AMD on track for launch of Kaveri in February 2014

[Shivansps]

nevermind

 

[CHADBOGA]

SA forum is a gold mine for special information .

It is also now the premier site for the mentally deranged to peddle all sorts of mad conspiracy theories and profound ignorance of the semi-conductor world.

 

[NTMBK]

It is also now the premier site for the mentally deranged to peddle all sorts of mad conspiracy theories and profound ignorance of the semi-conductor world.

Personally I'd say it's a tie with Alienbabeltech. Thankfully the most utterly deranged conspiracy theorist, mistercteam, got booted from the forum a while back. There are still a few characters whose insane rambling I could do without though.

 

[USER8000]

This could be a typo,but interesting if true:

http://wccftech.com/amd-kaveri-apu-...ce-uplift-richland-platform-details-unveiled/
http://wccftech.com/wp-content/uploads/2013/11/AMD-Kaveri-Architecture.jpg

4 cores,8 threads?

 

[inf64]

Yeah wccftech... They have no clue what they are talking about. Kaveri is 2M/4T part .

 

[mikk]

Up to 20%/30% is a big disappointment. Especially the GPU is close to a fail. The up to claim doesn't bode well, up to projections are usually exaggerated.

 

[inf64]

Ok.

 

[monstercameron]

Up to 20%/30% is a big disappointment. Especially the GPU is close to a fail. The up to claim doesn't bode well, up to projections are usually exaggerated.

trinity wasn't exaggerated. neither was richland but keep on keeping on...

 

[frozentundra123456]

Up to 20%/30% is a big disappointment. Especially the GPU is close to a fail. The up to claim doesn't bode well, up to projections are usually exaggerated.

Personally I am skeptical of those numbers, so I would reserve judgement until we see real data. But if they are correct, I would consider it underwhelming as well. Gpu performance would still be well below a HD 7750 gddr5, and CPU performance well below an Intel i5. So on the desktop it would still be in the limbo area of more gpu than casual users need and beaten by a low end discrete card. The saving grace could be if they can work out the asymmetric crossfire issues so that it would give a nice boost with a lower end card like a 7750/7770.

 

[monstercameron]

Personally I am skeptical of those numbers, so I would reserve judgement until we see real data. But if they are correct, I would consider it underwhelming as well. Gpu performance would still be well below a HD 7750 gddr5, and CPU performance well below an Intel i5. So on the desktop it would still be in the limbo area of more gpu than casual users need and beaten by a low end discrete card. The saving grace could be if they can work out the asymmetric crossfire issues so that it would give a nice boost with a lower end card like a 7750/7770.

so you are worried that a cheaper than intel i5 part will be slower? what about i3? also how can the igp be underwhelming if it performs better than the already decent a10-6800k?

the intel fans expectations are too high, I wonder if it is because neither intel nor amd care anymore to indulge their e-peen fantasies?

 

[CHADBOGA]

so you are worried that a cheaper than intel i5 part will be slower? what about i3? also how can the igp be underwhelming if it performs better than the already decent a10-6800k?

the intel fans expectations are too high, I wonder if it is because neither intel nor amd care anymore to indulge their e-peen fantasies?

Weren't desktop users right to be underwhelmed by Haswell's meagre IPC increase?

What's the difference here?

 

[monstercameron]

Weren't desktop users right to be underwhelmed by Haswell's meagre IPC increase?

What's the difference here?

the difference is that there are so many variables unaccounted for, true-audio, mantle, ipc increase, bandwidth increase, possible power usage decrease, 28nm process and HSA, yet some are still underwhelmed...

 

[frozentundra123456]

so you are worried that a cheaper than intel i5 part will be slower? what about i3? also how can the igp be underwhelming if it performs better than the already decent a10-6800k?

the intel fans expectations are too high, I wonder if it is because neither intel nor amd care anymore to indulge their e-peen fantasies?

I'm not worried, are you? I do in fact consider the i3 overpriced and probably intels most poorly positioned part. It is in the same gray area cpu wise as amds apu are graphics wise. Better than casual users need ( a Pentium is more than enough) but easily exceeded by a low end i5 for not that much more cost.

Oh, and btw, my fantasies involve Jennifer Lawrence, not cpus.

 

[Erenhardt]

Weren't desktop users right to be underwhelmed by Haswell's meagre IPC increase?

What's the difference here?

What are haswell IPC increases? What is projected kaveri IPC increase? Can someone roughly compare those two generations?

What is dedicated PCIe SSD interface?

 

[mikk]

trinity wasn't exaggerated. neither was richland but keep on keeping on...

It was. Richland 40% faster claimed AMD. For Trinity AMD claimed up to 56% faster GPU and up to 29% faster CPU. The average improvements were much less. Up to is a typically marketing trick.

 

[monstercameron]

It was. Richland 40% faster claimed AMD. For Trinity AMD claimed up to 56% faster GPU and up to 29% faster CPU. The average improvements were much less. Up to is a typically marketing trick.

Somewhat surprisingly, AMD isn't making any lofty claims about the performance of the 35W Richland parts versus their predecessors. At CES in January, AMD touted a 40% jump in graphics performance and a 10-20% boost in CPU performance over Trinity, but that was for ultra-low-voltage models aimed at ultrathin laptops. The gains likely won't be as dramatic for APUs with higher thermal envelopes, which suggests Richland's eventual desktop incarnations may benefit the least.

so there goes your richland myth...

also did you really grab those trinity numbers from this article...
http://news.softpedia.com/news/AMD-s-Trinity-56-Faster-Graphics-29-Better-Productivity-262835.shtml

the following articles will be a refresher of trinity
http://www.anandtech.com/show/6332/amd-trinity-a10-5800k-a8-5600k-review-part-1
http://www.bit-tech.net/hardware/cpus/2012/10/03/amd-a10-review/
http://www.fudzilla.com/home/item/29340-amd-a10-5800k-reviewed

 

[Shivansps]

the difference is that there are so many variables unaccounted for, true-audio, mantle, ipc increase, bandwidth increase, possible power usage decrease, 28nm process and HSA, yet some are still underwhelmed...

