AMD laptops have a hidden performance lag of 10 seconds. This is why | GeekComparison

Stylized illustration of computer component.
Enlarge / When running on battery, your new Ryzen 4000 series laptop will only deliver its true performance after about 10 seconds in a full workload.

In an embargoed Friday morning presentation, Intel Chief Performance Strategist Ryan Shrout led a group of tech journalists through a presentation aimed at tearing down AMD’s Zen 2 (Ryzen 4000 series) laptop CPUs.

Intel’s latest laptop CPU design, Tiger Lake, is a really compelling release, but it comes on the heels of some crushing disruptions in that space, leaving Intel looking for an angle to avoid bleeding its rival’s market share . Early Tiger Lake systems performed incredibly well, but they were configured for a 28 W cTDP, rather than the much more common 15 W TDP seen in production laptop systems, and reviewers were not allowed to test battery life.

Because of this, reviewers like yours really compared Intel’s i7-1185G7 with 28W cTDP to AMD Ryzen 7 systems at half the power consumption – and while Tiger Lake came out on top overall, the power difference kept it from being a decisive factor or was a crushing blow to AMD’s market share gain among the OEM vendors that are in fact to buy laptop CPUs in the first place.

Enter the battery

Intel’s original Tiger Lake launch presentations attempted to draw attention to battery vs. no-battery discrepancies in AMD’s performance, but those attempts mostly went unheard. Shrout’s presentation on Friday was an attempt to retell that story, this time with enough additional information to get people excited.

We can easily see this discrepancy between performance on battery and off battery performance in the PCMark 10 Applications benchmark and also in many of Intel’s RUGs – scripted workloads based on production applications, which the company calls “Realistic Usage Guides”. However, the same discrepancy between performance on and off battery is not visible in more widely used industrial benchmarks, such as Cinebench, PassMark or Geekbench.

Intel’s engineering team lists the reason Why we don’t see the discrepancy in Cinebench in the final image from the gallery above – during Intel’s testing, the Ryzen 4000 CPUs only ramped up power and voltage to somewhere between eight and eleven seconds. after heavy workloads started.

Independent confirmation

We were able to confirm Intel’s findings over the weekend, working with an Acer Swift 3 SF314-42 laptop (with a Ryzen 7 4700u CPU) and an MSI Prestige 14 Evo laptop (with a Core i7-1185G7). In the graphs above, we repeatedly compress small chunks of the Linux 5.3 kernel source and graph throughput over time on each CPU.

The 4-core/8-thread i7-1185G7 easily outperforms the 8-core/8-thread Ryzen 7 4700u in both single- and quad-thread workloads, even after the Ryzen 7 4700u jumps to full performance too late around the 12-second mark. In the unrestricted workload, where the Ryzen 7 is allowed to flex its entire octa-core muscle, things are much closer – and the 4700u even records a narrow victory in the last four seconds.

However, there are a few things we should mention here. First and most obvious, Intel is 100 percent correct in its claims that AMD’s Zen 2 laptop CPUs slow ramping power and voltage to their maximum state. This makes for a sharp corresponding and diminished performance during those first few seconds.

We have reached out to AMD representatives for comment on this design decision. While AMD representatives have asked further questions about our observations, we have not received a response for the record as of this writing.

The devil is in the details, just like the heat

But Intel still plays games with its own power consumption. In the screenshot above, we can see the MSI Prestige Evo 14 with Core i7-1185G7 during a Cinebench R23 run. We haven’t had this laptop long enough to fully rate it, and especially to rate its battery life, which we’re very curious about since we’re prohibited from testing that metric in two earlier i7-1185G7 systems.

But we can see that, rather than calling the i7-1185G7’s cTDP down to something approaching the typical Ryzen 7 4000 cTDP of 15W, as widely expected – MSI has opted in this laptop to crank it up even further than what we saw in previous prototypes. This production i7-1185G7 system has a variable PL1 that reaches up to 36W during a Cinebench R23 run, in addition to the 51W PL2, which is unchanged from the prototypes.

During this Cinebench R23 run, the laptop ran at its full PL2 power limit of 51W for the first 10 to 15 seconds, with temperatures reaching a blistering 98°C. After that initial extremely powerful power and heat-generating burst, the CPU settled down to maintain an average power consumption of 34W. In contrast, a Ryzen 7 Pro 4750U with 8 cores/16 threads – with a cTDP of more than 25W – consumes an average of 27.9W, peaking at 29.9W.

While we’re digressing from the CPUs themselves into the laptop design realm, it might be worth noting that system fan activity also differed significantly between the MSI Prestige 14 Evo – which almost immediately hit the level of fan noise that was similar with gaming laptops – and the HP EliteBook, which took over a minute to max out the fans and remained much quieter than the MSI throughout its run.

The battle continues

While Intel didn’t specifically tell us what conclusions to draw from the performance slowdown in Zen 2 laptop CPUs versus Tiger Lake’s instant-on performance, the company was clearly hoping for something between “AMD is gaming the benchmarks” and “the Turns out Intel was the winner all along.”

We do not think that such ready-made conclusions can be drawn here. Intel’s findings regarding the slow performance improvement of the AMD Zen 2 laptop CPUs are obviously factually correct – we had no trouble confirming it, and it also explains why many of Intel’s preferred benchmarking techniques show larger performance deltas in Team Blue’s advantage over the more commonly used industry benchmarks such as Cinebench, PassMark, and so on.

But this negates the greater efficiency of the AMD systems, aside from the delayed shift to maximum performance (and battery consumption) in the CPU. Running Cinebench R23 for a full five minutes, a Ryzen 7 Pro 4750u system renders more scenes than the Intel i7-1185G7, and it does this with less total power consumed. There’s no clever trick to explain that away.

We also believe that a reconciliation argument should be made on both sides. Intel’s faster shift to the highest performance state brings some practical benefits, but we’re not sure they’re as convincing as the charts make out. In practice, we’ve now spent quite a bit of time with both Zen 2 and Tiger Lake laptops – and the Tiger Lake systems don’t really to feel faster in terms of a subjective experience. This strongly argues that it often doesn’t make much sense to ramp up CPU power profiles that quickly – if the human running the system doesn’t notice the improvement in latency, it’s probably better to save the battery.

… and continues

There are some exceptions – the most notable is probably the boot time. Tiger Lake systems boot up – and resume from sleep – phenomenally fast – and we suspect their willingness to immediately boost performance to the max has a lot to do with it. One developer we spoke to speculated that JavaScript just-in-time (JIT) compilation could be another short workload that was nevertheless easily detectable by humans.

The best news for consumers, we suspect, is that the “which system is better” argument is so hard to answer conclusively in the first place. This level of competition means neither team can rest on their laurels, and consumers are less likely to buy systems no one would like, if fully informed about the differences.

Leave a Comment