AMD A10-5800K APU for Socket FM2

[brandonwh64]

Once again, Great review dave! Keep up the good work!

[Fourstaff]

Originally Posted by Frick

IMO we still have. Throw some light gaming on it and it'll still deliver.

Would recommend i3+discrete more often actually, you can get quite a lot more power for slightly more money (ok, $100+ more).

Originally Posted by Mussels

my laptop has an A6 APU, and i can play starcraft II on it. getting serious gaming done on integrated graphics is pretty mind blowing.

My POS laptop can play ScII, your point is invalid lol

[james888]

I can not find a good review with starcraft 2 as the benchmark. A person wants me to build a computer for them that can play starcraft 2 on high, 1080p for the cheapest possible option. I don't think this is enough, but maybe with a 6670.

[Frick]

Originally Posted by Fourstaff

Would recommend i3+discrete more often actually, you can get quite a lot more power for slightly more money (ok, $100+ more).

Or even more power than that if you get an A10 + HD 6670 for the same money.

That technology isn't perfect though but in some games it works quite well:

[Crap Daddy]

Originally Posted by Frick

Or even more power than that if you get an A10 + HD 6670 for the same money.

That technology isn't perfect though but in some games it works quite well:

http://www.nordichardware.se/images/...tlefield_3.png

http://www.nordichardware.se/images/.../DG_Skyrim.png

Now come on, who would shell more money for a HD6670 when an i3 Ivy + a HD7770 for the same money would be better than the crossfired APU. I think here lies the major problem for this chip. While undoubtedly the graphics is great it just isn't enough for a desktop part used for gaming. On a desktop you don't do "light" or even "lighter" gaming. You don't use 720p resolution. That's for laptops. As for a HTPC, maybe I'm wrong, I don't see an advantage for this chip compared with an i3 apart from needing much more power to achieve the tasks.

[Mussels]

having used APU crossfire i can say its not that great. thats not the real draw here - the real draw, is that the onboard video in entry level systems kicks the pants off intels.

[3870x2]

One issue with the PC gaming market is you can't just go out and buy any computer and play a game on it.

Now you can. If something like this had been available since early 2000s, I believe the PC gaming market would be much greater than the console market.

People get discouraged when they see a game on the PC they want to play, then buy it and find out they can't even play it on the lowest settings.

It is better late then never i guess, and this is a gigantic step forward. Too bad this had to come out in the decline of the PC as of the last 5 years. Laptops and tablets have taken over a large chunk of that market.

[SIGSEGV]

thanks dave for this great review

Originally Posted by Mussels

having used APU crossfire i can say its not that great. thats not the real draw here - the real draw, is that the onboard video in entry level systems kicks the pants off intels.

yeah, you're right, i have a dell optiplex 990 with i5-2500 and HD2000 graphic, i can't even play my most favourite games like Dota 2 and Counter-Strike: Global Offensive , it drives me crazy..

[CAT-THE-FIFTH]

There is also the 65W TDP A10-5700 which is only slightly lower clocked. In the UK it is around £90 which is around the same price as the 100W TDP A10-5800K.

[cadaveca]

Originally Posted by Absolution

Tl;dr: Wanted to see these "new" 7000 series cores to be paired with 7000 series cards in dual graphics mode.

I did too, and almost bought a 7770 for this review, but alas, it's 6670 only. 6570 with A8 chips now.

Originally Posted by Absolution

I also dont understand why there is an inconsistency in the comparison of benchmarks.

Why is an intel mobile chip included in the power test?
http://tpucdn.com/reviews/AMD/FM2_AP.../power_eps.gif

Why is the A10-5800 (with DG) included with no comparison with a 3850 (with DG)?
http://tpucdn.com/reviews/AMD/FM2_AP..._full_load.gif

In some cases, the 3850 with DG is there but not the A10 with DGL...
http://tpucdn.com/reviews/AMD/FM2_AP...ebench_gpu.gif

The Intel mobile chip has what some would consider twice the GPU power(7850m = 7770), plus four threads of IVB goodness. It's power consumption is quite similar to the APUs, for the full system too, but the desktop version of these same chips, i5 3220 and 7770, cost a fair bit more than the APU.

