The Kaveri Pre-Launch Thread (A10-7800 and A10-6800k @3,5 Ghz)

[Abwx]

The problem is they won't get much extra performance from the extra 30W they have on the 95W Kaveri's. Less than 10% in all probability.

From 65 to 95W there s 46% variation , increasing frequency
by 15% at same voltage would increase TDP to 75W and then
increasing voltage by 12.5% would yield 95W TDP.

That s not the most favourable case since 10% higher voltage
is quite a lot but still would allow 20% higher frequency within
a 95W TDP envelloppe so your 10% are rather the most pessimistic
estimation..

 

[ggadrian]

While I was googling Kaveri last night, I came across another discussion on the Linustechchips website. They said that, according to the text accompanying those benchmarks, that the 6800K had been underclocked (their word normalized) to 3.5 GHz, to match the Kaveri clock speed. Presumably to compare how much better or worse the new architecture would fare at the same clock speed.

Discussion here: http://linustechtips.com/main/topic/97361-first-kaveri-apu-benchmarks-surface/

If so, I can see why this can cause some confusion amongst this crowd, assuming this 'revelation' is correct. Especially since, if indeed the Richland was underclocked, i.e. how would the Kaveri fare against a fully clocked Richland?

I do not know how to read Chinese, though, so I can neither confirm or deny what the associated text says r.e. the 3.5 GHz normalization thing... just thought I'd share that tidbit for the collective here.

I am waiting, along with the rest of us, for the day that NDAs are lifted and we see more benchmarks from the 'usual suspects'. In the meantime, it's an interesting sneak peek, if nothing else. And who knows how 'reputable' this source is in any case? For example, were we dealing with 'alpha', or more mature, Kaveri drivers? So many things remain unclear...

AMD's CES press conference is in less than two days now. I am curious as to when the Kaveri NDA's will actually be lifted? I've seen multiple dates now, but Jan 6th seems like a 'natural' time to lift those. Of course, work on Kaveri's 'release' drivers might not be fully ready by the 6th... in which case we may be waiting a bit longer. We've seen Jan 13th, 14th, and 24th as other potential 'launch' dates.

Edit: Here's the associated 'translated' text from the baidu.com article:

About the launch date, is currently listing the A10-7700K as available the 14th of this month at a price of 151€ (205$).

If someone understands spanish here's the link.

If the price listed is correct, I believe that they're not going to sella lot of them because it's ridiculously expensive for it's price. Here do you have other CPU prices in the same web for comparison (tax included):

-A10-5800K 99€ (135$)
-A10-6800K 128€ (174$)
-i3-4130 105€ (143$)
-i5-4440 155€ (211$)

So we're talking that this APU (not event the top-of-the-line one) is priced in the i5 range, good luck with that. I just hope that the price is wrong and the fix it in the release date or AMD is not going to sell even one of those.

 

[inf64]

While I was googling Kaveri last night, I came across another discussion on the Linustechchips website. They said that, according to the text accompanying those benchmarks, that the 6800K had been underclocked (their word normalized) to 3.5 GHz, to match the Kaveri clock speed. Presumably to compare how much better or worse the new architecture would fare at the same clock speed.

Discussion here: http://linustechtips.com/main/topic/97361-first-kaveri-apu-benchmarks-surface/

If so, I can see why this can cause some confusion amongst this crowd, assuming this 'revelation' is correct. Especially since, if indeed the Richland was underclocked, i.e. how would the Kaveri fare against a fully clocked Richland?

Read the topic from the start, this has also been discussed.

 

[Homeles]

About the launch date, is currently listing the A10-7700K as available the 14th of this month at a price of 151€ (205$).

If someone understands spanish here's the link.

If the price listed is correct, I believe that they're not going to sella lot of them because it's ridiculously expensive for it's price. Here do you have other CPU prices in the same web for comparison (tax included):

-A10-5800K 99€ (135$)
-A10-6800K 128€ (174$)
-i3-4130 105€ (143$)
-i5-4440 155€ (211$)

So we're talking that this APU (not event the top-of-the-line one) is priced in the i5 range, good luck with that. I just hope that the price is wrong and the fix it in the release date or AMD is not going to sell even one of those.

Pre-launch price listings are almost always inflated.

 

[blackened23]

If the IPC improvements are truely this significant, I dont understand the logic behind leaving the FX lineup to the wasteside. Why not 6 and 8 core variants of these improvements? Or am i missing the fact that this is exclusive to the fact that APU's incorporate a GPU which gives them a computational edge. I am a newbie to the whole onboard graphics concept in a system where dedicated graphics is almost certainly used.

IPC improvements? It looks like a disappointment as far as I can tell.

This is a desktop LGA A10 CPU being beat my a mobile i3. A. Desktop. LGA. CPU. How does a desktop CPU get beat by a mobile chip? If that leak is true, it's not going to be a good state of affairs for the Kaveri whatsoever.

