The Official Kaveri Review Thread (A10-7850K, etc)

[AtenRa]

Also, for people that "didnt" like the games used in the first review,

Free to Play Games: AMD A10-7850K APU vs. Core i3 + GT 630


Dota 2, Warframe, LoL, World of Tanks, World Of warplanes

 

[SPBHM]

You still dont get it right ??? for the same price Intel Core i3 needs a dGPU to compete and still is slower in Gaming. Not to add that the power consumption both in Idle and full load in Gaming will go up substantially and you have another device with another fan in the PC case.

GT 630 is pointless, see this again

http://i.imgur.com/YQrIp5y.png

it doesn't even add a significant enough advantage over the HD 4600,
so Pcper should have compared the 7850K to an i5 4430 (closest priced Intel CPU+IGP), and we know how it would look, one much faster CPU and the other much faster IGP, if they wanted to have a significant comparison with cheaper CPU + VGA vs 7580K they made an awful choice with that 4330+GT 630.

 

[Arkaign]

You still dont get it right ??? for the same price Intel Core i3 needs a dGPU to compete and still is slower in Gaming. Not to add that the power consumption both in Idle and full load in Gaming will go up substantially and you have another device with another fan in the PC case.

I fully agree that Kaveri >>> i3 + $53 GPU for gaming.

However, does it make any sense to buy a Kaveri for gaming? No, no it doesn't unless you *must* have a SFF case and no dGPU at all. Richland + 7750/7770/7790 obliterates i3 + cheap GPU or Kaveri period, for similar $$.

Kaveri is the best of the niche, but that niche doesn't make a lot of sense to me.

A BIG BIG caveat to that though is that I'm purely speaking of desktop users. For laptop users, the page opens much better for Kaveri products if things work out. Let's recap though.

Small budget people :

(A)- Wants an HTPC build. Doesn't game.
(B)- Wants an HTPC build, does game.
(C)- Wants a budget gaming build.
(D)- Wants a budget office/social media/email PC.

Okay.

Group A is a small market. Still, Kaveri makes no sense to that group, at least the $100+ SKUs.

Group B is another small market, but COULD be a market for Kaveri provided the user demands a SFF case. As a personal example, I run an FX6350 + 7870 in my HTPC currently, and I don't care in the least that it's not super slim.

Group C is an entirely lost market for Kaveri because it's simply bad value (ditto most Intel options honestly). AMD's other products are far more compelling because dGPU is still the winner for respectable gaming, even on a budget.

Group D is another entirely lost market for Kaveri because it's a poor match for that balance of midrange CPU + overkill iGPU for desktop apps. A great example is my largest law firm, 95% of the people are on IGP, and nobody, I repeat nobody ever says 'wow I wish my video performance was better'. For web, email, office suite, quickbooks, timeslips, yadda yadda, even relatively ancient stuff like AMD 4250 and Intel GMA 4500 is more than enough power. The handful of people who are super power users with 2 27" monitors use things like Intel HD 4600 or AMD 5450. Again, nobody complains in the least because that's already more than enough power for what they do.

When you get to laptops though, now THAT opens the door to a bunch of people who would like something that can game, but don't have a lot of money. Unlike the desktop world, there's no cheap 1080p gaming combo like a 760k + 7790. You have a bunch of low end garbage, and then super expensive stuff. Something with ~7850k performance on a 1366x768 14" monitor for $500ish could be a big hit. College students, gamers who want a notebook they can use on the road that's not useless, etc.

It remains my opinion that it's a waste with current pricing on the desktop side of things, a niche inside a niche, too expensive to make sense. I concede that that's merely my opinion, but I do not think I'm alone.

 

[Shivansps]

The test should have been G3220+HD7790 and also the 760K+HD7790, you can actually build that for the same price.

Also with the steam in-home streaming ill say that gaming HTPC is not longer needed.

 

[Arkaign]

The test should have been G3220+HD7790 and also the 760K+HD7790, you can actually build that for the same price.

Also with the steam in-home streaming ill say that gaming HTPC is not longer needed.

Well, that's good for some people (with room for both a gaming PC and extra PCs/TVs/etc).

A college student might not have much room to fit two systems in, so a SFF HTPC/Gaming/Homework box fits the bill decently. Sure we're not talking a major market there, but it's something at least.

Fully agreed about the better dGPU vs testing though, it was incredibly lame to compare it the way it was done. In particular the 760K + 7790 is a stout combo. I know someone with an older 750K + 6770, and it's really quite nice, and works flawlessly with their 22" 1680x1050 display. He can play almost anything on high/very high, with only AA scaled back a tiny bit.

 

[coercitiv]

A10-6800k is on the far side of the performance power curve. Compare to the a10-6700 and they are roughly the same.

