Ars: AMD may be irrelevant

[pelov]

I can't understand why you're going down this road when all you're doing is projecting your personal opinion.
To actually answer your question, I'd choose the Intel chip with Integrated graphics for one sole reason. So that if I had to, I could use it. For whatever reason. In between discrete graphics card purchases, failed discrete GPU, RMA'ing.
It's my opinion, as you are displaying yours, but don't go on about it like your opinion is the only way and the "right" way. It might be for you, but that's as far as it goes besides a few who may share your outlook. K?

So to clarify, an "enthusiast" is one who chooses to waste 30-40% of the die (soon to be 50%) sitting there doing absolutely nothing for years but there just in case the graphics card craps out, at the expense of having a better performing CPU.

You have your opinions and I have mine. The difference here is you aren't able to convey yours very well :whiste: Apparently, failed GPUs are one of the most dreadfully common things we enthusiasts experience and we would definitely choose to have an underwhelming processor if it means we worry less about dead GPUs and RMAs.

I can't wait to hear your opinions of Haswell. Do we completely ignore the eDRAM? The 50% of the die? What about Broadwell? Are those all advancements driven by enthusiast needs? Oh boy...

 

[Idontcare]

To the people who make business decisions regarding market segments, "enthusiast" is placed above "performance", which is placed above "mainstream", etc.

You can argue with them what their working definition of enthusiast means since they are the ones who defined/created the market segment to begin with.

For people who self-identify as "enthusiasts", no one is going to tell someone with an i3 or a trinity that they are or are not an enthusiast. But the product slides from marketing will tell you whether or not those products are targeted at enthusiasts.

Ford has its idea of what makes a muscle car, and yet people take Nissans and add glass-pack mufflers them with spoilers and call them racers.

 

[pelov]

To the people who make business decisions regarding market segments, "enthusiast" is placed above "performance", which is placed above "mainstream", etc.

You can argue with them what their working definition of enthusiast means since they are the ones who defined/created the market segment to begin with.

For people who self-identify as "enthusiasts", no one is going to tell someone with an i3 or a trinity that they are or are not an enthusiast. But the product slides from marketing will tell you whether or not those products are targeted at enthusiasts.

Ford has its idea of what makes a muscle car, and yet people take Nissans and add glass-pack mufflers them with spoilers and call them racers.

While the term is relative, it's the same people who are whining about Bulldozer and Trinity's performance that are quite content with wasting 30-40% of the Sandy/Ivy die to useless graphics when an overwhelming majority of them use a discrete GPU. I was doing it to point out the hypocrisy in this line of thought and the thread.

I do think it brings up another point that's rarely talked about: Do we need the extra performance at all? If people are okay with having massive on-die graphics that they're not going to use and a mere 10% performance bump from generation-to-generation, do they really need the extra processing power? Just like the average person's computing needs are being fulfilled by tablets and smartphones, so too has the enthusiasts processing needs reached a point where they're willing to ignore the wasted space on a processor and the relatively meager performance increases. AMD and Intel certainly seem to think that's fine, but why don't some of the people here?

AMD isn't screwed because they don't have enough CPU performance on the desktop. That hardly matters anymore when people are raving about a 10-15% increase in performance while they're ignoring the two-fold increase in die space and graphical performance. No. AMD is screwed if they haven't had SoC improvements and ULV chips in the sub-10W range already brewing. And judging by the FM2 socket life expectancy and the 2 generations of APUs fitting into the same socket and chipset, I'd be very worried for AMD.

 

[sontin]

A perfect summary of A10 was done by SA:

Wow - trinity is fast enough. So why do we need HSA?

 

[Phynaz]

Sounds good. I am sure you are not against corel video studio pro X5 or arcsoft media converter for the AVDHC conversion, as they feature OpenCL and that Radeon HD7950 can be put to good use.

Why in the world would I want to take the massive quality hit of GPU encoding? Maybe that's okay for you, but those of that care about the quality of our work stick to CPU encoding.

Way to move the goalposts, first you were saying to save money over Intel to buy an SSD, now you say people should buy a $300 video card to make up the difference. Buy wait, I thought Trinity's GPU was great??

Next argument please. Oh wait, Anand's review is out, your spin on that is eagerly anticipated.

 

[Hacp]

Why in the world would I want to take the massive quality hit of GPU encoding?

Except there is no quality hit?

 

[Phynaz]

Except there is no quality hit?

What? Now I know you can't be serious. Everybody knows there is a quality hit, it's why the higher end products don't even use GPU encoding.

http://techreport.com/review/23324/a-look-at-hardware-video-transcoding-on-the-pc

 

[pelov]

Except there is no quality hit?

There can be, though this depends on the program and format in question. In general, CPU encoding is better than GPU or hardware accelerated fixed-function encoding like QuickSync. In certain applications there is no quality hit with QuickSync or openCL, whereas in others it might favor one or the other (and CUDA as well).

