AMD Carrizo Pre-release thread

[mrmt]

I expect something like that as well. AMD did not flat out denied that Carrizo will hit the desktop, their wording sounded to me as If they carefully blurred the waters in order to get rid of inventory.

That's AMD SOP, no? They didn't flat out denied that the FX line was dead but in fact, we didn't get anything new beyond Vishera. They didn't flat out deny that Mullins was dead despite the thing didn't get any meaningful design win. They won't flat out deny that FM2+ is dead. Given their bad market position, every single, errr, daydreamer that buys a FM2+ board with the hope it will get an opportunity to upgrade in the future is a win for them, much like the daydreamers that were expected for Steamroller or Excavator on AM3+.

 

[Kuiva maa]

That's AMD SOP, no? They didn't flat out denied that the FX line was dead but in fact, we didn't get anything new beyond Vishera. They didn't flat out deny that Mullins was dead despite the thing didn't get any meaningful design win. They won't flat out deny that FM2+ is dead. Given their bad market position, every single, errr, daydreamer that buys a FM2+ board with the hope it will get an opportunity to upgrade in the future is a win for them, much like the daydreamers that were expected for Steamroller or Excavator on AM3+.

That's totally different. AM3+ desktop processors shared dies with big opterons and those never materialized ,therefore no SR and EX on AM3+. FM2+ APUs share dies with their mobile brethren and those are coming. It stands to reason that some desktop SKUs will eventually surface. Not that it will change much in the greater shape of things (BD family arch has beem abandoned, EX is the end of the road).

 

[erunion]

It stands to reason that if socketed chips were coming AMD would have said so. Instead they gave a non-denial, which as mrmt pointed out, is what they always do when seemingly bad news comes to light.

My personal favorite non-denial from AMD ocuured when 28nm Kaveri's delay was leaked. AMD official response was along the lines of "28nm isn't delayed, Kabini will be here soon".

 

[Kuiva maa]

There were even rumours that kaveri was cancelled (on S/A more than 2 years ago) which were proven false. In retrospect I think the whistleblower was misinterpreted, around that time AMD probably realised that for whatever reason they couldn't deliver SR and EX for big servers and AM3+. Other than that and after that point what I recall was that AMD was always expecting kaveri to hit around Q4 '13, it was what, 1 or 2 months late? Really, they will have a bunch of mobile carrizo dies that won't make the cut for laptops, I don't see why 2-3 SKUs are out of the question here, especially when they were originally slated for FM2+ anyway.

 

[mrmt]

That's totally different. AM3+ desktop processors shared dies with big opterons and those never materialized ,therefore no SR and EX on AM3+. FM2+ APUs share dies with their mobile brethren and those are coming. It stands to reason that some desktop SKUs will eventually surface. Not that it will change much in the greater shape of things (BD family arch has beem abandoned, EX is the end of the road).

If HDL is indeed used clock speeds will take another hit, which means that Excavator might be an actual performance regression when compared to Kaveri.

 

[Kuiva maa]

If HDL is indeed used clock speeds will take another hit, which means that Excavator might be an actual performance regression when compared to Kaveri.

Yes, this is plausible and I agree. I am very curious to see the frequencies of laptop units.

 

[Abwx]

It s not plausible, if it was then there would be no 35W Carrizo to replace the 35W Kaveri, there s people explaining us that AMD will replace a 35W part by a less performing 35W part, that s what happen when fud rather than elementary logic is used as basis for an argumentation..

 

[erunion]

There were even rumours that kaveri was cancelled (on S/A more than 2 years ago) which were proven false. In retrospect I think the whistleblower was misinterpreted, around that time AMD probably realised that for whatever reason they couldn't deliver SR and EX for big servers and AM3+. Other than that and after that point what I recall was that AMD was always expecting kaveri to hit around Q4 '13, it was what, 1 or 2 months late? Really, they will have a bunch of mobile carrizo dies that won't make the cut for laptops, I don't see why 2-3 SKUs are out of the question here, especially when they were originally slated for FM2+ anyway.

Obviously not all rumors are true, nor is AMD unwilling to deny untrue rumors.
I was referring to those cases where AMD offers a non-denial: some statement that does not deny the rumor.

You're remembering the second delay of Kaveri, but forgetting the first.
Kaveri was delayed a full year. It should have been the 2013 product following Trinity but became the 2014 product following Richland, which was released as a stop gap but never existed on the roadmap.

This WCCF article from Oct 2012 accurately predicted the delay.
http://wccftech.com/amd-28nm-processors-delayed-2014/
Unfortunately, I can't dig up AMD statement I was talking about earlier.

After Richland was launch, AMD continued insisting Kaveri would be a late 2013 product right up until the end(and even though it was obvious it wouldnt happen). I imagine the company was determined to live up to some roadmap that promised kaveri and/or streamroller in 2013. That was the second delay.

