tag:support.primatelabs.com,2011-01-31:/discussions/geekbench/17124-scores-high-differencesinconsistenciesvariance-ios-iphonePrimate Labs: Discussion 2017-08-24T18:04:28Ztag:support.primatelabs.com,2011-01-31:Comment/428401952017-06-24T19:20:11Z2017-06-24T19:20:13ZScores: HIGH differences/inconsistencies/variance (iOS; iPhone)<div><p>Keyword variants: unusual unusually extreme score different inconsistent variable 'iPhone 7 Plus'.</p>
<p>Dear geeky primates:</p>
<p>The attached photo (PNG) is a screen capture from within the your 'Geekbench 4' app, showing the most recent historical CPU results from my iPhone 7 Plus (256GB) running the latest iOS 10.3.2</p>
<p>Generally speaking the tests where performed with as much external noise or internal limits reduced as much as possible, which may have included:<br>
-enabling Airplane Mode (turning the cellular radio OFF) but re-enabling Wi-Fi (as required to for parts of the benchmarks). -Bluetooth OFF -closed all apps (other than Geekbench 4) -have virtually all apps set to run only when being used wrt Location Services ( i.e. not set to "always") -running tests while device was relatively cool (room temp) vs hot (so as not to unintentional engage thermal breakers/throttling) -Siri turned OFF -Battery: low power mode turned OFF</p>
<p>NB: I did NOT go as far as turning OFF<br>
-Location Services (but virtually all are OFF as mentioned) -any existing Notifications</p>
<p>As you can see from the attached there is a HUGE variation in the benchmark results. I'm comfortable with a 10% deviation anytime or a 5% deviation when I try to control for any noise. The APRIL results are in line with the aggregates you've compiled. Then recently results drop by HALF. But seem to BOUNCH back. Disturbingly, the LAST results show MULTICORE results LOWER than SINGLECORE.</p>
<p>Given these results, a layperson would expect the COMPUTE benchmarks to vary in sympathy with these ERRATIC CPU benchmarks. This was not the case. That benchmark test (results not shown) was run at similar times (hence similar conditions) but were always in line with the aggregates.</p>
<p>What do you make of this monkey business?</p>
<p>P.S.<br>
I read both your white papers describing the nature of the benchmarks. I looked for answers in you Knowledge Base. I tried to review all relevant posts on your Discussion Board. I understand what is being executed. I did not find any answers. Nor did I see such extreme results (regardless of HW or OS). There is an 800 lb gorilla in the room. OK, just under 363 kg. Still, that's a lot of gorilla.</p></div>Bernie Hildebrandttag:support.primatelabs.com,2011-01-31:Comment/428401952017-08-24T18:04:26Z2017-08-24T18:04:26ZScores: HIGH differences/inconsistencies/variance (iOS; iPhone)<div><p>Hi Bernie,</p>
<p>Thanks for your message, and I apologize for the delayed reply. Based on your description, I am not sure what could be causing this issue. Could you please resend your PNG image for me to take a look at?</p>
<p>Regards,<br>
Colin<br>
Primate Labs Inc.</p></div>Colin