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

[Torn Mind]

Whooohoo, someone had a high hopes. Does "was to have" translate into "I wish it had"? My english is really bad..

Was is a form of "to be". Wish is a different verb with the infinitive "to wish". Under no circumstances should a verb communicating "what is" be confused with "what I want to be", but it is a common mistake in heated discussions to misinterpret that since human biology doesn't play nice with rationality.

In addition, supposed has two subtle differences in meaning, but one use case does not imply any sort of wishing, but rather a discrepancy between what actually is and what was claimed to be what actually is. Indeed, it can be synonymous with "intended" or "speculated", but certainly not cannot mean both at the same time.

The use of a perfect infinitive is the same as the past perfect tense. Kaveri preview news and Kaveri release both happened in the past, with the preview occurring before the release.

I'm not a grammar hawk by nature like English teachers are; it is an acquired taste, but I can figure out good enough to explain it.

 

[pcsavvy]

Wow, once again a thread about Kaveri ends up talking about everything but Kaveri. I find it interesting that whenever AMD comes up, the conversation strays to all matters then what the OP wanted to talk about. For instance is IGP actually a GP or is it some alien artifact left to bamboozle everyone. All I know about IGP at this time is APU's IGP can handle games better than Intel IGP though Intel is catching up. But gee I thought this was to discuss a Kaveri review not the finer points of the English Language, my bad.

 

[inf64]

Wow, once again a thread about Kaveri ends up talking about everything but Kaveri. I find it interesting that whenever AMD comes up, the conversation strays to all matters then what the OP wanted to talk about. For instance is IGP actually a GP or is it some alien artifact left to bamboozle everyone. All I know about IGP at this time is APU's IGP can handle games better than Intel IGP though Intel is catching up. But gee I thought this was to discuss a Kaveri review not the finer points of the English Language, my bad.

I don't think anything will change anytime soon . It's just the way it is, better just get used to it.

One more review, this time by XS forum user msimax. He has been long time OCer on XS forums, his review is pretty good I must say.

 

[monstercameron]

I have browsed a few reviews but I have seen many reviewers do igp and memory overclocks. Does anyone know of such a test?

 

[cytg111]

Reading up on my processor history here (out of the AMD game since K6-3) .. the last K10 iteration was Thuban2 right? And we are talking ~5 years ago on an 45nm node? Bulldozer is on its 4th iteration now and still struggling to rival thuban IPC and barely manages to edge it out in clocks. Correct? If so, howcome? What the hell went wrong? Does Intel hold all the patents to super fast x86 processors or what? Is it all in the processnode? Was thuban 'as good as it gets' on that node and given the mess with glofo there is just no room to improve on? Is it a matter of physics rather than anything else? (then why not stay with thuban and save the R&D money on bulldozer fail)

 

[Enigmoid]

Reading up on my processor history here (out of the AMD game since K6-3) .. the last K10 iteration was Thuban2 right? And we are talking ~5 years ago on an 45nm node? Bulldozer is on its 4th iteration now and still struggling to rival thuban IPC and barely manages to edge it out in clocks. Correct? If so, howcome? What the hell went wrong? Does Intel hold all the patents to super fast x86 processors or what? Is it all in the processnode? Was thuban 'as good as it gets' on that node and given the mess with glofo there is just no room to improve on? Is it a matter of physics rather than anything else? (then why not stay with thuban and save the R&D money on bulldozer fail)

Bulldozer is simply a fail architecture. Jaguar has similar IPC (though restricted by lower clocks). Its like intel and the P4 except intel could push clocks relative to the P3 and AMD can't.

Kaveri is a step in the right direction. IPC has increased though AMD is badly hamstrung by GloFo. Kaveri doesn't really improve efficiency on 28 nm bulk vs 32 nm SOI (and this shows when you compare it to richland 65W a10-6700) but it removes the MT penalty quite a bit

 

[monstercameron]

Reading up on my processor history here (out of the AMD game since K6-3) .. the last K10 iteration was Thuban2 right? And we are talking ~5 years ago on an 45nm node? Bulldozer is on its 4th iteration now and still struggling to rival thuban IPC and barely manages to edge it out in clocks. Correct? If so, howcome? What the hell went wrong? Does Intel hold all the patents to super fast x86 processors or what? Is it all in the processnode? Was thuban 'as good as it gets' on that node and given the mess with glofo there is just no room to improve on? Is it a matter of physics rather than anything else? (then why not stay with thuban and save the R&D money on bulldozer fail)

construction cores use less die space and resources while delivering more performance/IPS due to MOAR COARS and MOAR JIGAHERTz strategy

 

[SolMiester]

I might be missing something, but the only benchies I saw in which Kaveri was graphically more powerful than Richland, was in 1080p, and that was under 30 FPS so irrelevant.