I compared A10 6800K to A10 7850K. If you want to compare to A10 6700, please do, but do it against it's 7700 brethren. If the efficiency curve is in favor of Kaveri on the bottom and top TDPs, i'd be a bit surprised to see Richland do better at 65W.

 

[AtenRa]

With your logic people, AMD and Intel should only release two SKUs. A cheap $40-50 (Celeron/Pentium) and one at $200-250 (Core i5). There is no need for all those SKUs ranging from $40 all the way up to 1K. 🙄

Also,

From Steam Survey Dec 2013, there are countles of people playing Games using the GT630 or slower iGPUs/dGPUs. The two reviews above really shows what you get with A10-7850K as a whole package. People making the mistake comparing Kaveri as a CPU only, this is the most advanced APU in its category. We all know that there are better options for a strictly Gaming system, that doesnt invalidate the data of the review. And that is, as an APU Kaveri is the best choice by far giving you the best price/performance and lower power consumption in its class.


1: Intel HD Graphics 4000
4.82%

2: Intel HD Graphics 3000
4.25%

3: Intel HD Graphics 2000
2.25%

4: NVIDIA GeForce GTX 660
2.20%

5: NVIDIA GeForce GTX 550 Ti
2.00%

6: NVIDIA GeForce GTX 560 Ti
2.00%

7: NVIDIA GeForce GTX 650
1.50%

8: Intel Ironlake (Mobile)
1.40%

9: NVIDIA GeForce GTX 560
1.29%

10: NVIDIA GeForce GTX 670
1.29%

11: NVIDIA GeForce GT 630
1.25%

15: Mobile Intel 4 Series Express
1.16%

19: NVIDIA GeForce 9500
1.08%

27: NVIDIA GeForce 210
0.93%

31: NVIDIA GeForce GT 630M
0.91%

36: Intel HD Graphics 2500
0.85%

39: NVIDIA GeForce GT 220
0.73%

41: NVIDIA GeForce GT 620M
0.72%

42: ATI Radeon HD 5450
0.70%

44: ATI Mobility Radeon HD 5470
0.67%

 

[Erenhardt]

What I like about kaveri is it boosts the low end quite a bit.
What I mean by that? DualGraphics kaveri + r7 240/250.
gt630 and gt640 equivalent cards!
Suddenly a very bottom card performs somewhat close to the mainstream (from 7750 to more than 7770 depending on what is paired together).
But I agree 7850k it could be priced better. Well, it is nothing new that top of the line sku is not the best value.
but as pointed previously $130 microcenter deal is amazing! and not really expected for such a new product.

 

[Enigmoid]

I compared A10 6800K to A10 7850K. If you want to compare to A10 6700, please do, but do it against it's 7700 brethren. If the efficiency curve is in favor of Kaveri on the bottom and top TDPs, i'd be a bit surprised to see Richland do better at 65W.

Also out of all the richland products the a10-6800k probably has the lowest efficiency as it has such high clocks. And when you look at the a10-6700 it has nearly identical performance at much lower power consumption.

 

[coercitiv]

Also out of all the richland products the a10-6800k probably has the lowest efficiency as it has such high clocks. And when you look at the a10-6700 it has nearly identical performance at much lower power consumption.

Also out of all the Kaveri products the A10-7850K probably has the lowest efficiency as it has such high clocks. And when you look at A10-7700 it has nearly identical performance at much lower power consumption.

 

[frozentundra123456]

I fully agree that Kaveri >>> i3 + $53 GPU for gaming.

However, does it make any sense to buy a Kaveri for gaming? No, no it doesn't unless you *must* have a SFF case and no dGPU at all. Richland + 7750/7770/7790 obliterates i3 + cheap GPU or Kaveri period, for similar $$.

Kaveri is the best of the niche, but that niche doesn't make a lot of sense to me.

A BIG BIG caveat to that though is that I'm purely speaking of desktop users. For laptop users, the page opens much better for Kaveri products if things work out. Let's recap though.

Small budget people :

(A)- Wants an HTPC build. Doesn't game.
(B)- Wants an HTPC build, does game.
(C)- Wants a budget gaming build.
(D)- Wants a budget office/social media/email PC.

Okay.

Group A is a small market. Still, Kaveri makes no sense to that group, at least the $100+ SKUs.

Group B is another small market, but COULD be a market for Kaveri provided the user demands a SFF case. As a personal example, I run an FX6350 + 7870 in my HTPC currently, and I don't care in the least that it's not super slim.

Group C is an entirely lost market for Kaveri because it's simply bad value (ditto most Intel options honestly). AMD's other products are far more compelling because dGPU is still the winner for respectable gaming, even on a budget.

