[<({D34TH})>]

Spoiler: Warning: Spoiler!

Alongside the A10-7700K, the A10-7850K is a recent addition to AMD's A-Series line of APUs and uses the existing FM2+ socket. By using the new 28nm manufacturing process, these new APUs are able to run faster than the previous generation, even when running at a lower clock frequency. In addition, improvements in the graphical portion of the APU increases the capabilities of the integrated R7 graphics controller. Interestingly, the clock speed of these new APUs are actually lower than the models they are replacing

Quote:

Overall, the AMD A10-7850K is a great improvement over the previous generation APUs in AMD's A-series line. Especially in terms of gaming performance when using the integrated graphics, it is about 10% faster than the previous generation APUs. If you compare the gaming performance to an Intel CPU, the graphical capability is so much better that it really is not even in the same league.

In terms of CPU performance, the A10-7850K is overall an improvement over the A10-6800K. Especially in PCMark 08, the A10-7850K was about 10% faster than the A10-6800K. Our GeekBench benchmark indicated lower single-core performance, but as more and more programs use multi-core technology, that is really not even much of an issue.

However, there are a few issues we have with the A-series line in general. First, while the gaming performance is better than Intel, it is only better if you are using integrated graphics. As soon as you add a discrete video card, an equivalently priced Intel-based system is going to give you overall better performance in most games. And while the gaming performance is very good for integrated graphics, if you consider yourself even a moderate gamer you likely would want to invest in a discrete video card.

Similarly, if you are mainly concerned about performance in programs that do not use the GPU, an Intel-based system is going to give you much better performance for your dollar. Finally, in terms of future upgrade potential, the A-series APUs hit a wall much, much sooner than their Intel equivalents. The A10-7850K is the fastest A-series APU currently available, yet the CPU performance is much lower than even the Intel Core i5-4440 which is only a mid-range Intel CPU.

SOURCE

 

[james8]

10% CPU and 10% GPU?

glad I got Richland and didn't wait.

 

[Redeemer]

Why does AMD continue to have weak CPU single threaded performance?

 

[Tsumi]

Steamroller is supposed to focus heavily on improving single-threaded performance, seems odd that there is no performance increase and in some cases performance decreases...

 

[monstercameron]

WHY AMD!?! why must you troll us?

 

[sugarhell]

This is not an 'legit' review. The nda is on 16 and a lot of these results looks fishy. And puget-system is pretty much unknown.Easy clicks

 

[Durquavian]

Really wish they would have run 1080p. Think the difference in even ram would have been higher.

Don't forget it is 3.7Ghz to 6800k 4.0Ghz, no mention of turbo.

 

[boot318]

Quote:

Originally Posted by Tsumi

Steamroller is supposed to focus heavily on improving single-threaded performance, seems odd that there is no performance increase and in some cases performance decreases...

Remember Richland is going to have an 400MHz clockspeed advantage stock. I'm not sure if that eases the pain though.

 

[SandGlass]

Geekbench guys? Seriously, why would they use that? Geekbench is junk, even Linus Torvalds says so. It's wildly inaccurate, and not reflective of real world performance at all.

Also people know cinebench uses an intel compiler. Only two benchmarks that would be really nice to see would be 3Dmark, and Sisoft sandra which uses the Microsoft Visual C++ compiler.
FTC regarding Intel affecting Cinebench

 

[Exilon]

Quote:

Originally Posted by SandGlass

Geekbench guys? Seriously, why would they use that? Geekbench is junk, even Linus Torvalds says so. It's wildly inaccurate, and not reflective of real world performance at all.

Also people know cinebench uses an intel compiler. Only two benchmarks that would be really nice to see would be 3Dmark, and Sisoft sandra which uses the Microsoft Visual C++ compiler.
FTC regarding Intel affecting Cinebench

Little do we know, the Intel compiler also has a piece of code in it that says if CPU == 'Kaveri', make it slower than Richland.

 

[SandGlass]

Good point, AMD needs to get their butts off the ground and start working, they will fail badly otherwise.

 

[Gnomepatrol]

Quote:

Originally Posted by SandGlass

Geekbench guys? Seriously, why would they use that? Geekbench is junk. Also people know cinebench uses an intel compiler. Only two benchmarks that would be really nice to see would be 3Dmark, and Sisoft sandra which uses the Microsoft Visual C++ compiler.

http://www.ftc.gov/enforcement/cases-and-proceedings/cases/2010/11/matter-intel-corporation-corporation

Ok so dump the intel chip out of the picture and just compare the Kaveri and Richland chips. Not much of a difference between the two. 5-10% for all the promising. The last chip I had from AMD that I was impressed with was the 1055t that I bought second hand on OCN. My FX 8320 was a nice chip at 5.0ghz too.

