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

[BallaTheFeared]

Considering they all reliable have built-in benchmarks? Pretty high, actually.

There are plenty of recent non gaming evolved titles that feature built-in benchmarks though.

Metro, Rome, Grid all were used in a previous Anandtech R290 review yet absent here.

 

[monstercameron]

What 65w CPU?

what 55, 74, 84W intel part?

 

[ShintaiDK]

Interesting. 45W Kaveri consumes far more than 45W Trinity. It seems that they are messing with their TDP specs again.

It does seem so. Neither the 65 or 45W model seems to be anywhere that TDP. I guess its like the 140W+ TDP style FX series. The 95W one tho seems to be in the ballpark. So much for the "efficiency" PR spin.

 

[monstercameron]

for Hardwarecanucks did they test all the platforms with the same ram speed or am I misreading it?
http://www.hardwarecanucks.com/foru...31-amd-kaveri-a10-7850k-a8-7600-review-8.html

 

[ViRGE]

for Hardwarecanucks did they test all the platforms with the same ram speed or am I misreading it?
http://www.hardwarecanucks.com/foru...31-amd-kaveri-a10-7850k-a8-7600-review-8.html

They're testing it with the same RAM speed. Every system gets TridentX RAM running at 1600MHz.

 

[monstercameron]

They're testing it with the same RAM speed. Every system gets TridentX RAM running at 1600MHz.

why would they test an apu that way when it performs better with faster supported ram? those igp results feel invalid now...especially for the flagship.

 

[ShintaiDK]

This is what Damage ska Scott Wasson from Techreport writes:

Yeah, AMD's plan to protect Kaveri involved only sending out the A8-7600 and competitive parts. They obviously wanted to protect their new chip from the scrutiny it would get for being no faster in CPU benchmarks at 95W. Even the A10-6700T wasn't part of the original plan, just the slow-poke 6500T.

The fail is complete.

 

[sontin]

After reading other reviews - it's a turd.
I dont see any good point about Kaveri:
CPU performance is abnormal.
iGPU performance is not fast enough for a $180 product.

And every other features only exist on paper and are either disabled or not here right now.

 

[avtek21]

TR interesting playability and frame times of 45w/65w Kaveri versus i7-4770k

Confirms previous info about Musemage image manipulation acceleration with OpenCL.

 

[Enigmoid]

Disappointing. CPU gains are nowhere to be found.

Is it just me or has the quality of AT reviews gone down?

No power consumption or overclocked performance on the kaveri review. (Haswell review was lacking as well). Open the review on pages and pages of how HSA and hUMA are going to change everything; no notice that there was in fact no real change outside of compute tasks.

And the igp test settings this site uses are pretty bad. We go from 68 to 25 fps on bioshock for example, skipping the optimum playable range. 1080p xtreme is completely useless. I notice this a lot. Pick something for optimum playability not for the sake of the settings.

 

[Enigmoid]

why would they test an apu that way when it performs better with faster supported ram? those igp results feel invalid now...especially for the flagship.

Costs more. Fairness of the platform. You don't expect them to run tests with a HDD on one platform and a SSD on the other.

 

[monstercameron]

pcper review up
http://www.pcper.com/reviews/Processors/AMD-A8-7600-Kaveri-APU-Review-HSA-Arrives

 

[monstercameron]

Costs more. Fairness of the platform. You don't expect them to run tests with a HDD on one platform and a SSD on the other.

what costs more? going from 1600-2400? the premium is negligible if it exists. This isn't about fairness but stock for stock, if the APU is supports 2133-2400 give it that in the benchmarks.

 

[Arkaign]

So someone wants a small form factor slim unit that will be great at playing movies, great at gaming, isn't loud and hot. They want a GOOD APU. They only have about $400 to spend. Well gents, I give you :

LOL

 

[NTMBK]

If Mantle, Trueaudio, etc can take off and Kaveri can piggyback on the GCN GPUs' success, it might be worth recommending in 6 months. Without that it's a bit screwed.

