Performance Difference between Single/Dual Channel on Goldmount +

[ksec]

I am currently looking for some benchmarks on Gemini Lake, since I cant find anything on the desktop 10W version J5005, the closest I got was the mobile N5000. This is compared to a 14nm Boardwell CPU.

https://browser.geekbench.com/v4/cpu/compare/7351641?baseline=7236747

And by looking at the memory bandwidth I am guessing this is an 8GB Single channel DDR4 memory. I am wondering how much performance gain if dual Channel memory were used. Shouldn't the Goldmount Plus uarch be starved from memory bandwidth?

Also note this is a 15W vs 6.5W, I have no idea how much juice could be squeezed out of Gemini Lake's 15W, assuming running these test to be somewhat TDP limited.

I was wondering if a bumped up 15W Gemini Lake could replace a 14nm Boardwell CPU.


Edit: I got the desktop version from PassMark. ( I have no idea what those score were testing, compared to Geekbench 4 )

https://www.cpubenchmark.net/compare.php?cmp[]=3144&cmp[]=2530

Still a 10W vs 15W comparison. No idea on memory.

 

[IRobot23]

Its simple. Benchmark like AIDA64 will show you that dual channel is x2 faster. In this case you have dual vs single channel memory (ddr3 DC vs ddr4 SC) ... so about 40-50%, also you can see latency is half that.

Is it worth on low power chip? Well, I don't think so.

The problem is that you are looking at two different architectures. One is really great on single core performance.

 

[whm1974]

Do Goldmount based CPUs have enough performance to even take advantage of dual channel memory?

 

[Insert_Nickname]

Do Goldmount based CPUs have enough performance to even take advantage of dual channel memory?

The HD505/UHD605 actually has more EUs* then the venerable HD4000. So in that sense, yes, it'll likely benefit, but it'll still be low performance.

*16 for HD4000 vs. 18 for HD505/UHD605. The smaller HD500/UHD600 "only" has 12. Which is equivalent to GT1 for SKL/KBL. The frequency is a bit lower though.