Sorry I am rambling, I just was really hoping to go back AMD, but with a Microcenter next to me I guess not

 

[<({D34TH})>]

I just hope people don't start hating me for just reporting whats available right now. I will be adding new reviews as they keep coming.

 

[IvantheDugtrio]

So the question now is if Kaveri is worth anything. Based on these and other similar benchmarks I would say it's only pro over Richland is the support for HSA and Mantle (with IGP). At $40 more than the 6800k I don't see a good reason to buy the 7850k.

Sure LibreOffice supports HSA to some degree but who would choose that over Microsoft Office?

Sometimes it's better to improve on existing technologies than to attempt to re-invent the wheel in the hopes that everyone will adopt it.

 

[<({D34TH})>]

Quote:

Originally Posted by IvantheDugtrio

So the question now is if Kaveri is worth anything. Based on these and other similar benchmarks I would say it's only pro over Richland is the support for HSA and Mantle (with IGP). At $40 more than the 6800k I don't see a good reason to buy the 7850k.

Sure LibreOffice supports HSA to some degree but who would choose that over Microsoft Office?

Sometimes it's better to improve on existing technologies than to attempt to re-invent the wheel in the hopes that everyone will adopt it.

People who prefer not to pay for Office.

 

[IvantheDugtrio]

Quote:

Originally Posted by <({D34TH})>

People who prefer not to pay for Office.

That's still a very small group of people.

 

[Yeroon]

Quote:

Originally Posted by IvantheDugtrio

That's still a very small group of people.

ah, now, I bet everyone would like to not pay for office

 

[mdocod]

In an attempt to stay positive about this product launch (sort of difficult looking at these results), I do wonder if Kaveri can hit clocks on the CPU and GPU like richland (or better), and perhaps push into the upper DDR3 speed territory (3000?). If so then there's more headroom to improve into with overclocking here, so this may prove to be better once overclocked by a more respectable margin. Wish they had done power consumption comparisons, as that would offer some insight into whether or not they delivered on the compute efficiency improvements or not.

Considering the fact that the same GPU architecture is currently packaged and sold with factory clocked speeds of up to ~1.1ghz (near 1.2ghz boost on some products), and this thing is only churning 720mhz base.... there's "theoretically" up to ~60% improvement potential in shader bound workloads.

I am often frustrated by the position that a benchmark is somehow not "valid" because it doesn't meet the criteria of the person criticizing it. Especially when the position is brought up in a situation where the benchmark reflects real world workloads regardless.
1. The real world is not necessarily optimized for AMD, why should the benchmarks be? Why is it everyone else's responsibility to "fix" the performance of a non-mainstream product? A benchmark that is heavily optimized for an AMD chip isn't likely to translate to accurate real world performance comparisons. I see more similarities than differences in results between "benchmark-software" and "application-benchmarks," which tells me that most "benchmark-software" is actually better than folks give it credit for.
2. There are very raw hardware level benchmarks that simply look at cache and memory performance, latency, bandwidth, etc, that wind up being very telling about how a CPU performs. Piledriver has poor cache latency performance and memory controller performance compared to the last 3 generations of Intel chips. There is no way to "optimize" or "compile" software that depends on these performance traits in such a way to overcome these failings. A chip with "bad" performance in these raw cache performance traits can't perform better than a chip with "good" performance in these traits, it simply isn't possible.

To bottom line it: When people don't like the results of the benchmarks, they look for ways to discredit them.

PS: I have never paid for office. Been using OpenOffice (or some variant of it) for as long as I can remember.

 

[AlphaC]

No HSA support means 10% improvements isn't out of the question...

also the Stilt said A88X is weaker, we see that in this bench's memory results I think
Quote:

The A88X chipset offers nothing new compared to A85X.
In fact it is worse than it's predecessor.

A88X has an upgraded AHCI / RAID controller which makes it unable (as it currently appears) to run over 106MHz BCLK while AHCI is enabled. Personally I would rather see the top tier FM2+ motherboard using A85X FCH instead of the A88X.

http://www.xtremesystems.org/forums/showthread.php?288118-FM2-Chipsets&p=5221118&viewfull=1#post5221118

 

[Tsumi]

Quote:

Originally Posted by AlphaC

No HSA support means 10% improvements isn't out of the question...

20-30% claimed IPC increase with 10% lower clock speeds means at least 10% performance increase in everything non-HSA related. HSA supported software should see gains far greater than 10%, I would say 40% at minimum.