Intel’s Alder Lake line of CPUs will hit store shelves on November 4, and the chip giant is sharing some additional data about specs and pricing ahead of its debut next week. We’ve written several deep dives on Alder Lake already this year, but to recap:
Alder Lake is Intel’s next-generation desktop chip. It’s built on a refined version of Intel’s 10nm node (now rebranded as Intel 7), and it’s the first hybrid x86 architecture. It’s called a ‘hybrid’ architecture because it includes a mixture of larger and smaller cores. Anecdotally these are typically known as “big” and “little” cores, though Intel’s claimed performance targets for the Efficiency cores implies that “big” and “bigger” might be a better way to describe the relationship between the two core clusters.
Features and Performance Claims
Intel is promising a 1.19x overall uplift for Alder Lake when IPC and frequency are taken into account. The overall IPC uplift at identical frequency is ~1.15x. In gaming, Intel showed a spread of results ranging from parity with chips like the Ryzen 5950X to results showing itself 20-30 percent ahead. This isn’t particularly surprising; companies typically present a range of game benchmarks that collectively show varying relationships between themselves and their competitors:
Many of Intel’s performance claims, however, are significantly larger than 1.19x.
Alder Lake’s own leaks have contributed to user perceptions of this oddity. In some cases, leaked benchmarks have implied that Alder Lake could be much faster than the 1.19x uplift Intel has told us to expect. In other cases, we’ve seen leaks imply that AMD’s Ryzen 9 5950X can hold off the newcomer. When we asked Intel about the variation in some of its claimed improvements, the company told us that it’s measuring the execution speed of multiple workloads simultaneously in various content creation applications.
While most reviewers tend to run one application at a time, running multiple apps simultaneously isn’t necessarily unusual. A content creation workflow with a CPU-heavy and a GPU-heavy component can sometimes run both simultaneously, improving overall throughput.
That’s part of why Intel may be predicting the huge content creation gains we see the company claiming here. Alder Lake’s high-efficiency cores give it an extra eight threads to work with, in applications where boosting from 16 to 24 threads likely comes with its own rewards.
According to Intel, peak single-thread performance for 12th Gen Alder Lake is 28 percent above Comet Lake S, and ~1.14x above Rocket Lake S. Intel’s E-core performance claims are also significant, however, since the company is predicting that its “efficiency” cores are fully the match for a Comet Lake S core at the same frequency. As we said up top — this is less “big/little” and more “big/bigger.”
This is a very interesting slide. According to Intel, the Core i9-12900K is capable of 1.5x more multi-threaded performance at 241W as the Core i9-11900K at 250W. The reason this slide is interesting is because peak power consumption on the Core i9-11900K is well above 250W; the chip has been measured at nearly 300W peak power. Rocket Lake is a power-hungry architecture, so hauling down to 250W improves the comparison for Alder Lake.
There’s another very interesting admission in the slide above. According to Intel, Alder Lake at 125W offers 30 percent faster MT performance than Rocket Lake at 250W. Even if we allow for some corporate fudging, that’s much better power efficiency — but we also know that Alder Lake is just 50 percent faster than Rocket Lake at 241W.
Intel’s slide indicates it needs to increase Alder Lake’s power consumption by 1.92x (125W -> 241W) in order to increase performance by 1.15x (from 1.3x Rocket Lake performance to 1.5x Rocket Lake performance). In aggregate, this graph states that Intel takes a 3.7x power penalty in order to improve performance 1.5x over baseline.
It’s not clear how AMD would perform if we charted the Ryzen 5000 series on a similar graph, but x86 CPUs don’t generally compare well to chips like Apple’s M1 Pro and M1 Max right now when it comes to performance per watt. A major question for Alder Lake is whether it can change that impression and, if so, by how much.
Oh, one other thing on power consumption. After taking fire for the vast gap between PL1 and PL2, Intel has responded to the criticism by redefining PL2 to equal PL1 on unlocked processors.
Intel is claiming that Alder Lake will also offer enhanced thermal performance courtesy of a thinner die and thinner interface material. The company announced this will be offset with a thicker heat spreader, but did not offer any additional commentary on why it had thickened the heat spreader in the first place.
Pricing
Here’s the upcoming Alder Lake product stack, with pricing included:
At the top of the stack, the 24-thread / 16-core Core i9-12900K will be clocked at 3.2GHz / 5.2GHz (P-Core) and 3.4GHz – 3.9GHz (E-Core), with 20 PCIe 5.0 lanes and support for DDR5-4800 for $589. Chips like the Core i5-12600K (6P + 4E) will offer 10 cores and 16 threads for $289. The i5-12600K is priced to move against the Ryzen 5 5600X, which is currently selling for around $299.
Of course, all of this assumes the Core i5-12600K won’t hit market, skyrocket in price, and promptly vanish like water in the Atacama. No bets on that score.
We’ll have more details on Alder Lake — the kind that come with performance figures attached — in the not-too-distant future. For now, it’s enough to note that Intel is claiming some significant performance improvements and power efficiency gains. Leaked benchmarks point to Alder Lake being a contender, but power consumption against AMD is an unknown at this time.
And finally, Intel is claiming that Alder Lake will be a great overclocker. We’ll see on that score. It’s been years since either Intel or AMD shipped a CPU we considered a “great overclocker,” and headroom has been dropping on top-end chips for years. Either way, though, the chip’s hybrid architecture is the beginning of a new (hopefully fruitful) era in x86 design.
Now Read:
- AMD Will Offer V-Cache on AM4, Updated AM5 With Next-Gen PCIe 5.0 Coming in 2022
- The Apple M1 Pro and M1 Max’s Power Efficiency Should Rattle Intel, AMD
- Alder Lake Benchmark Results Surface, but Beware the Fine Print
sourse ExtremeTechExtremeTech https://ift.tt/2Zu5StW
ليست هناك تعليقات:
إرسال تعليق