 

[ViRGE]

Reading up on my processor history here (out of the AMD game since K6-3) .. the last K10 iteration was Thuban2 right? And we are talking ~5 years ago on an 45nm node? Bulldozer is on its 4th iteration now and still struggling to rival thuban IPC and barely manages to edge it out in clocks. Correct? If so, howcome? What the hell went wrong? Does Intel hold all the patents to super fast x86 processors or what? Is it all in the processnode? Was thuban 'as good as it gets' on that node and given the mess with glofo there is just no room to improve on? Is it a matter of physics rather than anything else? (then why not stay with thuban and save the R&D money on bulldozer fail)

Intel doesn't have a patent on super fast x86 designs, but they do have very, very good design teams. The route Intel is chasing - higher IPC with only slight clockspeed improvements - is very difficult and fairly expensive for the R&D involved. AMD literally can't afford to beat Intel at their own game, so if they did the same thing as Intel they would always be second-fiddle to Intel. They had to try something different, and something different in this case didn't really work.

Whether AMD would have been better off evolving K10 as Intel has with P6/Core is a question only AMD's engineers can really answer. But it's hard to imagine a K10 evolution doing worse than BD in everything but the most ideal MT integer workloads.:|

You're right though about process node playing a part. It's hard to make precise comparisons when Intel and AMD don't use the same fab (or even the same fab tech on the same node sizes), but GloFo has a pretty lousy image and they have earned it. Intel's process advantage means they can produce processors of equivalent or better performance that use less power, all the while being cheaper to produce over the long run (die size). If Intel is a first tier fab and TSMC a second tier, then GloFo would be third tier.

Bulldozer is simply a fail architecture. Jaguar has similar IPC (though restricted by lower clocks). Its like intel and the P4 except intel could push clocks relative to the P3 and AMD can't.

Kaveri is a step in the right direction. IPC has increased though AMD is badly hamstrung by GloFo. Kaveri doesn't really improve efficiency on 28 nm bulk vs 32 nm SOI (and this shows when you compare it to richland 65W a10-6700) but it removes the MT penalty quite a bit

It's something of a pessimist view, but you're not wrong. BD is Netbursty by design, and that has worked out about as well for AMD as it did for Intel.

At the same time I think AMD was expecting the real-world IPC to be higher than what they got, so there's a question about whether performance is matching their initial predictions when they settled on going the BD route. Some of the decisions AMD has made only make sense if IPC was expected to be higher (particularly the shared decoder in pre-SR designs).

 

[B-Riz]

I got a few pages in and realized something; everyone is nitpicking this thing for what it is not.

We persons, on this forum, live and breathe the computer tech everyday, we know the process changes and differences between generations, we care about gritty details, we however, are not the market for this product.

And the nitpickers would not buy one anyway for their intended purpose (PC Gaming!), they already have their i5 or i7 and super bad ass graphics card, or dedicated card for that matter.

Any large PC integrator / seller is going to love this chips and sell it to someone ON A BUDGET getting a new box with Windows on it.

It is x86, it is fast enough for everyday use and has some really good graphics.

The AT summary was spot on, it is perfect for Grandma and Grandpa or a non-enthusiast person needing / wanting a new computer without paying the Intel tax.

I am going to pick a low end chip up sometime to play around with it, but I am not going to bitch about what it does not do when I have the i7 and 7950 for everything else...

AMD should be praised for raising the capabilities of the low end inexpensive cpu / apu.

Lest we all forget the lovely Celeron 300 slot 1; look how far the low end has progressed, the entry point for affordable capable computers has never been cheaper, but the chips have gotten much much better.

 

[monstercameron]

just thought that this would be perfect for the unused memory controllers on kaveri althought I dont know if this is technically possible, platform and controller itself.
http://www.legitreviews.com/sandisk-gets-first-ulltradimm-ssd-design-win-ibm_134177

 

[el etro]

If AMD wnat to blow i3 to water, why not to sell a unlocked a8 with 65w or 95w tdp and 512 GCn cores $20 more expensive?

 

[monstercameron]

If AMD wnat to blow i3 to water, why not to sell a unlocked a8 with 65w or 95w tdp and 512 GCn cores $20 more expensive?

wait you mean the a10? the current pricing is just the release price with BF4 combo, the price will come down soon enough!

 

[jpiniero]

The AT summary was spot on, it is perfect for Grandma and Grandpa or a non-enthusiast person needing / wanting a new computer without paying the Intel tax.

Bay Trail would be better for them really.

That being said, OEMs might like it because it makes the discrete GPUs they typically sell kind of redundant. The 6750 that Anand used in the review is probably faster than 90% of discrete GPUs sold by OEMs these days.