Group D is another entirely lost market for Kaveri because it's a poor match for that balance of midrange CPU + overkill iGPU for desktop apps. A great example is my largest law firm, 95% of the people are on IGP, and nobody, I repeat nobody ever says 'wow I wish my video performance was better'. For web, email, office suite, quickbooks, timeslips, yadda yadda, even relatively ancient stuff like AMD 4250 and Intel GMA 4500 is more than enough power. The handful of people who are super power users with 2 27" monitors use things like Intel HD 4600 or AMD 5450. Again, nobody complains in the least because that's already more than enough power for what they do.

When you get to laptops though, now THAT opens the door to a bunch of people who would like something that can game, but don't have a lot of money. Unlike the desktop world, there's no cheap 1080p gaming combo like a 760k + 7790. You have a bunch of low end garbage, and then super expensive stuff. Something with ~7850k performance on a 1366x768 14" monitor for $500ish could be a big hit. College students, gamers who want a notebook they can use on the road that's not useless, etc.

It remains my opinion that it's a waste with current pricing on the desktop side of things, a niche inside a niche, too expensive to make sense. I concede that that's merely my opinion, but I do not think I'm alone.

I agree totally. I am not interested in a SFF desktop, so Kaveri holds zero interest for me there. More gpu power than the casual user will ever need, but not as much as a cheaper cpu paired with a discrete card. Cpu performance is adequate for casual use, but so is a pentium, or celeron that is cheaper and more power efficient. And if you want strong cpu performance, then get an FX or i5.

But I am somewhat interested in its application for mobile, depending on how much graphics performance has to be sacrificed to fit into a mobile package. It also depends on price, as current A10 models seem to run around 600.00, which is getting uncomfortably close to something with a decent discrete card. My grandson (a teenager) currently has an old gaming laptop with a core 2 duo and a GTX 260m, which I might replace with something like a Kaveri model if the performance were similar or somewhat improved and the price was suitably lower than an i7 with a discrete card like the GT 650m or 750m.

 

[Enigmoid]

Also out of all the Kaveri products the A10-7850K probably has the lowest efficiency as it has such high clocks. And when you look at A10-7700 it has nearly identical performance at much lower power consumption.

Toms

techreport

http://www.hardware.fr/articles/913-9/cpu-consommation-overclocking.html

Can't really find any other tests but

Uses little less power as the a10-7850k but is a little slower too.

 

[ashetos]

Toms

techreport

http://www.hardware.fr/articles/913-9/cpu-consommation-overclocking.html

Can't really find any other tests but

Uses little less power as the a10-7850k but is a little slower too.

These benchmarks are not so fair concerning power efficiency for 2 reasons:
1. They are CPU-only benchmarks and the product is an APU.
2. They complete a workload once and race to idle counts as total platform power is minimized due to smaller runtime of experiment.

These two limitations can be circumvented by a single benchmark, PC games:
1. PC games utilize the whole APU.
2. PC games are not a one-time run, but a throughput benchmark, where performance equates FPS for a fixed detail level; this way, race to idle doesn't count.

As long as VSYNC is off, PC games can be a great benchmark of power efficiency of the chip.

I would expect the 45W cTDP chip to be considerably more efficient than the 95W TDP chip.

 

[Enigmoid]

These benchmarks are not so fair concerning power efficiency for 2 reasons:
1. They are CPU-only benchmarks and the product is an APU.
2. They complete a workload once and race to idle counts as total platform power is minimized due to smaller runtime of experiment.

These two limitations can be circumvented by a single benchmark, PC games:
1. PC games utilize the whole APU.
2. PC games are not a one-time run, but a throughput benchmark, where performance equates FPS for a fixed detail level; this way, race to idle doesn't count.

As long as VSYNC is off, PC games can be a great benchmark of power efficiency of the chip.

I would expect the 45W cTDP chip to be considerably more efficient than the 95W TDP chip.

1. I am looking at CPU performance only. Sorry if I wasn't properly explicit.

2. I'm looking at pure efficiency on a given task. I agree that low power CPU's get unfairly penalized by idle power but idle power does exist and for a mobile platform this matters in the sense of how much of my battery will be used for a given task (though obviously higher draw will drain the battery faster).

45W will be much more efficient. I agree.

But the proper richland vs. kaveri efficiency we need a comparison at the same clocks, kaveri will perform better and we'll see how much the power draw differs. Richland was really pushed to the max and fully utilizes its boost capabilities. Kaveri is clocked more conservatively (no reason to as most reviews show hitting 4.2 ghz fairly easily) and that is a problem.

 

[burninatortech4]

These new APU's are amazing for the lower end of the spectrum. Especially from a laptop hardware perspective and even more so when the price becomes more realistic (7850k -129.99 at CA Microcenter!) If the newegg price drops that low sales should explode, especially for the a8's (you'd think AMD would realize this).

 

[Deceneu]

There is an amount of reviews around the web about Kaveri ,but none is complete in my view.