The problem I see is that it really only outshines Richland in high graphics settings- when all of the compute shaders are turned on, GCN really gives VLIW4 a kicking.

But the problem is that while it can go 50% faster, 6FPS is just as unplayable as 4FPS. The low GPU clock and lack of bandwidth have killed it, in my opinion.

And where the hell is the GPU turbo?! They just brought out fantastic new Turbo technology in Hawaii to make the most of TDP- even more important in an APU, where they need to balance power consumption between CPU and GPU- and it's completely MIA in Kaveri, leaving it with only 720MHz. Disaster. What's the point of Mantle dropping CPU usage, if your GPU can't turbo up an extra 50MHz to use the spare power budget.

 

[ShintaiDK]

And where the hell is the GPU turbo?! They just brought out fantastic new Turbo technology in Hawaii to make the most of TDP- even more important in an APU, where they need to balance power consumption between CPU and GPU- and it's completely MIA in Kaveri, leaving it with only 720MHz. Disaster. What's the point of Mantle dropping CPU usage, if your GPU can't turbo up an extra 50MHz to use the spare power budget.

I wouldnt say Hawaii uses turbo. It seems to be a matter of throttle only.

 

[NTMBK]

I wouldnt say Hawaii uses turbo. It seems to be a matter of throttle only.

It dynamically adjusts clock speeds between 1GHz and ~700MHz dependent on power consumption and temperature- I guess you could call it throttling, but whatever, semantics. Either way, that's what I expected in Kaveri, but nope. 720MHz.

 

[monstercameron]

apparently you can modify the tdp via the bios...or so according to phoronix

Kaveri also introduces a new feature of a configurable TDP. The A10-7850K has a 95 Watt TDP by default, but from the BIOS/UEFI it's possible to drop the APU down to running at a 45 Watt TDP but at the cost of reduced performance. The configurable TDP feature works on any Kaveri APU but AMD only has optimized it for the low-end A8-7600.

 

[ShintaiDK]

It dynamically adjusts clock speeds between 1GHz and ~700MHz dependent on power consumption and temperature- I guess you could call it throttling, but whatever, semantics. Either way, that's what I expected in Kaveri, but nope. 720MHz.

720Mhz is the boost clock for the IGP. The baseclock is 654Mhz.

 

[NTMBK]

I know you're taking the piss, but if I could flash that PS4 and run Windows & Steam on it then I would genuinely already own one.

 

[SlowSpyder]

The performance isn't bad. But I think the top Kaveri part should be ~$150, and lower models priced from there down. It isn't a clean sweep with the lower clock speeds, so the price increase is a bit too steep compared to the top Richland part that can be found for ~$140.

 

[monstercameron]

I know you're taking the piss, but if I could flash that PS4 and run Windows & Steam on it then I would genuinely already own one.

I think half the internet would have already...

 

[Enigmoid]

AMD's management team should be fired.

Send out 45W - 65W SKU models. Review sites compare them to a10-6800k.

But like guru 3D, Application performance is utter crap.

I've been thinking about this for days really, I mean AMD put well over 2 Billion transistors in this APU and that shows in both OpenCL and graphics performance. Kudo's for that. But the actual processing performance of the product is barely a notch faster compared to Trinity and Richland, and I just don't get it anymore. See it's now 2014 and the APU serial processing performance is not even at the performance level of a Phenom II X4 970 from early 2011.

 

[monstercameron]

http://www.guru3d.com/articles_pages/amd_a8_7600_apu_review,1.html
http://www.bit-tech.net/hardware/cpus/2014/01/14/amd-a8-7600-kaveri-review/1
http://www.hardwareheaven.com/revie...ring-gigabyte-g1sniper-a88x-introduction.html

 

[JDG1980]

Kaveri's GPU seems to be bottlenecked by DDR3, at least on many of the gaming benchmarks. Looks like any meaningful performance jump over Richland levels requires EDRAM, DDR4, or GDDR5.