 

[inf64]

IPC improvements? It looks like a disappointment as far as I can tell.



This is a desktop LGA A10 CPU being beat my a mobile i3. A. Desktop. LGA. CPU. How does a desktop CPU get beat by a mobile chip? If that leak is true, it's not going to be a good state of affairs for the Kaveri whatsoever.

1st of all you are looking at the ST test. What happens in MT test? right, mobile part is destroyed performance wise. 2nd that mobile i3 IvyBridge is running the ST test at 2.6Ghz. 3rd it "beat" it by 91/88=1.03 or 3%, hardly a beating since this is miniscule difference at best. We can say it's a rough match. 4th the IB core has 2x SIMD resources per core. And 5th the "lowly" mobile part also was on par with x5650 Nehalem part that runs this test at 3.06Ghz. Facts are sometimes funny huh.

Kaveri will have no problems in its market segment since 6800K has no problems in that same market segment. IPC is 10-15% better on average, clocks are ~10% lower. Overall it should be faster part, in some workloads much faster as there will be outlier x86 tests that show difference north of 30% (check OP).

 

[Vesku]

It's a mobile i5 not i3 and as inf64 points out it can ST turbo to 2.6GHz which is why it scores well against a *gasp* server Intel Xeon chip in addition to the A10-7800.

Still, it seems to me that if this preview is accurate then AMD has failed to get much out of GlobalFoundries 28nm process.

 

[inf64]

My bad, i5 .

 

[blackened23]

1st of all you are looking at the ST test. What happens in MT test? right, mobile part is destroyed performance wise. 2nd that mobile i3 IvyBridge is running the ST test at 2.6Ghz. 3rd it "beat" it by 91/88=1.03 or 3%, hardly a beating since this is miniscule difference at best. We can say it's a rough match. 4th the IB core has 2x SIMD resources per core. And 5th the "lowly" mobile part also was on par with x5650 Nehalem part that runs this test at 3.06Ghz. Facts are sometimes funny huh.

Kaveri will have no problems in its market segment since 6800K has no problems in that same market segment. IPC is 10-15% better on average, clocks are ~10% lower. Overall it should be faster part, in some workloads much faster as there will be outlier x86 tests that show difference north of 30% (check OP).

You can spin it however you want but Kaveri is not going to change the medium for which APUs are used. A desktop LGA CPU should not be beat by a mobile chip. The Kaveri will have the same niche as the A10-6800k has - Light usage and HTPCs. Those who thought it was going to be a massive step beyond the A10-6800k were clearly wrong - the IPC is higher which will be offset by lower clockspeeds. One step forward two steps back.

It will be an okay HTPC chip. A very good one in fact. But the most important market for combined CPU/GPU chips are mobile, and AMD won't make any meaningful headway there. Haswell will still be better by a lot, and Broadwell will be even better still.

I was hoping AMD would make better headway for the mobile market, but if these results are accurate, I can't imagine much will change. The mobile APUs will be severely cut down from their desktop variants, and if the desktop LGA APU performs like this - that gives us a rough idea of what the mobile APUs will be like as well.

 

[Arkaign]

My bad, i5 .

It's allright, that "i5" is just a rebadged dual-core + HT POS anyway lol. I've long had problems with what they stick 'U' on. That mobile "i5" would get utterly annihilated by a basic desktop i3. Hell, it would lose against the standard mobile i3 (non-'U' edition).

The 'U' products exist in a weird alternate universe. That's why you have sub 2Ghz dual core "i7" models. Stupidity.

 

[Arachnotronic]

AMD should change its ticker symbol on the NYSE to "VIA".


No thread crapping. Go to OT if you want to do this.
Markfw900

 

[Arachnotronic]

IPC improvements? It looks like a disappointment as far as I can tell.

This is a desktop LGA A10 CPU being beat my a mobile i3. A. Desktop. LGA. CPU. How does a desktop CPU get beat by a mobile chip? If that leak is true, it's not going to be a good state of affairs for the Kaveri whatsoever.

Vastly broader/deeper R&D dollars, a two generation process lead (22nm + FinFET), and a better design to build off of?

Not sure why hope springs eternal for AMD. They're finished. Will be interesting to see BYT-M eat into AMD's main PC revenue stream next year, dealing the final blow to this disappointment of a company.

 

[jpiniero]

The 'U' products exist in a weird alternate universe. That's why you have sub 2Ghz dual core "i7" models. Stupidity.

You do realize that it's only a matter of time that the majority of Intel's revenue comes from 17W and lower parts?

 

[NTMBK]

It's allright, that "i5" is just a rebadged dual-core + HT POS anyway lol. I've long had problems with what they stick 'U' on. That mobile "i5" would get utterly annihilated by a basic desktop i3. Hell, it would lose against the standard mobile i3 (non-'U' edition).