 

[monstercameron]

Bay Trail would be better for them really.

That being said, OEMs might like it because it makes the discrete GPUs they typically sell kind of redundant. The 6750 that Anand used in the review is probably faster than 90% of discrete GPUs sold by OEMs these days.

by that logic why not go further and say an ARM based soc would be better for them?

 

[NTMBK]

by that logic why not go further and say an ARM based soc would be better for them?

Because ARM based SoCs don't run Microsoft Office. (Apart from the Surface, and I really wouldn't inflict Windows RT on a grandparent.)

 

[monstercameron]

Because ARM based SoCs don't run Microsoft Office. (Apart from the Surface, and I really wouldn't inflict Windows RT on a grandparent.)

because they can tell the difference?

 

[monstercameron]

interesting, different boards, drastically different power profiles.
http://uk.hardware.info/reviews/513...d-asrock-asus-and-gigabyte-energy-consumption

and there was hullabaloo about AMD not staying near TDP...

 

[B-Riz]

Bay Trail would be better for them really.

That being said, OEMs might like it because it makes the discrete GPUs they typically sell kind of redundant. The 6750 that Anand used in the review is probably faster than 90% of discrete GPUs sold by OEMs these days.

Respectfully, Bay Trail is irrelevant as it is a different market segment compared to this chip.

OEMs using this chip in "comparable" current systems sold at Best Buy and Wal-mart are not using and have not used discrete GPUs for years.

Discreet GPUs are just extra cost and hassle for an integrator.

This new AMD chip has graphics comparable to the $50 replacement a B&M retailer sells that someone may buy to upgrade from that mediocre Intel HD.

Big picture, go to Best Buy tonight and look at the entry level mid tower desktops (the market for this chip). They will be Celeron or Pentium with the Intel HD. Go to the vid cards and see what is sold for $50 - $60, the equivalent to this new chips graphics capabilities.

 

[el etro]

wait you mean the a10? the current pricing is just the release price with BF4 combo, the price will come down soon enough!

I forgot this. Anyway i still waiting the launch of that hypothetical processor. Would cannibalize even their own a10 sales...

 

[NTMBK]

because they can tell the difference?

They try to install any piece of Windows software that didn't come from the Microsoft Store, and it doesn't work?

 

[monstercameron]

They try to install any piece of Windows software that didn't come from the Microsoft Store, and it doesn't work?

because grandma will be installing the lastest version of handbrake regularly. The point is set and forget for non-technical users. They hardly ever explore the system. Install the critical programs like MS office, a few card games and an internet browser and they are good.

 

[frozentundra123456]

I got a few pages in and realized something; everyone is nitpicking this thing for what it is not.

We persons, on this forum, live and breathe the computer tech everyday, we know the process changes and differences between generations, we care about gritty details, we however, are not the market for this product.

And the nitpickers would not buy one anyway for their intended purpose (PC Gaming!), they already have their i5 or i7 and super bad ass graphics card, or dedicated card for that matter.

Any large PC integrator / seller is going to love this chips and sell it to someone ON A BUDGET getting a new box with Windows on it.

It is x86, it is fast enough for everyday use and has some really good graphics.

The AT summary was spot on, it is perfect for Grandma and Grandpa or a non-enthusiast person needing / wanting a new computer without paying the Intel tax.

I am going to pick a low end chip up sometime to play around with it, but I am not going to bitch about what it does not do when I have the i7 and 7950 for everything else...

AMD should be praised for raising the capabilities of the low end inexpensive cpu / apu.

Lest we all forget the lovely Celeron 300 slot 1; look how far the low end has progressed, the entry point for affordable capable computers has never been cheaper, but the chips have gotten much much better.

You have hit on the problem, actually, but I view it much differently than you. People that want cheap, inexpensive boxes dont need the graphics power of an APU, a pentium, or even celeron is fine for them. The APUs are kind of "tweeners" that are more gpu than 90+ % of the users need, but not really powerful enough for uses who want graphics power. They are relegated to a niche, those who want as much graphics power as they can get without adding a discrete card. It also doesnt help that the low end "dual core" APUs with only one module are pretty weak in cpu performance. Their real place is in mobile, but so far, due to using a lot of power, they lose much of their performance advantage in the mobile space. Perhaps mobile Kaveri will improve on this, but that is yet to be seen, and by the time it comes out, it will be nearly time for Broadwell mobile.

 

[monstercameron]

some compute programs
http://developer.amd.com/community/application-showcase/

 

[Lonbjerg]

some compute programs
http://developer.amd.com/community/application-showcase/

I think they need to take a look at that PR site again.
This link was under VLC:
http://openclnews.com/apps/P60

Try clicking it ^^

AMD PR "FTW" lol