The GeekBench Challenge

BuzzPoet

I just discovered this nice little tool, which you can find at http://primatelabs.ca/geekbench/ and I thought we could share the results for different low end boxes. So I decided to run it on the three boxes that I have now, at AlienVPS, BuyVM, and ChicagoVPS.

I should note that these are the AlienVPS 192 (1 core), BuyVM 128 (1 core), and ChicagoVPS 1 GB (4 core). I don't know if if the 32 bit version of GeekBench uses more than 1 core, but all VMs are running 32 bit OSes.

Also, for the AlienVPS and BuyVM, all services were turned off. Unfortunately, I couldn't do that for the ChicagoVPS, which may have contributed to the large variance.

You can see the individual results here: http://browse.geekbench.ca/user/buzzpoet/geekbench2

Here are the machines:

AlienVPS: AMD Opteron(tm) Processor 6128
BuyVM: Intel(R) Xeon(R) CPU L5520 @ 2.27GHz
ChicagoVPS: Intel(R) Xeon(R) CPU E5620 @ 2.40GHz

The averages over 3 tests were as follows:

AlienVPS: 2165
BuyVM: 2698
ChicagoVPS: 4953

For reference, I also tested two of my personal computers, which are a Core 2 Duo 2.8 GHz with Linux Mint 10 (32 bit), and a Core 2 Duo 3.0 GHz with Debian 6 (64 bit), which scored 3768 and 4315, respectively.

rajprakash

I'm sure WHT has enough of these posts to find all the info you need about any provider you need.

Furthermore, you need to identify what the node load avg was a the time of the run. If you run it on an empty box, would the results be different than an fully loaded (not overloaded) box at peak operating times, or a overloaded node at non-peak times, or an overloaded node adn peak times, or ... you get what I'm saying?

BuzzPoet

The results may change as load changes throughout the day, or on different days of the week, so it's a valid point that the tests should be repeated at different times. However, whether your VPS is running on an empty vs overloaded node is a crucial difference between different providers that you'd want to factor in, so these numbers are more important than the ideal numbers you'd get on all empty nodes, for example.

Francisco

The reason you see such a difference is because of the core count

Being able to rip 2600 on a single core is frantastic.

Francisco

miTgiB

/me sees your 2600 and raises to 5800

http://browse.geekbench.ca/geekbench2/view/433043
1 core 256mb KVM on E3-1270

dmmcintyre3

128mb Hostigation KVM: http://browse.geekbench.ca/geekbench2/view?id=456781

96mb SecureDragon.net Xen: http://browse.geekbench.ca/geekbench2/view?id=456790

128mb BuyVM OpenVZ: http://browse.geekbench.ca/geekbench2/view?id=456785

NordicVPS

Score: 5837 - 1 core 128mb KVM on E3-1270 - http://browse.geekbench.ca/geekbench2/view?id=454725

yomero

Bah, Sandy bridge just scores like hell because the memory controller, not really because its raw processing power.

PS: Nice to see the new LET version in "stable" xD

NordicVPS

@yomero Sandy still scores high in both integer and floating Performance compared to other CPUs

circus

xen 128 MB:

• vooserver --> http://browse.geekbench.ca/geekbench2/view/437945
• evorack --> http://browse.geekbench.ca/geekbench2/view/437952
• inception --> http://browse.geekbench.ca/geekbench2/view/437942

Btw, anyone knows how to change the geekbench result title?

jh

Do you guys use the paid version of Geekbench? I'd quite like one of ours Geekbenched but don't really want to pay the $20.

NordicVPS

No, I just use the free version

yomero

@NordicVPS Yes, it scores higher, 3.4 Ghz plus new microarchitecture, but the big difference in the scores compared with Nehalem are mainly because the memory controller.

salmon

VPS Package 1 Volume drive: 3231
http://browse.geekbench.ca/geekbench2/view?id=457805

jh

4 cores, 600MB RAM on a dual 5640 system: 5715

dmmcintyre3

http://browse.geekbench.ca/geekbench2/view?id=490865
128mb Hostigation OVZ

Kairus

http://browse.geekbench.ca/geekbench2/view/493167
eNetSouth two cores of Intel(R) Xeon(R) CPU E31270 @ 3.40GHz

In comparison to my Windows desktop: http://browse.geekbench.ca/geekbench2/view/493168 (proc is actually at 4.6ghz)

I think it might beat my desktop proc if I had four cores allocated, although the Windows test was only the 32-bit version.