Never seen new tech so much ignored.

Overclocking and clocks observed during usage infromations are almost non existent so is testing with various cards or trying crossfire.

Even the small amount of game tests is useless for someone that wants ot buy.

Maybe techpower-up to do some real gaming tests but they don t have any review.

After reading a few i don t know for example what clocks where found for the NB and how does the the turbo and CnQ work with the actual motherboards.

Having in mind the way FM2 boards performed by default ,and i own 2 boards from MSI and GIgabyte i can assume the real performance was not revealed by this reviews.
The NB clocks for example were bouncing to even 1500Mhz.
Imagine the performance of the iGPU in this conditions even if high speed RAM was used ,which was not the case in many reviews.

Why are this hardware testing sites so limited in delivering information to the public.
Xbit for example was the most ignorant of them all.

Also in most tests they ve used Gigabyte boards that throttle the heck out of AM3 + and FM2 CPU-s in the last 3-4 years.The BIOS-es this guys made in the last 3 years are a shame for this big name in the industry.

 

[mnewsham]

Picked up the A10-7850k and the Asus a88x-pro from microcenter as a bundle deal for $200, then I picked up 8GB(2x4GB) DDR3 2400Mhz CAS11 RAM for $75, Thermaltake case for $30 and a new PSU for $45. Going to pick up a decent HSF when everything is delivered later this week or early next week.

Total upgrade cost was less than $400 and I have the option of adding an r7 260 or an r9 270x down the line if the IGPU isn't enough for me. (though for 90% of what I do it should be)

 

[SPBHM]

digital foundry tested the 7600

http://www.eurogamer.net/articles/digitalfoundry-2014-amd-a8-7600-kaveri-review

nothing new but, an interesting video comparing the framerate from different memory speed in real time during the gameplay

solid performance at 720p medium for all their games,

 

[bgt]

digital foundry tested the 7600http://www.eurogamer.net/articles/digitalfoundry-2014-amd-a8-7600-kaveri-reviewnothing new but, an interesting video comparing the framerate from different memory speed in real time during the gameplaysolid performance at 720p medium for all their games,

Nice chip the 7600:whiste: I have the 6800 but it uses far too much power. I am really spoiled now by the Haswell chips with their low power draw🙂

 

[bgt]

Oops, the 7600 is not yet available hereD: pity

 

[AtenRa]

Cinebench R11.5:

• AMD Jaguar A6-5200 2.0GHz (4cores) : 1.99 pts
• AMD Steamroller A10-7850K @ 2.0GHz (4cores): 1,90 pts

AMD Steamroller A10-7850K @ 2.0GHz + 2× 4 GB DDR3-2400MHz, 11-13-13-30-2T dual-channel
AMD Jaguar A6-5200 2.0GHz + 8 GB G.Skill Trident 2400MHz @ 1600MHz (2x4GB) single-channel

source: http://forum.cnews.cz/viewtopic.php?f=2&t=25865

Ehm, from the same link Kaveri at 2GHz is faster in single thread and same performance with 4 threads against Kabini at 2GHz.

Kaveri (Steamroller) @ 2GHz

Kabini (Jaguar) A6-5200 @ 2GHz

 

[AtenRa]

I know but we know only notebook result AMD A6-5200 for single-threaded (1xCPU): 0,50 pts.
Desktop (Mini-ITX) with better cooling can be in single-threaded (1xCPU) higher.

I dont expect single thread to gain anything from mobile to desktop. I dont believe mobile A6-5200 throttles in single thread.

 

[Enigmoid]

Ehm, from the same link Kaveri at 2GHz is faster in single thread and same performance with 4 threads against Kabini at 2GHz.

Kaveri (Steamroller) @ 2GHz

Kabini (Jaguar) A6-5200 @ 2GHz

Basically the same and basically the same. 0.03 points is basically the variation between runs. Jaguar is a pretty nice design.

 

[pw257008]

I dont expect single thread to gain anything from mobile to desktop. I dont believe mobile A6-5200 throttles in single thread.

i don't think it throttles with all four cores either. i pulled up Performance Monitor on a W8 Laptop with an A6-5200 in it and ran CB15 and it ran at 1.96-1.98 Ghz the entire time.

 

[Dresdenboy]

Basically the same and basically the same. 0.03 points is basically the variation between runs. Jaguar is a pretty nice design.

And it drives the PS4 and XBox1.

What we actually see here are 2 SMT FPUs vs. 4 FPUs.

 

[Red Hawk]

Yeah the 7850K's price right now is kind of ridiculous. It's the price I paid for my 2500K over two years ago. I could buy a much cheaper CPU from either AMD or Intel, throw in a 250X, and have equal or better gaming performance on the CPU side and much better performance on the GPU side. The 7850K's price needs to come down in order to be appealing for simple budget gaming.