8 ms·
Here [1] is an actual test and benchmarks. Showing M1 is faster than a Laptop with Nvidia Discrete GPU. Quite amazing. [1] https://petapixel.com/2021/03/10/re
by ksec 6y ago
Here [1] is an actual test and benchmarks. Showing M1 is faster than a Laptop with Nvidia Discrete GPU.
Quite amazing.
[1] https://petapixel.com/2021/03/10/real-world-test-photoshop-for-apple-silicon-is-a-game-changer/ https://petapixel.com/2021/03/10/real-world-test-photoshop-f...
- arghwhat 6y agoThey write that any GPU workload scores similar to the intel internal graphic[1]. For the total scores, they're intentionally using an old beta of the benchmark as it includes an older CPU workload that the M1 happens to excel at. Seems a little like they're cherry picking results. Interesting, but a little dishonest. ---- [1]: Quoting the article: > Unsurprisingly, the M1 Mac mini loses to the competition in raw GPU performance, more-or-less matching the onboard graphics of the quad-core Core i7 that’s in the 13-inch MacBook Pro (full review here).
- danlugo92 6y agoFact check: even if the M1's gpu performance matches that of an i7 as opposed to a discrete gpu it: 1) Is an entry level chip matching a high-end's performance. 2) it does so in around a 5th of 6th of the thermal workload.
- danlugo92 6y agoThermal load *
- africanboy 6y ago1) macs are not entry level, entry level is a 350$ laptop mounting an i3 2) it doesn't matter unless you work with the laptop sitting on the bare skin of your legs EDIT: sorry! forgot the basic rules: don't dismiss bitcoins, don't engage with Apple users. My fault.
- ben-schaaf 6y agoThe thermal load matters for battery consumption.
- africanboy 6y agomy 400$, 6 years old Chinese entry level laptop does 8 hours on battery...
- valuearb 6y agoMy raspberry pi does 9 hours.
- dvfjsdhgfv 6y agoYour post is a poor fit for the collective mindhive here and it makes me uneasy. I better downvote it.
- africanboy 6y agoKeep calm and everything will be fine.
- blunte 6y agoFor the historical HN crowd, a MacBook Air less than $2000 is definitely entry level. My high end Dell XPS was $2600, and the M1 Air blows it away for half the price.
- africanboy 6y ago> For the historical HN crowd, a MacBook Air less than $2000 is definitely entry level. so basically 99% of the laptops are entry level... Anyway the full optional Macbook Air on the Italian Apple website costs € 2.349,00
- asutekku 6y ago
- jeswin 6y agoThe i7 (and Intel) has always had mediocre graphics performance, mostly just for 2D, office/productivity and media encoding/decoding. What'll be interesting is how the M1 will compare upcoming 4th gen Ryzen laptop parts with RDNA2.
- JoshTko 6y agoThose chips will probably not be out until later this year and will take another couple months to show up in laptops. By then we will see the midrange versions of the M1. In Macbook pros.
- MengerSponge 6y agoI agree wholeheartedly, but a small correction: those chips are available now if you're very very lucky. For example, the latest ROG Zephyrus is available from my neighborhood Best Buy. The M1 has been readily available for months, and we'll probably see Apple's new pro chips by the time AMD's supply chain issues are sorted.
- andy_ppp 6y agoMy bet is Ryzen will be slightly competitive this time around (at maybe double the power usage), but Apple will increase performance by 10-20% per year and eventually AMD will start lagging behind.
- danlugo92 6y agoNot only is your post an example of the Moving goalposts fallacy [0]. Is also an apples to orange comparison. The M1 was being compared to an i7, and that was I was responding to. It will also not "be interesting", since if we part from "M1 does not beat current discrete GPUs" as per grandparent comment, it will obviously not beat an upcoming one, whatever the wattage will be. [0] https://en.wikipedia.org/wiki/Moving_the_goalposts https://en.wikipedia.org/wiki/Moving_the_goalposts
- neogodless 6y ago4th generation Ryzen parts are expected to replace their Vega-based integrated GPU with an RDNA2 integrated GPU. That comment was not about discrete GPUs.
