kaveri bf4 good performance?

[monstercameron]

just found this video of the amd guys messing around with BF4, they tried a few configurations and this apu seems quite capable.

http://www.reddit.com/r/APUsilicon/...kaveri_bf4_gameplay_tldr_720p_ultra_playable/

close to the end they cranked it up to ultra at 720p and they were able to almost maintain 30fps.

note: the video was of the indoor scene without any ai

for us to stay on topic here is what I would prefer we discuss, resolution, framerates, apis, comparison[other os, competition's hardware], value of demonstrating the indoor scene, technical issues such as bandwidth.

please no cmt vs smt, this/that sucks, financials, i know better than x engineer, faildozer, predictions etc.

 

[raghu78]

Kaveri is turning out to be a good product even though long delayed due to manufacturing troubles at Globalfoundries. AMD's software innovation is what will make Kaveri shine. HSA with unified memory architecture, Mantle and True Audio being the key factors.

AMD has tied up with game developers and will bring support for Mantle in some of the most high profile games - BF4, Star Citizen, Thief. Further more all of EA games in development (there are 15 of them) use Frostbite 3 and support Mantle. AMD also is working with game developers to bring True Audio support.

I think AMD has quickly realized that none of the foundries can compete with Intel's manufacturing juggernaut. Since AMD depends on the foundries they are fighting a losing battle if they try to compete on pure silicon race. They have smartly focussed on software innovation to outgun Intel and will succeed finally.

I can't wait to see BF4 with Mantle running on an overclocked Kaveri. That will blow the Intel solutions out of the water.

 

[LogOver]

Mobile Iris Pro at 1366 x 768 (high): 28.5 average, 24 min

http://www.notebookcheck.net/Battlefield-4-Benchmarked.105583.0.html

 

[Silverforce11]

Mobile Iris Pro at 1366 x 768 (high): 28.5 average, 24 min

http://www.notebookcheck.net/Battlefield-4-Benchmarked.105583.0.html

There's a big difference jumping from high to ultra.

Iris Pro is also a power hog once you load it. Not impressed for the price nor power use.

 

[leper84]

So Kaveri performs as good or better than an Intel chip that costs 4X as much?

 

[LogOver]

There's a big difference jumping from high to ultra.

Iris Pro is also a power hog once you load it. Not impressed for the price nor power use.

Not a big jump, according to the video. Kaveri has a problem to maintain 30 fps@720p with high settings in scene with no open spaces and ai.

 

[Arkadrel]

Looks like its does:

1600x900 @medium settings and around 39-54 fps (usually 41+).
1280x720 @medium settings and around 49-65 fps (usually 52+).

I wouldnt use it at high or ultra, those where usually around 25-30+ but,
you really dont want to dip below 30 when playing games.

Overall not bad for a non optimised driver with kaveri.

 

[LogOver]

So Kaveri performs as good or better than an Intel chip that costs 4X as much?

Do we know prices and performance of mobile Kaveri chips?

 

[Arkadrel]

I was kinda hopeing for better than this, hopefully its a driver issue, or early test bed not reflective of final products results.

1280x720 @ultra (without AA) at around 30 fps is okay but its not fantastic.

So Kaveri performs as good or better than an Intel chip that costs 4X as much?

Mobile Iris Pro at 1366 x 768 (high): 28.5 average, 24 min
http://www.notebookcheck.net/Battlef....105583.0.html

@Leper84

It looks like it:

Iris pro 5200 does 28.5 avg fps @1366 x 768 medium settings.
Kavari A10 7850K does ~49-65 fps @1280x720 medium settings (un-optimised driver in video).

 

[ShintaiDK]

Dont compare desktop and mobile GPU. AMDs mobile GPU performance is far from desktop.

Also dont compare 2 different benchmarks.

Specially not when one is a youtube one. We dont exactly lack examples of cheating on that media.

 

[LogOver]

Iris pro 5200 does 28.5 avg fps @1366 x 768 medium settings.

Go and check your information again.