 

[coercitiv]

Obviously not all rumors are true, nor is AMD unwilling to deny untrue rumors.

I think this sentence means the exact opposite of what you were trying to convey.

 

[Shehriazad]

2015

-> Kaveri refresh for FM2+
->BGA Carrizo
->Mobile Carrizo.


#dealwithit.

The CPU side would likely be weaker than Richlands' and Kaveris' "high end" chips, anyway.
So there is nothing to look forward to. Sure the GPU side will be better....but the sales would probably take a hit just from the fact that the CPU side is weaker.

And FM2+/FM2 sales aren't exactly top notch to begin with.

If AMDs plans for Zen are still alive...we might get some new Zen APUs in 2016...but that is just wishful thinking/speculation on my part. For now you might as well consider FM2+ to be in the same "undead" state as AM3+ aka "It's dragging on"

 

[erunion]

I think this sentence means the exact opposite of what you were trying to convey.

I meant to convey "AMD is willing to deny false rumors". I believe that sentences says that, but sorry for a confusing chain of negatives.

 

[AtenRa]

If HDL is indeed used clock speeds will take another hit, which means that Excavator might be an actual performance regression when compared to Kaveri.

People said the same thing for Kaveri because it is made with 28nm planar vs 32nm SOI for Trinity/Richland. And although clocks were in fact lower, performance was higher for Mobile and equal in desktop with a TDP disadvantage.

Desktop
http://www.anandtech.com/show/7677/amd-kaveri-review-a8-7600-a10-7850k/11

Mobile
http://www.anandtech.com/show/8119/amd-launches-mobile-kaveri-apus/3

 

[USER8000]

If the newer AMD mobile chips have better TDP and power balancing between the different parts of the chips as indicated by a slide which has been posted on Anandtech before,I don't see why performance would regress. One of the main problems with some of the earlier AMD chips is that they did not boost highly as the theoretical specs indicated.

 

[mrmt]

If the newer AMD mobile chips have better TDP and power balancing between the different parts of the chips as indicated by a slide which has been posted on Anandtech before,I don't see why performance would regress. One of the main problems with some of the earlier AMD chips is that they did not boost highly as the theoretical specs indicated.

The problem isn't the mobile chips, but with the desktop chips. Kaveri brings more performance per watt than Richland, but it cannot clock as high as it and in the end AMD APU desktop users ended up 2014 with more or less the same performance they had in 2012. Carrizo seems to bring more of the same, as whatever IPC gains will be compensated by the lower HDL clocks, which means desktop users shouldn't see much increase in raw performance, if any, in terms of CPU performance.

 

[USER8000]

The problem isn't the mobile chips, but with the desktop chips. Kaveri brings more performance per watt than Richland, but it cannot clock as high as it and in the end AMD APU desktop users ended up 2014 with more or less the same performance they had in 2012. Carrizo seems to bring more of the same, as whatever IPC gains will be compensated by the lower HDL clocks, which means desktop users shouldn't see much increase in raw performance, if any, in terms of CPU performance.

I thought it was not coming to desktop??

Well TBH,I actually found something like the A8 7600 and A10 7800 more attractive than Richland since you could build a small and relatively general purpose box which could do some light gaming reasonably OK.

 

[mrmt]

I thought it was not coming to desktop??

Well TBH,I actually found something like the A8 7600 and A10 7800 more attractive than Richland since you could build a small and relatively general purpose box which could do some light gaming reasonably OK.

Since you talked about performance regression I assumed you were talking about desktops, not mobile. HDL raises density (reducing costs) at the expense of maximum clocks, hence Carrizo should not be something worth of note for the desktop user.

 

[Idontcare]

I thought it was not coming to desktop?

You know some random OEM will shove one into a small-volume no-name desktop form factor and whether it sells 1 or 50 units, the mere fact that a BGA carrizo-based desktop form factor can be bought somewhere in some obscure online e-store will be vaunted by fanboys everywhere as proof positive that AMD did in fact bring it to the desktop after all...:\

This will become a war of words amongst fanboys, to everyone else's chagrin, over the simplest of verbal technicalities. In the meantime the financial reality will be the same for AMD as it is for Via...ever receding in the rear-view mirror, becoming more and more irrelevant to the enthusiast and mainstream consumer.

 

[AtenRa]

The problem isn't the mobile chips, but with the desktop chips. Kaveri brings more performance per watt than Richland, but it cannot clock as high as it and in the end AMD APU desktop users ended up 2014 with more or less the same performance they had in 2012. Carrizo seems to bring more of the same, as whatever IPC gains will be compensated by the lower HDL clocks, which means desktop users shouldn't see much increase in raw performance, if any, in terms of CPU performance.

Perhaps this could be said for single thread performance, but in MultiThread performance Kaveri is way faster than Richland even with lower clocks.