- thereddaikon 6y agoBeating the integrated graphics in an i7 is nothing, especially on the older chips. Intel integrated has never been about performance. It really isn't saying much when comparing to the 10th gen chips. They use the old Iris graphics. Apple didn't adopt the 11th gen which are actually contemporary to the M1 and use an entirely new integrated gpu.
- ogre_codes 6y agoIt is an entry level chip. And it’s a fantastic one. But a comparison is a comparison even if lopsided, and when people are pushing this as “As fast as high end Intel”, it’s going to be squared against high end intel setups with discrete GPUs. The M1 is absolutely fantastic. But it does absolutely fall short in some (arguably small) ways. Pointing those out isn’t unfair.
- okamiueru 6y agoI don't understand what is behind the "filter" benchmarks, and it smells a bit fishy. Image filtering are (should be) highly GPU intensive tasks, so what does it mean to score low on "GPU" but somehow high on "filtering".
- agloeregrets 6y agoResults are a little strange. Theoretically speaking, the M1 is a 2.6 TFlop GPU at FP32. The Macbook Pro they are benching it with is just 1 TFlop. to get the exact same result on a GPU with 2.6X the FP perfromance seems....unusual. I would be interested to see if their benchmark is a limiting factor here as realword is dramatically different.
- ribit 6y agoI think this is more a testament to the quality of Adobe's GPU pipeline (or lack thereof). M1's GPU is surprisingly capable for a 10W part, and more importantly, it supports low-latency CPU/GPU communication and opt-in memory page sharing. The later is especially interesting for creative software, as CPU and GPU can work on the same data structure, eliminating the need for copies or GPU buffer management. Knowing Adobe however, it's probably most likely that they are still using OpenCL, so the lack of performance is not surprising. P.S. Or maybe the GPU test is simply memory bandwidth limited...
- ksec 6y agoThe workload on Photoshop should be a mix of CPU with GPU acceleration. And possibly NPU with M1. So the reference to GPU was even with a discrete GPU doing whatever acceleration Adobe has done for Nvidia, the M1 is still faster. And doing so with a Power Usage that is far less than its competitor. Of course if we only focus on "GPU" category than the results as listed is quite clear. Which leads me to believe the NPU and CPU may have played a big part of it.
- izacus 6y agoNone of those laptops has a modern Ryzen CPU in them though? For a more fair comparison? Also your post is a bit misleading: GPU performance is comparable to Intels, not nVidias which is explicitly called out.
- unix_fan 6y agoRyzen isn’t popular in the high end laptop class.
- bogdan 6y agoAny guesses on why? I'm trying to get a new dev laptop with a Ryzen 4xxx or 5xxx but none of the _big_ enterprise vendors (dell, lenovo) have it yet.
- neogodless 6y agoLenovo is much better than Dell and has announced a lot of Ryzen 5xxx laptops. Not sure if the Thinkbooks/Thinkpads are available yet but they are coming. AMD just caught up (to Intel) in 2020 in laptop efficiency, and the OEMs are starting to catch up in offerings in 2021. However, availability is a mess in general lately. Might be 2022 before really high-end laptops offer Ryzen regularly.
- redisman 6y agoThe M1 is honestly very good value and the overall best mobile chip right now which is rare for Apple. Unless you need a PC. Goes toe to toe with my big desktop machine on dev workloads.
- stagger87 6y agoPretty much all major laptop vendors have Ryzen 4xxx models. Some like Dell only put them in their lower tier models. https://www.amd.com/en/shop/us/Laptops https://www.amd.com/en/shop/us/Laptops should get you started on where to look. 5xxx models are just appearing, look up the Zephyrus G15.
- neogodless 6y ago
- sbarre 6y agoSo do this just mean that the legacy CPU architectures we've been working with for decades are just really outdated? Is this less about M1 being amazing and more about Intel being inefficient, in terms of utilization of power/speed/features? Because I have to assume the M1 chip can ignore a host of legacy compatibility/performance tradeoffs that must exist in x86 architecture. If this is the result, hopefully it means a more generalized leap forward (not just for Apple hardware) at some point in the coming decade..