And in some workloads, the APUs beat it out. Just a reference number. It's truly just as out of place as the 3770K is, but I only have so many parts on hand to test with, so there it goes.

The desktop that uses those parts(yes, a desktop), will be reviewed in coming weeks as well!
If you'd liek other testing, I'm always open to taking donations for hardware purchases.

Originally Posted by Mussels

reading the review and enjoying it, but the intro seems kinda long and rambling.

I put the real conclusion first, hence the "Rambling".

Originally Posted by CAT-THE-FIFTH

There is also the 65W TDP A10-5700 which is only slightly lower clocked. In the UK it is around £90 which is around the same price as the 100W TDP A10-5800K.

I am quite interested in these chips myself, but am not sure how OC'ing with them will be. That 35 Watts less power would be perfect to me, if you can still clock the iGPU.

[Vinska]

In these:



i7 3770k & FX-4100; both paired with the same GPU. 'dozer is slightly ahead of ivy...
...But people said 'dozer is shi-i-i-i-it! WTH? This is outrageous!

(╯° · °）╯︵ ┻━━━━┻

[Dent1]

Originally Posted by Vinska

In these:

http://tpucdn.com/reviews/AMD/FM2_AP...es/f1_2010.gifhttp://tpucdn.com/reviews/AMD/FM2_AP...s/sniperv2.gif

i7 3770k & FX-4100; both paired with the same GPU. 'dozer is slightly ahead of ivy...
...But people said 'dozer is shi-i-i-i-it! WTH? This is outrageous!

(╯° · °）╯︵ ┻━━━━┻

Next time don't listen to rumours

There are plenty of gaming and non-gaming scenarios where Bulldozer easily beat out Sandy and Ivy. Then you'll look at the next set of charts and Bulldozer will be at the bottom of the pile. Bulldozer was inconsistant from one set of charts to the next.

[Vinska]

Originally Posted by Dent1

Next time don't listen to rumours

There are plenty of gaming and non-gaming scenarios where Bulldozer easily beat out Sandy and Ivy. Then you'll look at the next set of charts and Bulldozer will be at the bottom of the pile. Bulldozer was inconsistant from one set of charts to the next.

If You rolled a one on Your "spot jokes" skill check, then:
That was mostly a joke

[kukreknecmi]

What i dont understand is, how can theese power consumptions from wall can be true. I have never saw a FX or i7 which consumes less then 10W!!! Without a discrete card, shouldnt theese numbers be around 40-50W ?



The more weird thing is, how can an FX 4100 with 6670 can consume 8W from wall??

[Fourstaff]

Originally Posted by kukreknecmi

What i dont understand is, how can theese power consumptions from wall can be true. I have never saw a FX or i7 which consumes less then 10W!!! Without a discrete card, shouldnt theese numbers be around 40-50W ?

The more weird thing is, how can an FX 4100 with 6670 can consume 8W from wall??

Not from the wall, from the 8pin CPU header

[cadaveca]

Originally Posted by kukreknecmi

What i dont understand is, how can theese power consumptions from wall can be true. I have never saw a FX or i7 which consumes less then 10W!!! Without a discrete card, shouldnt theese numbers be around 40-50W ?

http://tpucdn.com/reviews/AMD/FM2_AP...power_full.gif

The more weird thing is, how can an FX 4100 with 6670 can consume 8W from wall??

The Intel i5 3210m and 7850m can drop to 1 W from the wall, idle, desktop still functional. It varies, with most boards, so I ahve a specific timed interval when that is measured. If board software keeps numbers higher, then I'll catch it, and investigate.


I know, it's crazy, I expected 75 W or so, but nope. I do take pictures of the meters too, just for posterity, so I do have photographic evidence! I am ysing jsuit a $20 wall-wart meter for that numbertoo, so it might not be 100% accurate, but it should be pretty close.

Bulldozer and Piledriver can disable un-used modules, what's really surprising is that those three board managed to drop so low on idle, while the APU boards did not. The FX-4100 was installed to an ASUS Crosshair V Formula, same ram as the APUs, same drives, etc...