DanielM

HOME LAPTOP:
http://browse.geekbench.ca/geekbench2/view/493177

UPTIME VPS 1GB WINDOWS:
http://browse.geekbench.ca/geekbench2/view/493179

Carstensz Pyramid Server 512MB KVM Windows:
http://browse.geekbench.ca/geekbench2/view/493181

NQHost 128MB Russia VPS:
http://browse.geekbench.ca/geekbench2/view?id=493182

VMotion IT 256MB Xen VPS:
http://browse.geekbench.ca/geekbench2/view?id=493183

QualityServers 600MB Xen VPS:
http://browse.geekbench.ca/geekbench2/view?id=493185

UK2 256MB "Cloud" Xen VPS:
http://browse.geekbench.ca/geekbench2/view?id=493187

FitVPS 128MB OpenVZ IPV6 Only:
http://pastie.org/2667973

My laptop Wins

yomero

@DanielM You can get a file to submit the result offline in Geekbench (FitVPS)

DanielM

@yomero Where? the program closed once it failed to upload.

yomero

$ ./dist/Geekbench-2.2.0-Linux/geekbench_x86_32 --help
Geekbench 2.2.0 : http://www.primatelabs.ca/geekbench/

Usage:

   [ options ]

Options:

  -h, --help             print this message
  --unlock EMAIL KEY     unlock Geekbench using EMAIL and KEY
  --load FILE            load and display Geekbench result in FILE
  --save FILE            save Geekbench result to FILE
  --export-html FILE     export Geekbench result as HTML to FILE
  --export-xml FILE      export Geekbench result as XML to FILE
  --export-text FILE     export Geekbench result as text to FILE
  --upload               upload results to the Geekbench Result Browser
  --no-upload            don't upload results to the Geekbench Result Browser
  --stress               run stress test
  --sysinfo              display system information and exit

fan

ramhost.us 512mb kvm la: http://browse.geekbench.ca/geekbench2/view?id=493281

Kuro

Not exactly a lowendbox here but... http://browse.geekbench.ca/geekbench2/view/516892

screenshot of htop while geekbench was running: http://i.imgur.com/lz7W9.png

yomero

@Kuro

Damn! That's an Amazon server? What plan? :S

Kuro

@yomero

Cluster Compute Eight Extra Large Instance

60.5 GB of memory
88 EC2 Compute Units (eight-core 2 x Intel Xeon)
3370 GB of instance storage
64-bit platform
I/O Performance: Very High (10 Gigabit Ethernet)
API name: cc2.8xlarge

Normal price is $2.40/hr, picked one up for $0.54/hr though :P

yomero

OMFG! That is "cheap"

Joel

Edis.At // 256mb // (Intel(R) Xeon(R) CPU E5649 @ 2.53GHz // Score: 14236
OnePoundWebHosting // 256mb // (Intel Xeon E5345) 2.33 GHz // Score: 4513
AlienVPS // 192mb // (AMD Opteron(tm) Processor 6128) 2.00Ghz // Score: 1998

Centaur

Can someone maybe run a geekbench on a Hostigation OVZ-1024 package. I know it isn't exactly a LEB. But I would really be grateful.

Also, is 2600 slow on a 3 core VPS?

yomero

@Centaur said: Also, is 2600 slow on a 3 core VPS?

Probably yes. Maybe your node is overloaded, or your CPU is capped.

tux

My netbook give following results with opensource radeon driver.

Centaur

@Joel, is that Edis.at score for real. What package do you have with them and what location. Really good score. Almost as if you have the entire server for yourself. lol