Mantle, HSA and True-Audio are very, very dependant of devs implementation, its likely not gona worth it on Kaveri lifetime, maybe on the next or the next after that one. Lower power usage is a welcome change, but not a priority on desktop, unless its drastic.

People wants IGP and CPU performance out of a APU right now, thats the only thing, all other features are to set up a base for devs to implement support for them, and its likely gona end up useless as 64bits where to K8.

 

[Erenhardt]

Mantle, HSA and True-Audio are very, very dependant of devs implementation, its likely not gona worth it on Kaveri lifetime, maybe on the next or the next after that one. Lower power usage is a welcome change, but not a priority on desktop, unless its drastic.

People wants IGP and CPU performance out of a APU right now, thats the only thing, all other features are to set up a base for devs to implement support for them, and its likely gona end up useless as 64bits where to K8.

There are already AAA games with TrueAudio comming, like new Thief.
Both consoles have it so... :whiste:
Forza 5 seems to be using TrueAudio tech, but it is xbone exclusive ;/

TrueAudio doesn't need that amd hardware bit you find in some R series graphics card and kaveri. The sound simulation can be done on CPU, if you have enough FLOPS to throw at it.

 

[frozentundra123456]

There are already AAA games with TrueAudio comming, like new Thief.
Both consoles have it so... :whiste:
Forza 5 seems to be using TrueAudio tech, but it is xbone exclusive ;/

TrueAudio doesn't need that amd hardware bit you find on some R series graphics card and kaveri. The sound simulation can be done on CPU, if you have enough FLOPS to throw at it.

That may be true, but are they going to be playing them on an APU? I agree true audio might be nice, but I am not sure it will matter that much to the low/midrange gaming market that APUs appeal to, much less the casual user.

 

[mikk]

so there goes your richland myth...

also did you really grab those trinity numbers from this article...
http://news.softpedia.com/news/AMD-s-Trinity-56-Faster-Graphics-29-Better-Productivity-262835.shtml

You are wrong, I didn't grab it from this article. I grab it from AMD. It was an official claim from a Cebit slide earlier in this year. Up to is a marketing trick, you should better know this.

 

[Erenhardt]

That may be true, but are they going to be playing them on an APU? I agree true audio might be nice, but I am not sure it will matter that much to the low/midrange gaming market that APUs appeal to, much less the casual user.

IMHO hardware accelerated 3d audio is a big deal for cheap low end market. These PC parts are giving all they have to deliver playable performance. Offloading sound processing (not only 3d sound?) can release a big % of CPU time. Of course there may be no interest in it from average joe
If we look at high end i7 users, they don't need it. They can dedicate one or two cores to sound processing (don't know how much impact truesound will have) without having to deal with CPU bottleneck.

 

[NTMBK]

Mantle, HSA and True-Audio are very, very dependant of devs implementation, its likely not gona worth it on Kaveri lifetime, maybe on the next or the next after that one.

I don't know about you guys, but I tend to keep my PCs around for quite a few years. Given that the first Mantle game is hitting in one month's time (before Kaveri even launches!) and plenty more are coming, and the PS4 uses a TrueAudio block so we can probably expect plenty of TrueAudio optimized games, I think AMD are in a pretty solid place software-wise. I'm planning on building a Kaveri HTPC (unless the reviews prove diabolical), and I think it should keep up with console ports for many years to come.

 

[JDG1980]

I would suggest taking all synthetic benchmarks with a very large grain of salt.

When Kyle Bennett reviewed Vishera, he compared it to Zambezi, SB, and IB, all running at the same clock speed (4 GHz). The synthetic benchmarks on page 3 looked very disappointing, with three of the four showing regressions in Vishera compared to Zambezi. However, all the real-world benchmarks (pages 4 and 5) did better on Vishera than on Zambezi, both gaming and productivity.

Until we see reliable, real-world benchmarks published by a reputable site, we're all just guessing. No one really knows how much improvement Steamroller will see over Piledriver. My guess is that the CMT penalty will be almost entirely eliminated, and we will also see roughly 20% IPC improvement in single-threaded apps, with the caveat that this improvement is in comparison to Richland, not Vishera (due to the lack of L3 cache on the APUs). But I could be completely wrong.

 

[Fjodor2001]

So why is Kaveri clock frequency lower than expected? Crappy GF 28 nm process, fatter Steamroller core, or something else?

 

[inf64]

Probably a combination of several things. It's not much lower, 4.1Ghz base vs 3.7Ghz base- that's ~10% difference. Turbo on Kaveri is still not disclosed but it's safe to assume similar ~10% difference there too meaning Turbo will be in ~4Ghz range.

Still this thing has to outperform Richland by 10-15%(pick 12.5% as mean value) as per AMD's yearly big-core roadmap and since clock speeds are lower that only leaves us with IPC increases. You do the math: 0.9 (clock) x Y= 1.125. IPC therefore needs to be in the range of Y~=1.25 or thereabout. Anything lower and Kaveri will not outperform Ricland much (if at all) in x86 benchmarks. Remember PD had minor core changes/fixes and it was solidly faster in apps/games than BD.