Is it reasonable to expect that the ‘nominal’ processor speed (as to overall benchmark capabilities) of the Rock 5’s RK3588 CPU will generally lie somewhere between the Intel i3 and i5, or would expecting this level of performance from the RK3588 be merely dreaming?
Rock 5's RK3588 speed relative to Intel i3 and i5?
An i3-1220PE is 8 cores 12 threads with the performance cores being 4.2Ghz.
Yes you are dreaming as somewhere between a I3 and I5 means nothing as the difference between generations and models such as there ‘embbeded’ versions such as i3-8145U Processor or Celerons/Pentiums is huge.
Yes it might beat some very old low clocked I3s and the embedded versions that are in some older Nucs.
In terms of modern chips its still likely to struggle against a Intel Celeron N4020 but we will have to see how the faster single cores vs more cores pans out in use.
The nearest thing is prob something like a LENOVO CHROMEBOOK DUET as it has a Mediatek Helio P60T which it should beat but prob not by that much.
The GPU in the RK3588 probably beats all intel cpu based gpu as even the newer Iris are not great but haven’t looked at perf of them for a long while. The Mali G610/710 GPUs still don’t compete with the newer Radeon APUs but certainly beat much of Intel.
Aim at a bit higher than the above Chromebook and yeah stop dreaming.
. . .
This thread creator is trying to compare x86 and ARM by using consumer CPU as benchmark without telling for WHAT exactly he is gonna use it, right?
It’s wrong on so many levels that i can’t even point one correct thing about this.
My aging HP laptop has this processor and 8GB of RAM:
Intel® Core™ i5-6200U CPU @ 2.30GHz × 4
I pretty well know the capabilities of my laptop. If a multitude of generic CPU (not GPU) benchmarks were to be ran pitting the Rock 5’s RK3588 against this specific processor, with also the given of 8GB RAM, how well would the Rock 5 generally stack up, with both systems running Ubuntu Linux.
I agree that the Intel GPU on my laptop is terrible. I presume the GPU on the Rock 5 to likely be much better.
I’m considering a Desktop computer application for home use. Currently my Wife and I are using a Raspberry Pi 400 overclocked at 2.2 Mhz for this purpose, and we are looking for something better as a desktop, and more along the lines of our HP laptop as to its overall speed and processing capabilities, while at the same time enjoying the adventure of working with (and learning to understand) an ARM based SBC that goes well beyond the Raspberry Pi.
We are also considering the upcoming Khadas VIM 4, and the upcoming Rock Zero 2. And having fun learning. And tinkering…
If you got to the Rock5 thread I posted two video links of RK3588 computers being used. Will give you a idea of real world use. For Vim4 you mentioned then search YouTube for " leapsvideo " he uses one with Android 32 bit.
Remember use of SSD is faster than a SDcard. In real world retro game emulation then RK3588 is close to a intel N5105 . But chosen GPU driver, matter a lot.
1 Like
Its very subjective but seen as you have tried a Raspberry Pi 4 with quite a large OC you prob still realise as a desktop its an extremely basic experience even though raspberry do push it but hey the 2gb is £35.
Vim4 vs Radxa Zero 2 is basically G52 cores x4 vs x8 as much of the rest are slight improvements as the Radxa Zero 2 uses the same SoC as the Vim3.
Price wise the Rock5B competes likely with the Vim4 but likely a more powerful machine (pretty much guaranteed)
Radxa Zero 2 (Vim3 Like) actually hits a niche just above the raspberry pi where as a basic entry level desktop it actually works where the pi fails.
I think many of us are waiting for the price Radxa can do these as for many it could be a super cute and capable desktop board.
1 Like
From the Geekbench 4 and Geekbench 5 scores I’ve seen for the RK3588, it seems overall similar to the Geekbench scores that I’m seeing for my HP laptop and its Intel® Core™ i5-6200U CPU @ 2.30GHz × 4. (4 cores)
The RK3588 is roughly about 90% as fast on these tests (as to the overall aggregate final scores), but I figure that it’s early and there should be improvement coming for the RK3588. But then again the scores for the RK3588 are under Android, and on my laptop I’m running the current version of Linux Mint. And perhaps Mint taxes my CPU to a far greater extent than would android. ???
Should this be a ballpark indicator of near equivalence in the performance for both CPU’s?
No it does not. Rk3588 is very expensive just for the cpu chip. The current Price is subsidized price for developement purposes, i except the actual price to be more than $200.
From this thread I think it is safe to say that rk3588 can be used as normal daily desktop for home and basis office use cases like browsing, email and media playback. Anything more than this then you might need more powerful device like honeycomb from solidrun with a dedicated gpu.
We at Manjaro Arm project have used Arm devices as our daily driver for browsing, chatting with team on matrix and some video content and it does a great job at it. Dan uses N2 for a month and I use Amlogic S922x device and it can do the work.
Oh well my wife uses Pinebook Pro(RK3399 4GB ram) for her daily use of only classes and online studies and she didn’t really realize that she is using a non x86 device nor she realized that its not Windows until I informed her. She just thought it some kind of theme.
So yes depending on your work load arm devices are capable enough already for such use cases.
