahead.

AMD says the Zen architecture won’t just be limited to enthusiast desktop chips. Laptops and servers also expected to benefit from Zen in the future. Pricing will be a thornier issue: the performance of these chips means AMD will be entering a price bracket it’s never been in before, so don’t expect a bargain.

Related: Best desktop PCs from £500

Zen architecture will also make it to AMD APUs, meaning you’ll be able to buy a full chip with both Zen architecture and AMD Radeon graphics on board. These have proven relatively successful for budget gaming PCs in the past, so with better processing performance and graphics, AMD could really take a bite out of the budget gaming market

What’s most interesting is the rumoured power efficiency of the top-end 8-core chip, which is rumoured to be just 95W. That compares very well the 140W 8-core Intel core i7-6900K, but it remains to be seen how the two compare in terms of performance when both are at maximum clock speeds.

AMD Zen-based chips will run on the company’s new AM4 architecture, which will feature DDR4 memory-compatibility, USB 3.1, high-speed NVMe SSD support and SATA Express, putting it mostly in line with Intel’s X99 chipset.

NemEu

August 20, 2016, 6:08 pm

a single and 'chose' test isn't much trustable, let's see the end user general benchmark tests

comments powered by Disqus