AMD A10-5800K Performance On Ubuntu Linux

[phoronix]

Phoronix: AMD A10-5800K Performance On Ubuntu Linux

Continuing on from last week's initial benchmarks of the AMD A10-5800K Trinity APU on Linux, the Trinity memory performance testing, and then more benchmarks of the Radeon HD 7660D integrated graphics, here is the large Ubuntu Linux comparison of the AMD A10-5800K compared to the previous-generation AMD A8 Llano APU, an AMD FX-Series Bulldozer, and several Intel Core i3/i5/i7 CPUs from the Sandy Bridge and Ivy Bridge families.

http://www.phoronix.com/vr.php?view=17980

[bug77]

So it's pretty competitive with the entry level from intel's previous generation. In all fairness, Ivy Bridge isn't much faster than Sandy Bridge in CPU performance. Though the CPU itself might be less power hungry in IB.

[jakubo]

"...(at least) better than Sandy Bridge / i3" it keeps echoing...

[schmidtbag]

I'm pretty disappointed with these results but I've noticed trinity is tremendously impacted by memory frequencies, almost too much, so I wonder how different the tests would be with the most optimal RAM frequency. With most other systems, different ram speeds might get another few seconds out of encoding or another few FPS in a game, but with trinity it can be the difference between it being as bad as i3 or better than i5.

What I also think is weird is how little of a difference overclocking makes. Overclocking doesn't seem worth it considering how much more heat and power is used in turn.

[Luke_Wolf]

I'm pretty disappointed with these results but I've noticed trinity is tremendously impacted by memory frequencies, almost too much, so I wonder how different the tests would be with the most optimal RAM frequency. With most other systems, different ram speeds might get another few seconds out of encoding or another few FPS in a game, but with trinity it can be the difference between it being as bad as i3 or better than i5.

What I also think is weird is how little of a difference overclocking makes. Overclocking doesn't seem worth it considering how much more heat and power is used in turn.

I'm not. The results are that a $130 AMD processor is performing in the same range as a $200 Intel processor, and as it's on the APU side of things honestly it's performing where it should be for this generation, and these results actually speak rather well for where Vishera will be landing. What's important is that Trinity is faster than Llano and competing with i3s and i5s, thus proving that the bulldozer design was the right move for those who were doubting. Steamroller should be where AMD is competitive or better in the high end in most areas as opposed to being a mixed bag because it's removing the scheduler bottleneck that was creating a massive hit to the performance.

[bug77]

...Trinity is ...competing with i3s and i5s, thus proving that the bulldozer design was the right move for those who were doubting...

Wishful thinking. It's only competing with i3's that are dual-cores with hyperthreading and a low-power version of the i5, clocked 1.3GHz below A10-5800K. On top of this, both are part of intel's previous generation.

[coder543]

Wishful thinking. It's only competing with i3's that are dual-cores with hyperthreading and a low-power version of the i5, clocked 1.3GHz below A10-5800K. On top of this, both are part of intel's previous generation.

Nope, I'm not buying your argument. Looking through these results, I'd say this is competitive with low end i5 Ivy bridge processors... not i3s. And, for well below the cost of those i5s, you're also getting a GPU with some punch in it. If you have a workload that combines OpenCL in the mix, you're suddenly competing with i7s, for well below cost.

My i5-2500K might outperform this in a few workloads, but it cost me $230 back in the day, and that's before buying my $200 GTX 460. To get competitive performance for one-quarter the cost is noteworthy. I think AMD should charge more for these than they do -- their profit margins need to increase, and they deserve to be rewarded for their hard work here.

Finally, what's your deal with clock speeds? Everyone who knows anything about CPUs can tell you that clock speeds are marketing hype. The numbers are worthless -- invalid. You can't use them for any comparative shopping unless you're studying processors of the same brand and generation. Sure, my car may have ten million cubic miles of volume, but if it's all dedicated to the trunk, you're gonna be mighty disappointed with the passenger capacity.

[HokTar]

@Michael:
Thanks a lot for the dolfyn and lammps benchmark profiles!

Any idea why AMD is so misereble there (including the himeno, too)? Could that be really just the lack of compiler optimisations?
I mean they're still making a lot of money on Opterons where these workloads are crucial...

[bug77]

Nope, I'm not buying your argument.

It's not an argument, it's just numbers - the new CPU only touches the 2500K or the 3470 in maybe 5 tests out of the 20+ tests listed. And I have counted here the tests where is comes within 10%, not necessarily beating Intel.

It seems we're using the same setup, I'm on a 2500K and a GTX460 as well. Though I paid a tad under $200 for the CPU Still going strong, but I may take the plunge and go for a 660Ti soon.

[nightmarex]

Did I miss where the ram speed was listed in the llano and trinity rigs? BTW you should also include OC'ed llano as most users can get ~3.4ghz & 760-800mhz (GPU clock) on air. Since it's supposedly it's competition and you showed OC results of trinity.