A53 vs A72 performance

[enip]

I know A72 is performant while A53 is energy efficient. But what about the Spectre vulnerability? A72 is vulnerable, while A53 is not. The vulnerability is somehow fixed in linux. Hence in recent linux we have pure A53 vs A72 + the fix overhead.

So is A72 + the Spectre fix overhead is still more performant than A53?

[fire219]

Even with the Spectre patches, an A72 core is much faster than a A53 core of the same clock speed.

How much faster is a matter of implementation, and I don't have any good comparisons on hand. What I can tell you is that the implementation in the Raspberry Pi 4 is not a good one.

[enip]

Even with the Spectre patches, an A72 core is much faster than a A53 core of the same clock speed.

How much faster is a matter of implementation, and I don't have any good comparisons on hand. What I can tell you is that the implementation in the Raspberry Pi 4 is not a good one.

I wasn't thinking of pi but could you please elaborate the pi4 part? What is wrong with it?

[stuartiannaylor]

Even with the Spectre patches, an A72 core is much faster than a A53 core of the same clock speed.

How much faster is a matter of implementation, and I don't have any good comparisons on hand. What I can tell you is that the implementation in the Raspberry Pi 4 is not a good one.

I wasn't thinking of pi but could you please elaborate the pi4 part? What is wrong with it?

Its probably more to do with thermals as both the pi3 & pi4 are the usual low clock of raspberry.
Pi4 is clocked at 1.5 Ghz as opposed to an 1.8Ghz avg with the rk3399 even though its just a pair rather than quad.

A lot of the benchmarks are without cooling and to be honest unless its active it delays throttling more than it lower temp if passive.
The Pi4 isn't that bad its a reasonable priced Arm SoC SBC but that is it really.

It follows the strange tradition of an only incremental upgrade than previous where for function there is still much overlap with product and price to previous that are still existing sales product.
Its what it should of been makes the Pi4 a big disappointment as with the revenue, economies of scale and resources Raspberry have, the likes of Pine should of been blown out of the water.
The opposite really has happened and the RockPro64 is probably in a stronger place now there are no false expectations of what the Pi4 will be.
The incremental update managed to kill the Pi form factor with this 2x hdmi thing that still has the VC4 engine that struggles to play 1080p in Chromium.


For me what is bad about the Pi4 is that Raspberry puts on this big show as this education focussed benefactor that garners the bonuses of charity status but if you look at there latest offering, the decisions where purely in line with its commercial operations and there is a lot of snakeoil in a pot of raspberry jam.
IMHO was more of a blocking product released early due to fears of Panfrost and various other offerings becoming mainline kernel complete, it was more dumped on the community with a 'that will do them' than any design or investment in what should of been a game changing SBC.

The benchmarks are pretty poor for a quad A72 with the RockPi63 being 20-30% faster on average, but its Achilles heel is that VC4 GPU as it sucks mightily.
For me the Pi4 being a desktop replacement is pure Snakeoil but actually the RockPro64 with Panfrost is what I would consider low entry level desktop computing that with PCIe is capable multiple sata3 or even nvme ssd that make the desktop argument even more valid or pi's claims less so.

The Pi4 aint that bad but much of what is said is pure BS.

https://www.tomshardware.co.uk/arm-corte...33404.html