My go to selection for high-end applications has been APC3100 Celeron with 8Gb RAM and 64Gb storage for quite a while. My favourite target. Does anyone has similar experience with this target and how does it compare to the APC3200 Celeron ?
Regards,
Hello,
Due to the newer CPU Architecture and the trend to multi core, the APC3200 has a little less single core performance. But with AR6 you can use multi core on the Runtime and send some Load to the other Cores.
If your previous System was on high load, you might have to adapt some settings, if you switch to the new system.
Greetings
Michael
B&R publishes a huge pdf with all current CPU Benchmarks to compare them:
In Chapter 4.2.5 (page 88) APC3100 and APC3200 with all configurations are compared.
Unfortunately no benchmark is marked as single core or single threaded. You could compare them at some benchmark page:
But these results have to be taken with a grain of salt as the performance is highly dependent on the settings, wattage, cooling, boost speed etc. B&R uses.
It is definitely worth to discuss this with your local support as it depends highly on the application.
APC3200 celeron has f.e. half of the clock frequency the APC3100 celeron can handle. Multi-core is a plus for sure but for heavy load applications like robotics you might run into issues.
Hi @francis.pion ,
I think in US @austin.carpenter is our reference for this topic, I’m sure He is more than happy to review with you different options we have in our catalog (I’m so excited for example to test on real machines our new APC2300, both of them 
)
Ciao
Valerio
Half the clock speed is somewhat true. But with current CPUs it’s not that simple for two reasons:
- The APC3200 CPU (U300E) can boost up to 4.3 GHz. The APC3100 (Celeron 3965U) has a fixed clock speed of 2.2 GHz. Im not sure if B&R uses the boost as it’s used in “normal” PCs.
- From Kaby Lake (3965U) to Raptor Lake (U300E) Intel significantly widened CPU execution. Larger out-of-order windows, more execution ports and improved schedulers allow the processor to exploit more instruction-level parallelism, meaning it can execute more instructions per clock cycle (higher IPC). In this case comparing only the base clock speed has limited relevance.
This is correct for Windows, but not for hypervisor and ARembedded where a fixed clock speed is used. With high fixed clock speeds temperature could be an issue depending on passive or active cooling.
Exactly. Hence my question. For AR application only, how does the 2 compare ?
For Automation Runtime benchmarks on PCs, you want the Industrial PCs Benchmark Automation Runtime User’s manual.
Industrial PCs Benchmark Automation Runtime User’s manual | B&R Industrial Automation
The comparison has already been done on the AR benchmarking application, and as you can see, the xPC3100 has higher clock speeds and therefore generally better performance (lower values are better).
However, that’s not the whole story. Just looking at clock speeds, you can see the xPC3200 platform is more powerful (p-core at half the frequency, the score is not twice as high). The xPC3200 also benefits from multicore support, utilizing the e-cores to offload some processing power from the p-core. The application may need to be redesigned to take the most of out multicore support, but it will help some even without changes. Finally, the clock speeds shown in the AR benchmarking guide aren’t fully optimized for the majority of applications. With the i-series xPC3200, we are able to execute the vast majority of applications at a higher clock speed than was tested when utilizing custom BIOS settings. For example, the i5-PC3200 can execute the majority of AR-only applications at 2.5 GHz while maintaining 2x p-cores and 4x e-cores.
-Austin
2xP Core + 4xE cores at 2.5Ghz must be juicy in AR. I will project similar capability for the Celeron (aka raise the base clock) and get all the cores which should yield better performance.