In general, QuickSync is faster than GPU-acceleration but it's not necessarily better quality as that depends on the program. Straight CPU encoding w/o QuickSync is always better albeit slower than all of the others.

 

[Hacp]

What? Now I know you can't be serious. Everybody knows there is a quality hit, it's why the higher end products don't even use GPU encoding.

http://techreport.com/review/23324/a-look-at-hardware-video-transcoding-on-the-pc

A direct quote.

Aside from slight variations in artifact patterns, it's hard to discern much of a difference between the different solutions here. That's good news. It means the OpenCL acceleration doesn't degrade image quality in a noticeable way, regardless of the hardware used.
Handbrake is the only one of our three test apps that doesn't mess with color saturation, too. Really, I'd say it has the best output, hands down.

So no, if it is done right, there will be no difference in quality.

 

[Phynaz]

A direct quote.



So no, if it is done right, there will be no difference in quality.

Awesome, which editing and authoring packages use that version of Handbrake?

 

[Hacp]

Awesome, which editing and authoring packages use that version of Handbrake?

Edit it from the source, then shrink it with handbrake? How hard is that?

 

[Ferzerp]

Nissans and add glass-pack mufflers them with spoilers and call them racers

Wrong vowel IDC...

 

[Ferzerp]

A direct quote.



So no, if it is done right, there will be no difference in quality.

There would only be a quality loss if shortcuts were taken for speed.

Math is still math.

 

[Phynaz]

Edit it from the source, then shrink it with handbrake? How hard is that?

You have never done authoring I take it.

 

[alexruiz]

Except there is no quality hit?

There is a quality hit
How much? Depends on how much time you want to save

Why in the world would I want to take the massive quality hit of GPU encoding? Maybe that's okay for you, but those of that care about the quality of our work stick to CPU encoding.

Way to move the goalposts, first you were saying to save money over Intel to buy an SSD, now you say people should buy a $300 video card to make up the difference. Buy wait, I thought Trinity's GPU was great??

Next argument please. Oh wait, Anand's review is out, your spin on that is eagerly anticipated.

I see you are into video encoding also. I read Scott's review of OpenCL when he posted it, but he forgot something: He tested only one product, ArcSoft's. ArcSoft implementation is fast, but the quality output is borderline acceptable. Corel Video Studio X4 (and X5) have a more polished implementation. Not as fast as ArcSoft's, but quality is quite better. Furthermore, Corel is licensing MainConcept MPEG2 encoder, one of the best. It does much better job at MPEG2 encoding.

No, no goal post moving at all. I said clearly Better SYSTEM for the money I know you just got a HD7850, because I even gave positive feedback on the product you picked. What is a $300 video card doing in my system? Exactly what the $200 is doing in yours. Seems overmatched? System price is still lower. Seems that I am cheating? Deal with it, remember, better system for the money.

Oh, I can hear your answer "The HD7850 is for my son's i5 system" Ok, no problem, point taken, sounds reasonable. The HD7660G can still use openCL, and using good software will be very close to the quality of a good CPU encoder.

If you still want to get into CPU encoding, I would ask to use my FX8120 rig (paid $140 plus tax for the CPU) Right tool for the job right? I also pick the software, and let's stay with corel / mainconcept, or if you want to get snobish, let's use CCE. Well threaded to stay productive, as time is $$ right? Wait, you said you value quality! You are not going to leave those ugly pieces where the camera is moving in there, are you? Let's do some editing. Trim and fix exposure, normalize audio. How about adding a basic menu to the DVD? How is that spinner doing? Let's author it. Can we stay Corel or wanna get snobish and go Vegas Pro or Sony's DVD architect? Keep in mind that DVD architect creates a lot of temporary files. Furthermore, we might not be able to avoid creating the video_ts folder while compiling the VOBs AND create the ISO. How is that spinner doing?

Spin Anand's review, what spin? He basically said it is a tie. Wanna see a spin? How about what the new "AMD witchhunter" gave as conclusion... "If you've been following the x86 processor wars lately, you might be surprised to see that the A10-5800K beats the Core i3-3225 in overall performance, while the A8-5600K ties it. (You probably aren't surprised to see the pitiful Pentium G2120 buried at the bottom left of the scatter plot.)..."
However, he cannot recommend Trnity? Now that is spin!

ps. Compared to an i7 and z77 motherboard cost, what I paid for the FX8120 plus 970 motherboard gives me room for a 120GB Sandisk extreme SSD (toggle nand) and the $100 to jump not one, but 2 video card classes and have the "$300" GPU overmatch.

 

[Siberian]

Looking at tbeir finances, I don't know how they can afford another design cycle.

 

[Idontcare]

Looking at tbeir finances, I don't know how they can afford another design cycle.

