Rocket Lake has been a bit of a mystery at any time given that we first heard about the chip. There have been questions about no matter if it was a real 14nm backport of a 10nm style, no matter if it would keep the complete evaluate of clock pace Intel squeezed out of 14nm, and no matter if it will contend efficiently from AMD’s Zen 3 architecture. We cannot response that very last issue right up until the chips launch, but fortunately we can acquire a crack at the first two.
To start with of all, Rocket Lake isn’t centered on Willow Cove, the bigger-frequency CPU that Intel debuted earlier this calendar year. Rocket Lake’s CPU core, codenamed Cypress Cove, is centered on Sunny Cove, the CPU that powers its Ice Lake line of processors. This isn’t automatically a issue Intel was attempting to take care of difficulties with scaling to significant frequencies when it designed Willow Cove. Due to the fact it’s centered on an older 14nm method, Sunny Cove should not experience the exact declines.
Meet up with Cypress Cove
Cypress Cove is a backport of Sunny Cove, aka Ice Lake. Intel promises it is on the lookout for double-digit IPC scaling, which we presume refers to IPC scaling without the need of counting clock pace enhancements about earlier chips. Benefits of Cypress Cove include things like:
New support for AVX-512 baked into Intel desktop CPUs.
Integration of Intel’s new Xe graphics core
Help for up to 20 PCIe 4. lanes
The Xe-LP graphics are only shown as staying 1.5x speedier than Gen9 integrated performance. This states absolutely nothing superior about the real onboard GPU designed into the desktop versions of these chips.
To set this in perspective, Tiger Lake’s Xe remedy is ~2x speedier than Ice Lake’s GPU. As this graph from Anandtech demonstrates, Ice Lake’s Gen11 GPU was significantly speedier than any Gen9 remedy Intel at any time shipped. There are minimal figures, but they array from a 2x advantage to a 1.27x advantage.
If Intel was putting a complete Xe configuration on its die, we should really be capable to hope a little something like a 2.54 enhance in raw GPU horsepower. TGL laptops can have up to 96 EUs (execution units). We know that Xe is going to be significantly speedier than Gen 9 many thanks to performance enhancements, but it seems like Intel is going to use in between 24 EUs and 32 EUs in its desktop chips. If it made use of 32 EUs as a substitute of the 24 it experienced fielded beforehand, this would stand for a 1.33x performance advancement on width by yourself. We know, having said that, that Xe will make some important modifications to Intel’s in general GPU style, which is probably where by the rest of the enhanced performance is coming from. If Intel was working with a 96 EU configuration in Rocket Lake, we’d hope performance to be 2-2.5x speedier than Gen 9, not just 50 per cent.
I experienced floated the notion earlier this calendar year that Intel could possibly contend from AMD by positioning its have CPUs at price factors where by the designed-in Xe graphics could carve into markets where by people today want far better graphics on a price range. As an included reward, environment up a comparison from an AMD APU with a 64-96 EU Rocket Lake would probably have gone well for Intel, provided how Tiger Lake compares from the Ryzen 4000 sequence.
But if Intel is sticking to 32 EUs on desktop, which is not going to come about. Gen9 basically was not speedy ample for a 1.5x performance enhance to get you significantly in the way of handy gaming performance. If you commence at 30fps, a 1.5x enhance receives you to 45fps — a obvious gain although actively playing. If you commence at 15fps and choose up the exact +50 per cent, you’ve hit 22.5 frames for every next — well under the playable threshold for the vast majority of titles.
Intel, apparently, continue to sights its integrated graphics as worthless to its desktop line, which is deeply unfortunate. Corporations like Topaz Video Greatly enhance AI are actively doing the job to boost Intel GPU support, and a more sturdy iGPU in desktop would have assisted this a terrific offer.
Also, unlike Comet Lake, Rocket Lake will drop again to just eight cores. How significantly of an affect this will have on the products continues to be to be noticed. Consider, for illustration, that Intel hits the ~1.16 IPC uplift for Ice Lake and keeps each bit of its frequency at 14nm — or perhaps even hits slightly bigger clocks, given that it does not have to electrical power 10 chips in the exact die. That would be a 1.16xc IPC bounce combined with, say, a 5 per cent gain in real clock speeds, or about 1.22x uplift in general. An eight-core chip with a 1.22x performance uplift would be just slightly slower (theoretically) than a 10-core chip that lacked these pros.
A couple other notes: Intel does not rule out releasing chips without the need of integrated graphics, comparable to its present-day “KF” CPUs, and it states that the x4 PCIe lane off the CPU is “only validated for discrete graphics, storage, or Intel Optane.” Not guaranteed if any individual is possible to build a PCIe 4. audio card any time quickly, so this is unlikely to be an situation. There is also backward compatibility with the existing 400-sequence chipsets, with a new 500 sequence chipset coming as well, with unspecified features.
Intel’s single-thread performance bounce would set it nearer to par with AMD, but its multi-threaded performance could possibly not get significantly far better. Matching the performance of a 10-core with an eight-core is intrinsically impressive, but cleanly exceeding the performance of your 10-core with an eight-core in well-threaded code will require more than just a 1.16x IPC improve combined with a 5 per cent frequency gain. A 10 per cent advancement in sustained clocks combined with 1.16x IPC would do it, but it’s not apparent if Intel’s 14nm method has any headroom left.
Having AVX-512 into desktops is crucial to encouraging developers to use it, but I imagine the most vital narrative all around Rocket Lake for Intel is that it’s no for a longer period recycling an endless succession “14nm++++.” While these new chips are continue to designed on 14nm, they will introduce real, distinctive enhancements to the Intel ecosystem, with everything from PCIe 4. to Xe support in an iGPU.
I’m not guaranteed that’ll be ample — but it’s absolutely superb to see real levels of competition in the CPU market all over again.
Now Go through: