MIPS Loongson 3A Benchmarks On Debian

[maldorordiscord]

You definitely have no clue. Ask yourself what simulation speed is.

You can not simulate a single-core and scale this up to a quad-core.
They talk about a multi-core processor but simulate only a single-core.
http://en.wikipedia.org/wiki/Multi-core_processor
The result is not the same.

[ldesnogu]

You can not simulate a single-core and scale this up to a quad-core.
They talk about a multi-core processor but simulate only a single-core.
http://en.wikipedia.org/wiki/Multi-core_processor
The result is not the same.

Instead of stating the obvious, why don't you answer my question about sim speed?

Also as far as multi core goes, would you expect single thread apps to be faster on a multi core or on a single core? I'm wondering why I ask because you don't seem to answer even trivial questions, and this one is not easy

[maldorordiscord]

Instead of stating the obvious, why don't you answer my question about sim speed?

???? please rephrase your question.

Also as far as multi core goes, would you expect single thread apps to be faster on a multi core or on a single core? I'm wondering why I ask because you don't seem to answer even trivial questions, and this one is not easy

Single thread applications run faster on multi-core systems this is a fact!
The biggest impact of that effect is on single-core systems versus dualcore systems.
Sure this effect goes down with any more than 2 cores but its still faster.
The cause is any OS use the second core for so many tasks that the first core do have more power to calculate the "Single-thread-task". There are also many optimizations on a multi-core system to improve the single thread performance. On Modern operating systems a Dual-core is the minimum to chose to get the best single-thread performance result. This effect is ~20-30%

[maldorordiscord]

The performance per watt is very good. Measured by the SPEC CPU2000 benchmark a Loongson 3A at an average of ~ 9.26 W power consumption perform 788 points in the peak
compared to the Phenom II x4 955BE with a TDP of 125 watts and 2797 points in the peak the performance per watt, more than twice as high.

sources:
http://bbs.lemote.com/archiver/?tid-50629.html
http://www.hardwarezone.com.sg/revie...ec-cpu2000-v13

[smitty3268]

The performance per watt is very good. Measured by the SPEC CPU2000 benchmark a Loongson 3A at an average of ~ 9.26 W power consumption perform 788 points in the peak
compared to the Phenom II x4 955BE with a TDP of 125 watts and 2797 points in the peak the performance per watt, more than twice as high.

sources:
http://bbs.lemote.com/archiver/?tid-50629.html
http://www.hardwarezone.com.sg/revie...ec-cpu2000-v13

TDP != average power consumption.

[smitty3268]

Single thread applications run faster on multi-core systems this is a fact!

Nope, you fell into his trap.

Single-core systems are often faster than a multi-core system, simply because they can clock up to a higher speed. Because they can spend the entire power/heat budget on that 1 core without having to worry about others joining in.

Clever algorithms can be made to speed up cores when running single-threaded apps only, but it's still likely not as good as a CPU that is completely dedicated to a single core would be.

Also, you can run into issues with a process being bounced around from core to core if the OS isn't smart about it, which can cause excessive cache misses, etc. that a single core CPU doesn't have to worry about. It's especially a problem when not all the cores are equal - such as intel's hyperthreading or bulldozer's 2-int cores per package, which caused scheduling problems in OS's that weren't fully aware of the architecture.

And of course your OS may have to do a lot more locking in order to be safe to run on multiple cores, that it can skip on a single-core system. That's extra overhead if you don't need it.

On Modern operating systems a Dual-core is the minimum to chose to get the best single-thread performance result. This effect is ~20-30%

You don't get a 30% performance boost on single threaded apps in a multi-core CPU. That's just stupid talk. Unless your single-threaded CPU is 30% busy all the time, even when you are idling. Maybe if you are constantly playing youtube videos in the background or something it might be true.

[maldorordiscord]

TDP != average power consumption.

Sure but the TDP of the Loongson3A is 15 watt.
The ACP of the amd cpu is ~90watt
9ACP vs 90 ACP or 15watt TDP vs 125watt TDP

the result is always the same the loongson is better per watt.

[smitty3268]

Sure but the TDP of the Loongson3A is 15 watt.
The ACP of the amd cpu is ~90watt
9ACP vs 90 ACP or 15watt TDP vs 125watt TDP

the result is always the same the loongson is better per watt.