Example, in Cinebench R11.5 MT Kaveri A8-7600 65W TDP 3.1GHz is faster than A10-5800K 3.8GHz Trinity.
Also to note, A10-7700K at default 3.4GHz is faster than Trinity/Richland at 4GHz in this benchmark. So even if Kaveri cannot OC to the same levels like Richland, OC vs OC Kaveri is still faster.

Just to show how much faster an AM3+ FX SteamRoller would be even at 28nm planar. Now imagine how powerful it would be at 20nm SOI, such a loss for the consumers.

 

[ShintaiDK]

You know some random OEM will shove one into a small-volume no-name desktop form factor and whether it sells 1 or 50 units, the mere fact that a BGA carrizo-based desktop form factor can be bought somewhere in some obscure online e-store will be vaunted by fanboys everywhere as proof positive that AMD did in fact bring it to the desktop after all...:\

This will become a war of words amongst fanboys, to everyone else's chagrin, over the simplest of verbal technicalities. In the meantime the financial reality will be the same for AMD as it is for Via...ever receding in the rear-view mirror, becoming more and more irrelevant to the enthusiast and mainstream consumer.

Something like this?

http://www.newegg.com/Product/Produc...82E16883266592

 

[Idontcare]

Something like this?

http://www.newegg.com/Product/Produc...82E16883266592

You can practically hear in the distance that someone, somewhere, is screaming at their LCD that this is all the fault of contra-revenue...:| :\

 

[Erenhardt]

The problem isn't the mobile chips, but with the desktop chips. Kaveri brings more performance per watt than Richland, but it cannot clock as high as it and in the end AMD APU desktop users ended up 2014 with more or less the same performance they had in 2012. Carrizo seems to bring more of the same, as whatever IPC gains will be compensated by the lower HDL clocks, which means desktop users shouldn't see much increase in raw performance, if any, in terms of CPU performance.

Kaveri is TDP constrained.
If Carrizo could improve graphics performance thanks to better bandwidth efficiency (40%), the improved CPU efficiency could result in a nice boost.

Have in mind that in kaveri CPU and GPU fight for each W.

But it all comes down to what TDP will Carrizo get. At 90W TDP, carrizo will blow kaveri out of the water. Not only it could sustain higher clocks when igp is used, by the igp itself could be that much faster (not rly ):

 

[mrmt]

Kaveri is TDP constrained.
If Carrizo could improve graphics performance thanks to better bandwidth efficiency (40%), the improved CPU efficiency could result in a nice boost.

Hard to believe that when AMD itself isn't saying it will bring Carrizo to the desktop. In fact, they do not have even a recent desktop roadmap. Their desktop business sucks so bad that they have half a year worth of chip production choking the channel, and they are mostly Richland and Jaguar chips.

We should wait in may for the FAD. It will be when Lisa will show what 15% will do to the future of AMD.

 

[DrMrLordX]

Perhaps this could be said for single thread performance, but in MultiThread performance Kaveri is way faster than Richland even with lower clocks.

If you think the difference in Cinebench 11.5 is dramatic, take a look at y-cruncher. My 7700k's 1g time @ 4.7 ghz is disturbingly close to a stock 8350's time according to the y-cruncher database. I'm nipping at the heels of a 4 ghz 8350 with half the modules.

Kaveri is TDP constrained.

Have in mind that in kaveri CPU and GPU fight for each W.

Bollocks. That is just not true. Something in Windows is pegging the chip to its p5 state under iGPU load, period (probably a combination of ACPI drivers and microcode in the CPU itself, but still, I haven't been able to get iGPU-induced throttling to the p5 state to work under Linux. The Linux ACPI driver won't allow it/won't do it, even without me using cpufrequtils).

All the noise about Kaveri being TDP or cTDP constrained stem from people trying to push past 4.5 ghz on Gigabyte and/or ASRock boards, the p5 state throttling, and some people drawing erroneous conclusions.

Here is the bottom line: Under Windows, set your p5 state to be where you want it to be and the CPU will not throttle under heavy iGPU load, even if the CPU is also under load. I can push my 7700k to 4.7 ghz and 1028 mhz iGPU, load up y-cruncher or Prime95 or whatever and run Furmark in the background and not see any loss in performance. It'll run hotter, sure, and it might even crash under all that strain, but the chip isn't going to just start to slow down arbitrarily on account of a TDP constraint.

If you want to go past 4.5 ghz on Kaveri, you pretty much need an a88x-Pro or Crossblade Ranger (or maybe the A88x-Plus, but I make no guarantees.).

 

[NTMBK]

Something like this?

http://www.newegg.com/Product/Produc...82E16883266592

D: KILL IT WITH FIRE

I really feel for anyone who ends up with a PC like that in 2015. Slower than a Pentium 4...

 

[elemein]

D: KILL IT WITH FIRE

I really feel for anyone who ends up with a PC like that in 2015. Slower than a Pentium 4...

But it's got a better iGP...

And it comes with Windows 8.1 with Bing.

With Bing! :awe: