Originally Posted by wizard69
Nobody completely obsessed with benchmarks will be happy with Kaveri. If you take a more balanced view the processor really isn't that bad.
LOL. First you take exception to "reading biased reporting" and now with being "obsessed with benchmarks". So I guess your opinion is based merely on personal whimsy.
I'm no more "obsessed with benchmarks" than anyone else. They are an objective measure of the performance of a system component that provides insight into what to expect for real world performance increase. Yes, you can lie with benchmarks but Anand, despite your accusations of anti-AMD bias, is pretty decent to presenting representative enough benchmarks.
Your "balanced view" isn't balanced at all. AMD promised Core i5 level of CPU performance. Everyone was excited that they caught up to Intel but the benchmarks shows that they failed.
Which is baloney. Sure Intels best CPU does better in some tests but lets face it Haswell does much better than Sandy Bridge. As for your obsession with the GPU as a gaming feature that is baloney also. GPU's are very important in all modern operating systems.
Nope. CPU still dominates standard desktop workflows. Even for transcoding after Intel introduced quick sync. Sufficiently so that Apple has for relied on Intel's IGP since Sandy Bridge for it's entry level laptops (MBA, 13" MBP) and now for all of it's entry level machines including the 15" MBP.
Are you claiming that OSX is not a modern operating system or that Apple engineers are clueless?
And the gaming focus is exactly AMD's focus. Those are the benchmarks that Kaveri wins. That's the emphasis that AMD puts in its marketing literature.
It beats Kaveri if you hand pick the benchmarks to prove your point. Kaveri however delivers a far more balanced result.
Show any CPU benchmark where Kaveri beats the Core i5. CPU benchmarks and not the "CPU while gaming" benchmarks that AMD touts.
You don't like Anand. Try Tom's. Same result. Look at the content creation, productivity, compression, and media encoding benchmarks.
Even for OpenCL benchmarks.
"But the Intel chip does quite a bit better in our OpenCL-accelerated metric. In fact, we also see the Richland-based A10-6800K finish up ahead of any Kaveri-based APU."
"The same thing happens in Sony’s Vegas Pro 12 as Intel’s CPUs take first and second place, while A10-6800K slides past the Kaveri-based APUs. This is an OpenCL-accelerated workload, so it’s strange that AMD’s latest doesn’t turn in stronger numbers. But when we turn off OpenCL and run the same benchmark, completion time nearly doubles. So hardware acceleration is definitely helping, just not as much as we would have thought given AMD's more modern graphics architecture."
These are what people do with their minis. Decompress files, encode video, use iTunes, use iLife. Nobody can run benchmarks using Kaveri and OSX but Vegas, WinZip, Adobe, Handbrake and iTunes on windows WITH OpenCL turned on in many cases is indicative that for mini users the Core i5 is the far better chip than any Kaveri option except for gaming.
Well you are living in the past, the CPU stopped being important at least 3 years ago. Why do you think Apple pushed intel so hard to implement decent GPU's?
Kaveri consistently sucks ass in comparison to the Core i5 when doing things that most people care about.
Intel has been working on improving their GPU performance since the 1999 introduction of the Intel740 as part of their strategy to own the entire lower end of the graphics market. Until Intel's move to lock out 3rd party chipsets it hasn't mattered to Apple very much. Prior to 2010 Apple happily used 3rd party IGPs like the nVidia 9400M as part of their solution when Intel's IGP was considered too slow. Given that Apple choose the GMA900 and other Intel IGPs from time to time over nVidia or ATI IGP solutions Apple obviously felt that Intel's offerings during those periods were competitive or they wouldn't have used them.
Whether Apple pushed Intel on the IGP front is debatable. Certainly Apple's laggard implementation of OpenCL on Intel's platform indicated to great rush to push GPU acceleration in OSX.
Reading all the benchmarks wouldn't hurt. Kaveri doesn't do bad at all when benchmarks making use of the GPU are taken into account.
The only benchmarks I don't care about are gaming and synthetic computer. Those are the ones that Kaveri wins. While I care about gaming I'm not going to game on a IGP so those benchmarks are moot. I don't care about synthetic compute benchmarks because these generally aren't indicative of production software performance. I'd rather see Adobe or Handbrake CPU and OpenCL benchmarks.
That is fine for you. However if you are looking for a low cost node to put on the net somewhere Kaveri looks to be very interesting and is certainly better than any i3 based machine and gives the i5 based machines a run for their money.
Except nowhere do they give "i5 based machines a run for their money". At $600 the Mini is not a "low cost node".
Exactly so why do you do it????
I provide links to benchmarks to support my case. You provide handwaving.
Seriously dude my post was about the Mini a machine Apple doesn't exactly imbue with a great deal of performance to begin with.
The $800 mini has a huge amount of performance. Period. Any step back from the Core i7 is a huge reduction in performance.
It would be a huge difference if this was a discussion about the iMac, but in the Mini Kaveri would be fine. This is especially the case if Apple can trim a few bucks off the price of the machine.
A mini Kaveri would be fine if you didn't care about iPhoto, iMovie, zipping files, handbrake, iTunes, iWork, Xcode, etc performance getting significantly slower.
The workflow doesn't have to be dominated by GPU performance for a GPU to be important with respect to OS performance. Just about every app gets some benefit from the GPU if it used the Mac GUI at all.
The benefit is completely negated by the significant loss in performance in content creation, zipping, encoding, compiling, etc. The HD4600 is more than capable of providing all the GPU acceleration required for normal OS performance.
Hey I don't like Apple dragging its feet either. I honestly believe that the pace of Mac OS development has slowed in preference of iOS.
Given that they invested their limited resources in other aspects of OSX then GPU acceleration is obviously NOT the cornerstone of Apple's OS strategy as you repeatedly assert.
You haven't kept up with Intel failure or you have selective memory. in any event nothing I've seen indicates that Kaveri is getting its but kicked by i3 processors.
Since Intel's failures are so rampant why don't you list them since it should be easy. Oh wait, you don't mean the schedule slips? Yah, like AMD has any advantage there. Intel's broadwell desktop slip is relatively minor given that they're first to 14nm which is pretty hard.
That you close your eyes to Kaveri's shortcomings is immaterial. The benchmarks are all linked.
Wrong again. Apple went ARM just before the iPad release when it became obvious that Intel didn't get it. It is pretty hard for Apple to be dissatisfied with Samsung when Samsung is producing Apples designs. Even before the Apple designed hardware hit, Samsung's designs where ARM derived with help from Intrinsity. More so Apple Purchased P.A. Semi in 2008 specifically to go their own way in processor development, with Jobs specifically saying they where going to develop their own processors.
So yeah I'd call it a crash program. They literally went out and purchased the expertise to build their own chips and pushed out those design rather quickly. That move was motivated by Intel just not getting low power. Frankly it was a good move on Apples part as Intel did not have anything close to reasonable for a tablet until late last year. Now Apple has a chip that pretty much holds its own against Intel chips targeting the same market.
Too bad that Apple stated differently.
"When we look at future roadmaps, mid-2006 and beyond, we see PowerPC gives us 15 units of performance per watt, but Intel’s roadmap gives us 70. And so this tells us what we have to do," - Steve Jobs WWDC 2005
And they choose Intel over AMD precisely for that reason.
That Intel was not interested in becoming a ARM fab in 2006-2007 let Samsung and all it's resulting problems into the picture.
Intel's interest in phones and tablets has always been tempered by the low cost of those chips just like Apple not wanting to get involved in the low end of the computer business. Intel's drive to far lower power has far more to do with keeping ARM solutions out of the server and ultra book high ASP and high margin market. Something that I believe that they have succeeded in doing. There is no reason to go ARM for ultra books or servers any more.
Intel's margins are 60% and the ASPs are far higher than for ARM.
Job's interest in developing their own processors for the iPad and iPhone was a desire for a competitive advantage that Android (or Samsung) could not easily match.
And obviously that desire was largely fueled by Samsing copy-catting the iPhone.
In any event this ARM discussion pulls thread off course. HSA architecture has so much to offer Apple that I see Intel as having no choice but to match the capability or loose Apples interest. You can disagree with that but it seems lie Intel is following AMD in this regard.
By the time HSA matters at all Intel will have reached parity with AMD or surpassed it in terms of IGP performance. HSA software won't be out until later this year (what, summer?). That means no production software will use it before 2015 at the earliest. That means it won't matter for real before 2016.
Given the rapid GPU advances in the last 3 Intel CPUs I don't think that AMD's GPU advantage will last until 2016.
Edited by nht - 1/21/14 at 7:07am