Originally Posted by Fourstaff

Not from the wall, from the 8pin CPU header

Nope, that's from the wall. The other graph is from the 8-pin.

[Fourstaff]

Originally Posted by cadaveca

Nope, that's from the wall.

My bad I thought you measured everything through the 8pin

[cadaveca]

Originally Posted by Fourstaff

My bad I thought you measured everything through the 8pin

I used to, but now I include graphs for each, as I had requests to keep 8-pin, but add full system.

[kukreknecmi]

Originally Posted by Fourstaff

Not from the wall, from the 8pin CPU header

This is from 8pin


This is full system


So isnt full system means from wall?

[trickson]

Originally Posted by Assimilator

A10-5800K @ 4.4GHz / 100 W can't touch i5 2500K @ 3.3GHz / 95W in multi-threaded scenarios.

WOW just wow.
So even @4.4GHz this CPU is struggling just to make it? WOW just WOW!

[rangerone766]

Originally Posted by 3870x2

One issue with the PC gaming market is you can't just go out and buy any computer and play a game on it.

Now you can. If something like this had been available since early 2000s, I believe the PC gaming market would be much greater than the console market.

People get discouraged when they see a game on the PC they want to play, then buy it and find out they can't even play it on the lowest settings.

It is better late then never i guess, and this is a gigantic step forward. Too bad this had to come out in the decline of the PC as of the last 5 years. Laptops and tablets have taken over a large chunk of that market.

i couldnt have said things better!

[kukreknecmi]

Originally Posted by cadaveca

Bulldozer and Piledriver can disable un-used modules, what's really surprising is that those three board managed to drop so low on idle, while the APU boards did not. The FX-4100 was installed to an ASUS Crosshair V Formula, same ram as the APUs, same drives, etc...

Thx for info. Its a bit offtopic but yet its .. i dunno what to say, just crazy. fx4100 + 6670 and sub 10W lol. I wonder how much it goes up within very light to light load. Like 720p movies, or some web browsering(it's a bit hard to measure web browsing tho, need average). I guess i should expect sub 30w then?

[trickson]

At least AMD is doing some thing with there APU's Not like Intel has any thing to offer in that respect. I like what I see for sure.

[cadaveca]

Originally Posted by kukreknecmi

Thx for info. Its a bit offtopic but yet its .. i dunno what to say, just crazy. fx4100 + 6670 and sub 10W lol. I wonder how much it goes up within very light to light load. Like 720p movies, or some web browsering(it's a bit hard to measure web browsing tho, need average). I guess i should expect sub 30w then?

I am sure that part of the power savings there is how the VRM works on those boards, where it can disable phases when there is no need for them. It is possible that "out-of-the-box" the Intel BIOSes have these features enabled, but the AMD APUs do not. I measure at the 8-pin, and hte wall only, so where that actual power goes, I can only guess. Differences on installed parts, like LAN and such will ahve an effect as well. Both FM1 and FM2 APU boards are top-level Gigabyte products, A75-UD4H and the F2A85X-UP4, both have VRMs that should be capable ofthe same thing as the Intel boads.

The ASUS HD 6670 might have some fancy power-saving stuff too, I'm not sure. W1zz does GPUs, not me, so those sorts of things I gotta ask him about, even!


I didn't have a lot of time to do all those benchmarks...35 graphs, 6 systems with 8 configurations, is like over 200 tests. Plus overclocking, OS installs, blah, blah, blah...

It might be possible to get <30W, but i think that if you really want to go that low, there might be better options. Piledriver-based CPUs might make that even better...I haven't got any of those yet, so time will tell, either when AMD sends one, or I go buy one. I do plan on having FX4100 and FX8150 in that review, so I can take a look at light desktop use power draw, for sure. I might even just add a graph with that, might be interesting info.

[kukreknecmi]

With an A6 3500 undervolted, paired with matx board, ssd and pico-psu, idle consumptions can go as low as 12.5w idle and 28-38W at 1080p movies.

Trinity chips may be do similar, with custom p-states. A6 5400 and A4 5300 may be interesting to htpc builders, and will be interesting to compare 2-3Core Lianos.