The Rock5b supposed unsubsidized price is https://shop.allnetchina.cn/collections/rock5-model-b/products/rock5-model-b?variant=39514839482470 Regular price$149.00
The subsidized price for early adopters https://shop.allnetchina.cn/collections/rock5-model-b/products/rock5-model-b-rk3588-pre-order-promotion-voucher?variant=39528515272806 Rock5 Model B 8GB RAM USD 149 -> after discount USD99
No one really knows what the cost of the Vim4 will be as its not been added to pricing yet https://www.khadas.com/post/sbc-pricing-update so either they are going to reduce or models will likely be more due to extra ram and the Rock5b will be considerably less than Khadas models which are nice but often with a cost premium as non of their models could be said to be cheap.
Performance is subjective and to quote a singular user experience means nothing really especially from enthusiast reviews.
There is no solid reference apart from generally what are current market specs with very basic being Windows Intel Celeron N3350 4GB 64GB 14 Inch for approx $130 with Chromebooks coming in slightly cheaper with Intel® Celeron® N4020 Processo 4GB eMMC starting as low as $110.
Rock5B likely will exceed the expectations of those but is only a bare board whilst minimum desktops/mini pcs tend be of higher spec with the likes of Beelink GK55 Mini PC Windows 11 Pro,Intel J4125(up to 2.7GHz) 8GB LPDDR4 128GB SSD $185
Its the market that sets specs as that is the environment the majority of users know and the Rock5b prices are well priced with the likes of Khadas tending to be expensive with what users can get elsewhere.
1 Like
My vim3 cost me 139$ over two years ago. I’d happily, cheerfully, gleefully pay 200$ for the Rock5b.
Hi @Larry, you can easily compare different processor benchmarks. For instance, here’s Geekbench5 for an i5-6200U laptop running Manjaro compared to a RK3588 running Android 12: https://browser.geekbench.com/v5/cpu/compare/14428490?baseline=11836244
Hit search at the top for other comparisons meaningful for you.
Besides optimising yet to be completed, I don’t think this tells the whole story. A decent NVMe drive in the Rock5 will provide much higher storage throughput than what that laptop would have had. The new chip’s available memory bandwidth will likely exceed what a 6th gen i5 had available as well (looks like 34.1 GB/s). Then as noted, the Mali-G610 MP4 capabilities are likely in a whole other league than intel’s 520 igpu.
For me, what really stands out with Arm processors is their energy efficiency as denoted in CPU cycles per watt or similar metrics.
My Rock5 will be replacing an AMD APU system I built nearly a decade ago. While the old battleaxe has 4C/4T able to boost to 4.4Ghz, it takes 100W or more to get there.
What provides a “fast” user experience is the culmination of not only a processor but also graphics, memory, storage, OS, optimisations, and so on. As another poster mentioned, what a system is used for is also a key indicator.
Before long we’ll be able to answer this for the rk3588 on the Rock5!
1 Like
Dunno if that comparison is any good as Geekbench might be using a small core Cortex-A55 for the single core bench.
As looking have a gut feeling the difference is too big and usually the index of the smaller cores is at the start so wouldn’t be surprised as it lists as “ARM ARMv8 @ 1.80 GHz 1 Processor, 8 Cores”
The multicore is right but that also lends me to think yeah the single core was the A55.
I did benchmark the radxa cm3 on the io board some time ago:
https://browser.geekbench.com/v5/cpu/12386767
Which resulted in a single core score of 159 and multicore of 493
For comparison (probably not the most scientific method), the intel atom in my car achieves 97 on a single core and 272 on two cores.
https://browser.geekbench.com/v5/cpu/3987750
There’s also a core 2 duo in my profile on geekbench which achieves 224 respective 411 points
This should help you get a clue of the performance levels of arm socs compared to bigger computers
There are already RK3588 development boards on sale and available. Just because Radxa build a board with a RK3588 does not maker the same RK3588 faster!
Just search " RK3588 " on YouTube.
True @stuartiannaylor, so testing on an A76 core @ 2.4 Ghz could potentially gain as much as 1/3 more on the single core test. My thinking too is as an early Android development version, there’s much still on the table to optimise. Then as the linux kernel gets dialed in, it is likely to see another bump. And again with proper cooling, overclocking may offer even more perf on top of that.
Personally I find benchmarks to be less important than how quickly I perceive my regular tasks to go. Some of them are SMT others are single.
What’s truly remarkable is how what seemed performant a few years ago can appear slow next to newer tech. For instance, the ever increasing number of cores/threads, nvme gen4, RAM with higher clocks (or even overclocked) with min latency, increase in shader count and gpu freqs, etc.
I look forward to seeing how far we can take the RK3588 as a community!
Yeah I am thinking the single core could well be an A55 @ 1.8Ghz due to the indexing given to the cores.
Performance is very subjective as often I don’t really notice an upgrade until use and then downgrading after a time becomes hugely noticeable.
The price point Radxa has provided gives an excellent $/performance ratio that is extremely competitive with what is available in the market.
Its why I think a Rock5A of stripped minimal ‘desktop’ function could even increase that excellent $/performance ratio as it gives a real desktop experience than what has been essential more wishful thinking from Raspberry than reality.