bit-tech.net

AMD A10-7850K and A10-7700K (Kaveri) Reviews

Comments 1 to 25 of 50

Reply
Corky42 30th January 2014, 12:36 Quote
With rumors saying Dice is rolling out their long awaited BF4 Mantle patch in the next few hours/days and with AMD supposedly releasing 14.1 Beta drivers supporting Mantel, would there be a chance of revisiting the BF4 benchmarks ?
Bede 30th January 2014, 12:43 Quote
It'll be fascinating to see if Mantle can do anything for AMD, my feeling is it can't but I'm willing to be surprised.
Dogbert666 30th January 2014, 12:48 Quote
Yeah, if Mantle does launch today I'll probably give it a few days plus the weekend for issues to be found and removed, then get to testing on Monday.
Harlequin 30th January 2014, 12:50 Quote
Sheiken 30th January 2014, 13:17 Quote
YYEEEESSS! Perfect, can't wait to try this out when I get home and slot in my new hyperx beast 2400mhz in the kaveri rig!
Harlequin 30th January 2014, 13:19 Quote
oh and yes kevari shows scaling beyond 2133 ram - 2400 is faster as well ;)
rollo 30th January 2014, 13:24 Quote
looks like they tested for you.

14% faster on a 7850k with mantle.

the headline 58% is in single player with multiple gpus. ( anyone actually play bf4 single player )

Would like to see the multiplayer tested again on a none AMD cpu as its been held back by that CPU acording to Dice.

Will run some testing of my own later on, Enough it seems the 7000 series is unsupported on launch so I guess my 7970 will not work for testing purposes hmm.
Dogbert666 30th January 2014, 13:26 Quote
Yeah, we only used 2,133 since that what they're officially rated for, but certainly the faster memory the better. Advice is to get the fastest kit possible without breaking the bank.
Corky42 30th January 2014, 13:35 Quote
AMD needs to pull its finger out and publish the 14.1 Beta drivers to go with it, until then we still only have AMD's figures to go on, and the numbers they have published look a bit odd.
Yes we see a claimed %58 in a crossfire rig, but as another site claims it seems Mantel "shows itself most dramatically when you are in a heavily CPU limited environment."
sandys 30th January 2014, 13:35 Quote
Quote:
Originally Posted by rollo
looks like they tested for you.
the headline 58% is in single player with multiple gpus. ( anyone actually play bf4 single player )

Me, first game I completed this year :o

Only problem I found with the APU is that you have to do FSB overclocking for higher memory speeds, will be less of a problem with Kaveri as it has a 2133 but on my old A10 it was tricky to do more than 2100 as FSB went too high, have they put more multis than 2133 support in Kaveri or does it still need FSB overclock.
Harlequin 30th January 2014, 13:41 Quote
ocuk have some 2400 kits for £60 (8gb)
Parge 30th January 2014, 14:35 Quote
Quote:
Originally Posted by Corky42
AMD needs to pull its finger out and publish the 14.1 Beta drivers to go with it, until then we still only have AMD's figures to go on, and the numbers they have published look a bit odd.
Yes we see a claimed %58 in a crossfire rig, but as another site claims it seems Mantel "shows itself most dramatically when you are in a heavily CPU limited environment."

A crossfire setup will likely be CPU limited.
Corky42 30th January 2014, 14:42 Quote
Almost definitely. The thing that struck me as odd was that i got the impression Mantel was intended to improve the GPU side of things, but as it turns out it seems to be a greater benefit in CPU bound situations.

Cant wait to see how much benefit this brings to Kaveri, as unless I'm having a senior moment from what i can tell the CPU part of Kaveri is what is letting it down in terms of FPS.
Sheiken 30th January 2014, 16:51 Quote
Quote:
Originally Posted by Corky42
Almost definitely. The thing that struck me as odd was that i got the impression Mantel was intended to improve the GPU side of things, but as it turns out it seems to be a greater benefit in CPU bound situations.