 

[ShintaiDK]

Iris pro 5200 does 28.5 avg fps @1366 x 768 medium settings.
Kavari A10 7850K does ~49-65 fps @1280x720 medium settings (un-optimised driver in video).

Iris Pro laptop did 45.5FPS in that setting.

 

[Arkadrel]

Ohhhh your right <.<

That was the high quality mode, jeez at medium it does 45.5 fps avg.

Kavari isnt looking that special now lol.

 

[ShintaiDK]

The elephant in the room is obviously the Radeon HD 8970M, 4700MQ. 67FPS in that setting.

So unless someone claim that Kaveri=HD8970M performance+bandwidth. Then its not gonna happen.

 

[schmuckley]

I wonder how it does with discrete GPU?

 

[Arkadrel]

Iris pro and kavari are actually looking pretty strong compaired to some of the mobile discrete GPUs.

GeForce GT 750M = ~45.4 fps (Iris pro beats this, so does the kavari in video)
GeForce GTX 660M = ~53.4 fps (Iris pro doesnt beat it, Kavari in video does)

Radeon HD 7970M = ~61.9 fps (beats both, but if kavari is ~57 fps not by huge amounts)

but again desktop apu vs mobile parts.

Lets wait and see how kavari does in mobile format (inside a laptop).

 

[Vesku]

A bit silly to be "meh" about Kaveri performance just because a chip with 128MB of eDRAM is already pretty close. I don't think those two products will be anywhere near each other in terms of pricing.

It's interesting that they were showing off BF4 in a hallway with no opponents. Makes me think the CPU side of AMD APUs will still have trouble fully utilizing the iGP when memory bandwidth is not the limiting constraint.

 

[Enigmoid]

Iris pro and kavari are actually looking pretty strong compaired to some of the mobile discrete GPUs.

GeForce GT 750M = ~45.4 fps (Iris pro beats this, so does the kavari in video)
GeForce GTX 660M = ~53.4 fps (Iris pro doesnt beat it, Kavari in video does)

Radeon HD 7970M = ~61.9 fps (beats both, but if kavari is ~57 fps not by huge amounts)

but again desktop apu vs mobile parts.

Lets wait and see how kavari does in mobile format (inside a laptop).

Are you using notebookcheck?

It looks like the 7970m/8970m is having enduro issues as even the 670mx beats the 8970m at 768p med with 68.8 fps. The R7 250 gets around 67 fps on 768p medium.

If anything kaveri looks slightly weaker than the GDDR5 7730 (Mars).

Interestingly HD 4600 (mobile) pulls around 25 fps on that setting. Once you bump the clocks to desktop levels (+10%), it looks like kaveri is around twice as fast. As for mobile the HD 4600 is a good bit faster than the a10-4600m.

As for the notebookcheck benchmarks they do not appear to be consistent. 660 TI does better than the 680 on everything except 1080p ultra. Enduro issues. Also, the R7 250 does really well, normally its only slightly ahead of the 660m.

 

[Paul98]

I wouldn't call those frame rates playable. That is also single player so who cares, lets see 40+ players multiplayer on low to see how it really works. Though with mantle it could end up working just fine.

 

[sniffin]

64p BF4 on mobile Richland usually holds between 30-40fps when coupled with a discrete GPU and low mesh settings, so I wouldn't be to worried about CPU performance. More interesting is how the iGPU handles, and AFAIK GPU load is not greater in MP than it is in SP.

I like the 1 star rating for the thread, Shintai clearly does not approve. Hardly surprising!

 

[tweakboy]

Ive heard of the cloth designer, Kavali ,,,,

 

[Silverforce11]

and AFAIK GPU load is not greater in MP than it is in SP.

It is greater. 64MP is completely on another level compared to SP.

BF4 still looks great on medium, so as long as these APUs can handle 1080p medium gaming, it's a great product for a cheap build or HTPC + light gaming.

 

[Zor Prime]

BF4 on Kaveri is gonna look better in December when it becomes Mantle powered.

 

[Soulkeeper]

Outside of compute intensive apps, I really don't see the point in adding more SPs to these APUs untill they get ddr4.

 

[tweakboy]

Nice rig Soulkeeper. gl