Forums - PC Discussion - 2Ghz AMD Jaguar Benchmarks

Tagged games:

Hi I'm here again to show some new stuffs about the AMD Jaguar core that will be present in the PS4 in a 8-core form. The first think you need to know is what the specs of this Jaguar based CPU.

* The specs are abou 1 core only *

The name specs suggest a four-core Mobile ("M") chip running between 0.8 and 2.0GHz core clock... it is the same clock to have the PS4 CPU (2.0Ghz). The chip have a TDP of 25W... so that means the PC4 APU could have a good TDP too.

Now what about the Benchmarks?

Well you can found here https://www.osadl.org/CPU-benchmarks.qa-farm-cpu-benchmarks.0.html, look at the rack #9, slot #1, or "r9s1" but like I'm a good vgcharterz I will try to make a more compreensive table.

Single-core CPU Unixbench benchmak
System  Arch Name Cores BogoMIPS kDhry Whet Execl kCopy kPipe Index Effective
rack9slot1 "r9s1" x86_64 Jaguar "Kabini" @ 2.0Ghz 4 x 1 15968 13418 1224.9 1731.1 146.83 350.2 403.8 22.04. 18:11
rack8slot7 "r8s7" armv7l ARMv7 @ 996MHz 4 x 1 3156 217.5 3094 338.4 78.79 101.4 116.7 20.04. 06:07
rack1slot5 "r1s5" ppc PPC e500v2 @ 1.2Ghz 2 x 1 400 3611 543.3 258.1 73.67 178.9 119.2 03.03. 18:17
rack4slot0 "r4s0" i686 i3-2100T @ 2.5Ghz 2 x 2 19952 19232 2673.5 1044.0 247.20 537.7 495.5 22.04. 20:16
rack1slot8 "r1s8" i686 Atom Z530 @ 1.6Ghz 1 x 2 6399 3903 555.6 1099.0 98.86 230.8 228.8 22.04. 20:16

 

Multi-core CPU Unixbench benchmark
System Arch Name Cores BogoMIPS kDhry Whet Execl kCopy kPipe Index Effective
rack9slot1 "r9s1" x86_64 Jaguar "Kabini" @ 2.0Ghz 4 x 1 15968 53549 4898.6 6540.1 84.87 1383.1 840.7 22.04. 18:11
rack8slot7 "r8s7" armv7l ARMv7 @ 996MHz 4 x 1 3156 12310 870.5 630.5 34.33 403.2 206.9 20.04. 06:07
rack1slot5 "r1s5" ppc PPC e500v2 @ 1.2Ghz 2 x 1 400 7128 1088.4 414.4 0.44 354.0 54.0 03.03. 18:17
rack4slot0 "r4s0" i686 i3-2100T @ 2.5Ghz 2 x 2 19952 41981 9472.7 4240.8 64.56 1227.2 749.6 22.04. 20:16
rack1slot8 "r1s8" i686 Atom Z530 @ 1.6Ghz 1 x 2 6399 5363 989.2 1526.6 133.83 273.4 305.9 22.04. 20:16

What more? There are a picture with the detailed cache/memory bandwith plot:

Sources

http://citavia.blog.de/2013/04/17/2-ghz-amd-jaguar-benchmarks-15761535/
https://www.osadl.org/QA-Farm-Realtime.qa-farm-about.0.html

Now what theses benchs means? I don't know for sure but while the Jaguar seems weak in single-thread tasks compared to Core i3 in multi-thread it became better in almost all benchs... so a 8-core Jaguar can be a good solution in the end due at the multi-thread power.

I ask for some help to the tech guys here to say what these results means?

Thanks and good night.



PlayStation Nation

"Making haters eat crow since 1994"
Lord of Gran Turismo - Ethomaz the Undertracked

Around the Network
Ohhhhhh I forgot to put some others AMD CPUs in the table but you can look at the x86_64... there are a lot of models.

PlayStation Nation

"Making haters eat crow since 1994"
Lord of Gran Turismo - Ethomaz the Undertracked

The improvements over the previous Bobcat.



PlayStation Nation

"Making haters eat crow since 1994"
Lord of Gran Turismo - Ethomaz the Undertracked

Obviously the cores have power-management-features but I can't really look up the abbreviations.
Ethomaz, is this a machine you have access on? Is the listing from cpuinfo from you? Because if it is I can tell you other commands ;) The difference-picture regarding bobcat and jaguar is not revealing anything groundbreaking - higher clock, more cache, enhanced isa -> standard cpu-evolution.
The benchmarks are strange somehow - why did you choose exactly these for comparison? :) And the site is awful to me to look at but ok, they want to show as much as possible in a small format.
Ah, found things about the clock:
CPU scaling governor
CPU0: ondemand from 800 to 2000 MHz
CPU1: ondemand from 800 to 2000 MHz
CPU2: ondemand from 800 to 2000 MHz
CPU3: ondemand from 800 to 2000 MHz
And according to the cpuinfo-info it can do these in 100mhz-steps, nice.
Overall at least you should have explained what the benchmark-table shows (means: what is tested exactly) because at first look they can mean anything ;)