Cant wait to see how much benefit this brings to Kaveri, as unless I'm having a senior moment from what i can tell the CPU part of Kaveri is what is letting it down in terms of FPS.

Well you actually say it yourself. It improves on the GPU part of things, to such an extent that the cpu becomes less of a factor and/or bottleneck.
rollo 30th January 2014, 17:01 Quote
Still is the bottleneck acording to dice though.
azazel1024 30th January 2014, 17:22 Quote
Quote:
Originally Posted by Dogbert666
Yeah, we only used 2,133 since that what they're officially rated for, but certainly the faster memory the better. Advice is to get the fastest kit possible without breaking the bank.

And I think that is part of the problem. The price isn't THAT much more to go from DDR3-1600 2x4GB to DDR3-2400 2x4GB, but it is another $15+. Considering the much lower price of an Intel core i3 Haswell processor, combine that with using less expensive memory, and you are a good part of the way to a low end discrete GPU...which will provide better performance in most CPU AND GPU tasks then, at barely more money.

I see the appeal to Kaveri and it is a good step forward on the GPU front, but on the CPU front it doesn't do much. It seems to maybe hold the line, but AMD was already so far behind the line that doesn't mean much. If you want CPU performance, Intel, if you want an inexpensive machine and don't really care about GPU performance, pretty much Intel again. If you really care about gaming, probably Intel again for the CPU and discrete GPU(s). If you need a real budget gaming rig, or an HTPC that can also double as a gaming rig and are very size limited, THEN Kaveri/AMD makes sense.

Thats a relative edge case. Even basic office tasks, the performance difference (especially because a lot of the stuff is single threaded/light multithreaded) between a low end Intel Pentium Haswell processor even and AMD means you'll have a better entry level computer experience with the Intel Pentium, with a fine GPU that'll handle your Youtubes and Facebooks and flash games and browser GPU acceleration, but generally have better single thread performance and light multithreading and probably also cost around $50-60, cheaper than a bottom of the barrel Kaveri.
Harlequin 30th January 2014, 17:31 Quote
its a lot closer than you want to think

the top of the line 7850k is $184 on newegg

the i3 4330 is $135

so the $50 difference gets you a WORSE gfx card - to get one comparable its $75 (R7 240)

and I would suggest you check ram price - teamgroup CL10 2400 is $2 more than 1600 CL11
Dogbert666 30th January 2014, 17:47 Quote
azazel1024, I agree to an extent, and I went to some length to outline in the conclusion that Kaveri certainly isn't a one stop solution for all user needs. Harlequin is also right; I didn't check 2,400MHz prices but in the UK I was able to find 2,133MHz kits that were just £1 or £2 dearer than £1,600MHz ones. DDR4 is coming and manufacturers need to shift the higher speed DDR3 kits.

It depends what you mean by basic office tasks, but while I won't argue that single threaded performance favours Intel's processors, I would say that the entry level experience doesn't rely on CPU-intensive tasks.
sandys 30th January 2014, 17:56 Quote
overclockers has 8Gb 2400 ram for 56quid

In office scenarios I am hard pressed to tell the difference between my 4.9Ghz 3770 and my daughters underclocked a10.

She can game nicely on her a10 chip, I was impressed anyway, it isn't even as good as Kaveri being one of the old 5700s.
Corky42 30th January 2014, 18:14 Quote
Quote:
Originally Posted by Sheiken
Well you actually say it yourself. It improves on the GPU part of things, to such an extent that the cpu becomes less of a factor and/or bottleneck.

No. So far all figures show it does little to improve the GPU side of things, that isn't saying it doesn't improve the GPU side at all. But figures show Mantel has more impact on CPU bound games than GPU bound ones.
Quote:
In Battlefield 4, we have more robust information. AMD claims that the FX-8350 + R9 290X will see an average of 23.8% better performance under Mantle, while the Core i5-4670K + 290X will see a performance improvement of just 7.5%.

