The Enigma Called Geekbench

flowrider

macrumors 603
Original poster
Nov 23, 2012
6,116
2,321
I am really puzzled by the last release of Geekbench, Geekbench 4.1. After releasing Geekbench 4.0 last year, which changed some parameters that drastically lowered scores I was puzzled. Now Geekbench has released Geekbench 4.1 that raises scores. My cMP's configuration has not changed through all all of this. But, now I am even more puzzled. There seems to be no baseline. To me, the scores are meaningless.

Lou
 

Attachments

Sharky II

macrumors 6502a
Jan 6, 2004
615
29
United Kingdom
In each iteration of geekbench, they change the machine which set the baseline score. The baseline machine gets faster, so your score will go down in comparison.
 

pastrychef

macrumors 601
Sep 15, 2006
4,520
1,027
New York City, NY
I still use GeekBench 3 to compare performance between two systems. The scores make much more sense to me. Plus, the OpenCL benchmark in GeekBench 4 has never worked for me.
 
  • Like
Reactions: bax2003

Synchro3

macrumors 68000
Jan 12, 2014
1,979
835
I am really puzzled by the last release of Geekbench, Geekbench 4.1. After releasing Geekbench 4.0 last year, which changed some parameters that drastically lowered scores I was puzzled. Now Geekbench has released Geekbench 4.1 that raises scores. My cMP's configuration has not changed through all all of this. But, now I am even more puzzled. There seems to be no baseline. To me, the scores are meaningless.

Lou
The Geekbench 4.1 result is much more plausible than the Geekbench 4.0.3 result, especially the multi core result.

For comparison my computers.

Mac Pro 5,1 Xeon W3690: https://browser.geekbench.com/v4/cpu/2296391
Core i7-7700K (no overclock): https://browser.geekbench.com/v4/cpu/2297126
 
Last edited:

AidenShaw

macrumors P6
Feb 8, 2003
18,635
4,633
The Peninsula
Never compare new version of Geekbench to previous version obviously.
This is good advice for just about any benchmark suite.

They are updated to be relevant for current systems, and that can eliminate any meaningful comparison with the older version.

Sometimes it's because a component was "broken" by newer systems.

For example, consider an application where the core code actively uses about 20 MiB of RAM. If you run that code on a system with 512 KiB of cache, it gives the memory system a good workout and is a good test.

On the other hand, if you run it on a Xeon with 40 MiB of cache, the core code sits in cache and it screams. It's "broken", however, because you can no longer compare the real life performance of the Xeon to the older system. The Xeon may seem to be hundreds or thousands of times faster, but on a code that doesn't fit in cache it may only be 5-10x faster.
 

jerwin

macrumors 68030
Jun 13, 2015
2,836
4,633
There's a white paper:

http://geekbench.com/doc/geekbench4-cpu-workloads.pdf

In addition:

Users can expect a 2% increase in single-core scores, and at least a 5% increase in multi-core scores. Note that the multi-core score increase depends on the number of processor cores – systems with more cores will see a larger increase in the multi-core score.
Did geekbench 4.0 also break out the cryptography workload? Some mobile chips have cryptography engines that hit above their weight.

Granted, end to end encryption is normal now, and it's disappointing to use a device that can't keep up with the workload.
 
Last edited:

Sharky II

macrumors 6502a
Jan 6, 2004
615
29
United Kingdom
You're right - I think they've basically just changed the baseline again in the 4.1 update... very silly to keep moving the goal posts. Perhaps they are trying to get the score more similar to the previous version or something, the scores are higher than before.

It also takes much less time to run now, around 60% of the time of the 4.0 update.
 

jerwin

macrumors 68030
Jun 13, 2015
2,836
4,633
Ryzen did come out recently, and apparently it didn't do so well on GB 4.0.

Since the purpose of a benchmark is to model real world performance, occasionally tweaking the benchmark is to be expected.
 

Theophany

macrumors 6502a
Nov 16, 2008
633
186
NW London.
I am really puzzled by the last release of Geekbench, Geekbench 4.1. After releasing Geekbench 4.0 last year, which changed some parameters that drastically lowered scores I was puzzled. Now Geekbench has released Geekbench 4.1 that raises scores. My cMP's configuration has not changed through all all of this. But, now I am even more puzzled. There seems to be no baseline. To me, the scores are meaningless.

Lou
Synthetic benchmarks are always a bit sketchy, in fairness. They serve a purpose insofar as they allow you to compare different hardware setups across the same version of the benchmark, but the tasks a benchmark uses may not be representative of your daily workload.

For a photographer, it would make more sense to time a Lightroom import and convert to DNG or render a large file in Photoshop whereas a dev would compare compile times and a video pro Premiere export times.

Things like Geekbench are great for willy waving, but they're not terribly relevant to each individual's usage case. I think the reason for the raised score in GB 4.1 is because the CPU tasks have been changed/tweaked to adequately benchmark Ryzen chips.
 

splifingate

macrumors 6502
Nov 27, 2013
439
267
ATL
I am really puzzled by the last release of Geekbench, Geekbench 4.1. After releasing Geekbench 4.0 last year, which changed some parameters that drastically lowered scores I was puzzled. Now Geekbench has released Geekbench 4.1 that raises scores. My cMP's configuration has not changed through all all of this. But, now I am even more puzzled. There seems to be no baseline. To me, the scores are meaningless.

Lou
My scores fluctuate, even within the same build.

Why does Synchro3, using a Titan X, derive a lower Metal score than I--using a 980ti--in GB4.1?

Why do I not fully realise a noticeable improvement in perceived performance between a 5,1 and a 1,1?

Should not the 'Baseline' be between users of the same hardware--using the same software--to compare?

http://cbldf.org/wp-content/uploads/2015/06/Zippy-239x300.jpg
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.