I don't doubt loongson is better per watt - the lower TDP chip almost always wins those comparisons.

I'm just saying your numbers are crap and don't say much.

[maldorordiscord]

Nope, you fell into his trap.

Single-core systems are often faster than a multi-core system, simply because they can clock up to a higher speed. Because they can spend the entire power/heat budget on that 1 core without having to worry about others joining in.

Clever algorithms can be made to speed up cores when running single-threaded apps only, but it's still likely not as good as a CPU that is completely dedicated to a single core would be.

Also, you can run into issues with a process being bounced around from core to core if the OS isn't smart about it, which can cause excessive cache misses, etc. that a single core CPU doesn't have to worry about. It's especially a problem when not all the cores are equal - such as intel's hyperthreading or bulldozer's 2-int cores per package, which caused scheduling problems in OS's that weren't fully aware of the architecture.

And of course your OS may have to do a lot more locking in order to be safe to run on multiple cores, that it can skip on a single-core system. That's extra overhead if you don't need it.

You don't get a 30% performance boost on single threaded apps in a multi-core CPU. That's just stupid talk. Unless your single-threaded CPU is 30% busy all the time, even when you are idling. Maybe if you are constantly playing youtube videos in the background or something it might be true.

Be sure i did not fell into his trap! I know many tests single core vs dual-core cpu with only single-thread workload.
I gave him the right answer not the answer he expected and not the answer he think what is right.
Higher clock speed is a invalid argument for an cpu like the Loongson because the cpu design do not allow you to clock higher and back in the time of the athelon64 dualcores vs singlecore the cpu design also do not allow a higher clock speed you can clock the dualcore versions as high as the single clock version if your cooling system is good. because its the cpu design what limit your clock speed.
Your algorithms argument is also invalid because you can speedup a single thread so much by multicore assisted calculations that no single-core cpu can compete with that.
You can do Speculative calculations on the second core to speed up.
You can use the second core to assist caching the hard-drive and pre-caching the hard-drive.
You can accelerate by driving hardware like mouse and keyboard and the graphic card with the second core.
An stupid os is also a invalid argument because linux is not stupid. we do not use windows!
Intel's hyperthreading is complete bullshit and the bulldozer pseudo core system is also complete bullshit.
"Maybe if you are constantly playing youtube videos in the background or something it might be true."
Its always true because the second core do: pre-caching, Speculative calculations, driver handling soundcard,graphiccard,mouse,keyboard,raid, sata, harddrives, encoding cryptography, and so one and so one.
You can also improve the single thread performance with multicore by using thread level speculation (TLS)
The second core also can load the memory ram speculative from the hard-drive.
You can Boosting Single-thread Performance in Multi-core Systems through Fine-Grain Multi-Threading.

and so one and so one. you are just wrong.

[smitty3268]

Be sure i did not fell into his trap! I know many tests single core vs dual-core cpu with only single-thread workload.
I gave him the right answer not the answer he expected and not the answer he think what is right.
Higher clock speed is a invalid argument for an cpu like the Loongson because the cpu design do not allow you to clock higher and back in the time of the athelon64 dualcores vs singlecore the cpu design also do not allow a higher clock speed you can clock the dualcore versions as high as the single clock version if your cooling system is good. because its the cpu design what limit your clock speed.
Your algorithms argument is also invalid because you can speedup a single thread so much by multicore assisted calculations that no single-core cpu can compete with that.
You can do Speculative calculations on the second core to speed up.
You can use the second core to assist caching the hard-drive and pre-caching the hard-drive.
You can accelerate by driving hardware like mouse and keyboard and the graphic card with the second core.
An stupid os is also a invalid argument because linux is not stupid. we do not use windows!
Intel's hyperthreading is complete bullshit and the bulldozer pseudo core system is also complete bullshit.
"Maybe if you are constantly playing youtube videos in the background or something it might be true."
Its always true because the second core do: pre-caching, Speculative calculations, driver handling soundcard,graphiccard,mouse,keyboard,raid, sata, harddrives, encoding cryptography, and so one and so one.
You can also improve the single thread performance with multicore by using thread level speculation (TLS)
The second core also can load the memory ram speculative from the hard-drive.
You can Boosting Single-thread Performance in Multi-core Systems through Fine-Grain Multi-Threading.

and so one and so one. you are just wrong.

Nope, you did. You lose.