LOL, people were saying that of SUN for a decade as well. All it took for them was an eccentric billionare to bail them out and now they can live on to bleed billions from Oracle for another 10yrs or more.

AMD just needs to find itself courted by some eccentric billionaire that is willing to put a ring on it.

 

[pelov]

LOL, people were saying that of SUN for a decade as well. All it took for them was an eccentric billionare to bail them out and now they can live on to bleed billions from Oracle for another 10yrs or more.

AMD just needs to find itself courted by some eccentric billionaire that is willing to put a ring on it.

If we're to believe rumors...

Though with the in-house A6, that option looks unlikely. I wouldn't be surprised to see them lay down a fat bid for some AMD graphics IP, though.

 

[RampantAndroid]

I honestly won't believe Apple switching over. They are probably always considering it...I don't see why they would actually do it, and if my rMBP had come with an AMD chip instead of the Ivy Bridge based quad core, I probably would not have bought it...

 

[pelov]

I honestly won't believe Apple switching over. They are probably always considering it...I don't see why they would actually do it, and if my rMBP had come with an AMD chip instead of the Ivy Bridge based quad core, I probably would not have bought it...

The overwhelming majority of people buying Apple aren't doing it after reading benchmark reviews.

The display is what sells that rMBP, not the processor.

 

[happysmiles]

Wow - trinity is fast enough. So why do we need HSA?

so more programs run more efficiently on using CPU and GPU also that it can be ported/run on a wider range of hardware

 

[NTMBK]

If we're to believe rumors...

Though with the in-house A6, that option looks unlikely. I wouldn't be surprised to see them lay down a fat bid for some AMD graphics IP, though.

I don't know, Apple's push for custom chips actually makes it more likely that Apple would go AMD. AMD have been offering for a while now the option of putting external IP into custom versions of their Bobcat: http://semiaccurate.com/2012/02/02/amd-opens-up-bobcat-to-3rd-party-ip/ Presumably this will extend to Jaguar, too. Apple could optimise a semi-custom chip with AMD, similar to how they did the A4 and A5 chips, to get precisely what they want for their next Macbook Air. And Apple have enough volume to make a custom design worthwhile. It also fits with rumours of Apple looking to go ARM in their future Macbooks- they could put a single Jaguar core, several A6/7 cores, and a big blob of GCN on the same die, glued together by HSA, with the Jaguar core handling legacy x86 code.

 

[NTMBK]

I honestly won't believe Apple switching over. They are probably always considering it...I don't see why they would actually do it, and if my rMBP had come with an AMD chip instead of the Ivy Bridge based quad core, I probably would not have bought it...

The retina display is pretty much the entire reason why Intel are frantically trying to beef up their graphics in Haswell, and why AMD might have a chance. Do you think Apple really wanted to have to include both NVidia and Intel graphics in their rMBP? And do you think such a solution would work in the Macbook Air, with its even smaller thermal envelope? The next Air will have a retina display, with either Haswell GT3, Jaguar, or low power Trinity- and no discrete chip alongside it.

 

[ShintaiDK]

I don't know, Apple's push for custom chips actually makes it more likely that Apple would go AMD. AMD have been offering for a while now the option of putting external IP into custom versions of their Bobcat: http://semiaccurate.com/2012/02/02/amd-opens-up-bobcat-to-3rd-party-ip/ Presumably this will extend to Jaguar, too. Apple could optimise a semi-custom chip with AMD, similar to how they did the A4 and A5 chips, to get precisely what they want for their next Macbook Air. And Apple have enough volume to make a custom design worthwhile. It also fits with rumours of Apple looking to go ARM in their future Macbooks- they could put a single Jaguar core, several A6/7 cores, and a big blob of GCN on the same die, glued together by HSA, with the Jaguar core handling legacy x86 code.

Both Atoms and Bobcat allows 3rd party IP. Its standard business.

Macbooks will not have ARM chips. ARM is so horrible slow it hurts. Not to mention that Apple said OSX and IOS will never merge.

Apple also wants performance and supply guarantee. Something AMD cant offer. Haswell backed by Intels fabs is exactly what Apple desires.

 

[ShintaiDK]

The retina display is pretty much the entire reason why Intel are frantically trying to beef up their graphics in Haswell, and why AMD might have a chance. Do you think Apple really wanted to have to include both NVidia and Intel graphics in their rMBP? And do you think such a solution would work in the Macbook Air, with its even smaller thermal envelope? The next Air will have a retina display, with either Haswell GT3, Jaguar, or low power Trinity- and no discrete chip alongside it.

Jaguar is very slow, Trinity is slow. None of the IGPs is enough if one excludes the other. How fast does a low power Trinity run again? Are we talking the 17W A6-4455M? The 2.1Ghz 1M/2T 2MB cache chip with 256 SP at 327Mhz? Remember Haswell offers a single chip solution in that segment instead of a 2 chip solution.