walsufnir said:
Obviously the cores have power-management-features but I can't really look up the abbreviations.
Ethomaz, is this a machine you have access on? Is the listing from cpuinfo from you? Because if it is I can tell you other commands ;) The difference-picture regarding bobcat and jaguar is not revealing anything groundbreaking - higher clock, more cache, enhanced isa -> standard cpu-evolution.
The benchmarks are strange somehow - why did you choose exactly these for comparison? :) And the site is awful to me to look at but ok, they want to show as much as possible in a small format.
Ah, found things about the clock:
CPU scaling governor
CPU0: ondemand from 800 to 2000 MHz
CPU1: ondemand from 800 to 2000 MHz
CPU2: ondemand from 800 to 2000 MHz
CPU3: ondemand from 800 to 2000 MHz
And according to the cpuinfo-info it can do these in 100mhz-steps, nice.
Overall at least you should have explained what the benchmark-table shows (means: what is tested exactly) because at first look they can mean anything ;)

No, these are not mine... the site have the specs and info: https://www.osadl.org/Profile-of-system-in-rack-9-slot-1.qa-profile-r9s1.0.html

I found the site hard to understand too so I tried to make a compilation with one of each Arch with similar number of core and clock. I need to study these results to see if they are applied to real compute tasks or for games.



PlayStation Nation

"Making haters eat crow since 1994"
Lord of Gran Turismo - Ethomaz the Undertracked

Around the Network
ethomaz said:
walsufnir said:
Obviously the cores have power-management-features but I can't really look up the abbreviations.
Ethomaz, is this a machine you have access on? Is the listing from cpuinfo from you? Because if it is I can tell you other commands ;) The difference-picture regarding bobcat and jaguar is not revealing anything groundbreaking - higher clock, more cache, enhanced isa -> standard cpu-evolution.
The benchmarks are strange somehow - why did you choose exactly these for comparison? :) And the site is awful to me to look at but ok, they want to show as much as possible in a small format.
Ah, found things about the clock:
CPU scaling governor
CPU0: ondemand from 800 to 2000 MHz
CPU1: ondemand from 800 to 2000 MHz
CPU2: ondemand from 800 to 2000 MHz
CPU3: ondemand from 800 to 2000 MHz
And according to the cpuinfo-info it can do these in 100mhz-steps, nice.
Overall at least you should have explained what the benchmark-table shows (means: what is tested exactly) because at first look they can mean anything ;)

No, these are not mine... the site have the specs and info: https://www.osadl.org/Profile-of-system-in-rack-9-slot-1.qa-profile-r9s1.0.html

I found the site hard to understand too so I tried to make a compilation with one of each Arch with similar number of core and clock. I need to study these results to see if they are applied to real compute tasks or for games.

Ok looking forward to your investigations but please try to get info about what compilers where used and which processor-flags as these factors can change results quite a bit.



walsufnir said:

Ok looking forward to your investigations but please try to get info about what compilers where used and which processor-flags as these factors can change results quite a bit.

I will try to find these options.

I don't know if you think that interesting but AMD released two days ago the Optimization Manual for Jaguar: http://support.amd.com/us/Processor_TechDocs/52128_16h_Software_Opt_Guide.zip

Edit - I fixed a mistake in my table... the i3 is a 2.5Ghz model.



PlayStation Nation

"Making haters eat crow since 1994"
Lord of Gran Turismo - Ethomaz the Undertracked

Those are some odd points of comparison tho I mean a 5 year old single core Atom CPU, a 2 year old low end dual core i3, an 2 year old Cortex A9 based CPU. I mean most of those have long been discontinued.




@TheVoxelman on twitter

Check out my hype threads: Cyberpunk, and The Witcher 3!
zarx said:
Those are some odd points of comparison tho I mean a 5 year old single core Atom CPU, a 2 year old low end dual core i3, an 2 year old Cortex A9 based CPU. I mean most of those have long been discontinued.

There is a i7 in the result and some others processors but none "similar" to four-core Jaguar... I tried to get the most close to what Jaguar is... for example there are just 4 PPC processors listed and just one with more than one core.



PlayStation Nation

"Making haters eat crow since 1994"
Lord of Gran Turismo - Ethomaz the Undertracked

Around the Network
Ok, here are some explanations:All benchmarks were done using Unixbench

kDhry: Dhrystone-benchmark
mainly integer- and string-performance. NO floats. http://en.wikipedia.org/wiki/Dhrystone

Whet: Whetstones
mainly integer- and string-operations: http://en.wikipedia.org/wiki/Whetstone_(benchmark)

execl:
"This test measures the number of execl calls that can be performed per second. Execl is part of the exec family of functions that replaces the current process image with a new process image. It and many other similar commands are front ends for the function execve()."

kcopy: file copy
"This measures the rate at which data can be transferred from one file to another, using various buffer sizes. The file read, write and copy tests capture the number of characters that can be written, read and copied in a specified time (default is 10 seconds)."

kPipe: pipe throughput
"A pipe is the simplest form of communication between processes. Pipe throughtput is the number of times (per second) a process can write 512 bytes to a pipe and read them back. The pipe throughput test has no real counterpart in real-world programming."


So overall these benchmarks show how good a cpu will perform in everyday use - obviously there are no tests that especially consider gaming (meaning floating-point operations). Still interesting read.