So we see a bigger performance gain using arguably the poorer of the two CPUs, and reading up on why it makes sense as a big part of Mantel is reducing draw calls, so making a CPU bottleneck less of an issue.

A game that proves Mantel helps more when the CPU is an issue, is when you look at the improvement when AMD used Mantel and the Star Swarm demo they saw a 319% improvement at 1080p

AMD even say
Quote:
AMD stresses that Mantle works best when a game is CPU-bound, not GPU-bound – if a game is GPU-bound, it tends to stress the card to such a degree that Mantle-like optimizations will not help much at all. That said, those who will see the best benefit of Mantle will have modest PCs
Harlequin 30th January 2014, 18:30 Quote
Quote:
Originally Posted by Dogbert666
azazel1024, I agree to an extent, and I went to some length to outline in the conclusion that Kaveri certainly isn't a one stop solution for all user needs. Harlequin is also right; I didn't check 2,400MHz prices but in the UK I was able to find 2,133MHz kits that were just £1 or £2 dearer than £1,600MHz ones. DDR4 is coming and manufacturers need to shift the higher speed DDR3 kits.

It depends what you mean by basic office tasks, but while I won't argue that single threaded performance favours Intel's processors, I would say that the entry level experience doesn't rely on CPU-intensive tasks.

as said in another comment - OCUK have CL 10 2400 8GB kits for £56 - im using one with an A10-5800k and its awesome. will post comparable benchies if you want?
Corky42 30th January 2014, 18:42 Quote
Speaking of what RAM to use with Kaveri, some benchmarks show you also get an improvement when using dual-rank DIMM memory.
Quote:
The best bandwidth is apparently achieved when Kaveri APU has both memory channels populated by 2-rank memory modules or with four single-rank modules, depending on what you can get your hands on.
supermuchurios 30th January 2014, 21:47 Quote
Where's Alienware (AMD) Andy when you need him
azazel1024 30th January 2014, 23:49 Quote
Yes, the price difference isn't a whole heck of a lot, but it still inches you closer to a GPU with better performance, and better CPU performance to boot going with the Intel.

Also looking at performance, I specifically mentioned a Pentium processor, which is a lot more than $50 cheaper. Looking at the Kaveri numbers, it at least seems to me that a decent Ivy Bridge Pentium processor is going to be faster in a lot of the single thread and light multithreaded CPU tasks even...and that is more like $100+ cheaper, which deffinitely buys you a nicer GPU.

Heck, just looking at the numbers, the 2.6Ghz G1320 in my sever is likely faster in a resonable number of those single threaded benchmarks, and that is a $40 processor.

Heavy multithreaded stuff might in that case favor the Kaveri processor though.

Overall my point is just that Kaveri seems to mostly just fill the niche of a, in general, a better performing system where a discete GPU is generally not an option. There ios an edge case where it might be a better performing system for some GPU and CPU tasks for an equivelent cost system incorporating a low end current/last generation Intel CPU and a discrete GPU...but then you are very much in to "what performance do you care about the system being best at?"

Even you jump in to the very bottom of the barrel on computer requirements, Intel can deliver a system that is possibly cheaper. A super low end Ivy Bridge board and the lowest end Celeron price comparing seems to be around $10 cheaper than the lowest end AMD CPU and board that I can find (I am excluding the Sempron, which doesn't have an integrated GPU, but even then, it only pushes the price to being a couple of dollars more than Intel).

So, on price, AMD is only sometimes, occasionally the low price winner and it is ONLY the low price winner in a relatively narrow set of requirements.

I am really, really hoping for a better Arch once the orginal Bulldozer Arch timeline has run its course and AMD is moving on to the next Arch. I am not holding my breath as it seems like Intel's process and node advantage is part of what is delivering it significantly better performance at a lot power power levels.
Harlequin 30th January 2014, 23:54 Quote
you sat `faster in a lot of apps`


name 1 app , outside of a benchmark , in the real world you will actually notice or care what is under the hood - especially at this end of the market?
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.



Discuss in the forums