The 'U' products exist in a weird alternate universe. That's why you have sub 2Ghz dual core "i7" models. Stupidity.

Intel's logic (not that I agree with it) is that the i3/5/7 badge is an indication of relative performance within a given market. The Ultrabook i7 is the best you can get in an ultrabook; the mobile quad core i7 is the best you can get in a full sized laptop; and the desktop i7 is the best you can get in a desktop.

I wish that they would bring back the "M" branding, so people can distinguish better. Core i5-M.

 

[Asterox]

IPC improvements? It looks like a disappointment as far as I can tell.

This is a desktop LGA A10 CPU being beat my a mobile i3. A. Desktop. LGA. CPU. How does a desktop CPU get beat by a mobile chip? If that leak is true, it's not going to be a good state of affairs for the Kaveri whatsoever.

If you are wise and do not jump ahead of time, then one crapy benchmark as Cinebench R15(powered by Intel ICC) means nothing.For a more accurate assessment of something, you need more samples but from reliable sources and obviously you've never heard of such a practice.

Here's a great example of Cinebench precision, here are two examples in which slower processor outperforms faster clocked CPU from the same CPU family and this is completely logical?It should be noted there is nothing faked on this photos, so if you have only one Cinebench test in which you'll believe if you do not have nothing reliable for good comparison?

 

[Enigmoid]

The scary thing is if Kaveri is having trouble boosting at 3.5 Ghz + 512 GCN shaders at 720mhz at 65W then I cringe when I think about how much the clocks will have to be reduced at 35W. A10-5750m gets around 2.7 ghz (fully loaded) to 3.2 ghz (1 or 2 cores) which means major clock reductions may be seen on mobile. AMD really needs a 45W mobile SKU to compete with intel for the larger notebooks (15.6" or 17" or high-powered systems like the MSI GX series).

 

[littlebluesmurf]

If you are wise and do not jump ahead of time, then one crapy benchmark as Cinebench R15(powered by Intel ICC) means nothing.For a more accurate assessment of something, you need more samples but from reliable sources and obviously you've never heard of such a practice.

Here's a great example of Cinebench precision, here are two examples in which slower processor outperforms faster clocked CPU from the same CPU family and this is completely logical?It should be noted there is nothing faked on this photos, so if you have only one Cinebench test in which you'll believe if you do not have nothing reliable for good comparison?

Kaveri is also supposedly slower than a lower clocked Trinity going by your image. :whiste:

I also ran the same benchmark on my A4 5300 which boosts to 3.6Ghz and I managed to get a score of 75. That leaves around a 11% clock speed difference between Kaveri boost speed and this Trinity A4 boost speed. 75 + 11% = ~ 83.

Kaveri either brings little ipc improvement or the supposed Kaveri screenshot is a fake.

 

[Homeles]

The scary thing is if Kaveri is having trouble boosting at 3.5 Ghz + 512 GCN shaders at 720mhz at 65W then I cringe when I think about how much the clocks will have to be reduced at 35W. A10-5750m gets around 2.7 ghz (fully loaded) to 3.2 ghz (1 or 2 cores) which means major clock reductions may be seen on mobile. AMD really needs a 45W mobile SKU to compete with intel for the larger notebooks (15.6" or 17" or high-powered systems like the MSI GX series).

AMD's at a very, very severe process disadvantage compared to Intel at the moment. There's not really much chance of them competing with Intel for the high end stuff anyway.

However, do keep in mind that lower power chips are almost invariably better binned than their higher TDP counterparts. As process variability continues to increase, this will only become more true over time. Also, frequency scaling is exponential when accounting for voltage, so it may not actually have so much trouble at those clocks, and at that TDP.

 

[Vesku]

If GF 28nm did well at mobile clocks I think AMD would have launched Kaveri mobile for the Winter shopping season.

Edit: Oh, and we would have seen GF 28nm Kabini chips.

Seems like any improvements AMD makes in chip design just tread water due to node shrink troubles.

 

[AtenRa]

If GF 28nm did well at mobile clocks I think AMD would have launched Kaveri mobile for the Winter shopping season.

Edit: Oh, and we would have seen GF 28nm Kabini chips.

Seems like any improvements AMD makes in chip design just tread water due to node shrink troubles.

Are you sure AMD uses GF28nm now or in the near future ???

 

[Abwx]

Kaveri either brings little ipc improvement or the supposed Kaveri screenshot is a fake.

The main IPC improvement will be on the integer side of things,
it was known that FP wise they ll just remove the MT penalty
hence the 11.7% improvement in Cinebench at same frequency
but other tests suggest an healthy IPC increase in integer
related tasks.

 

[PPB]

The main IPC improvement will be on the integer side of things,
it was known that FP wise they ll just remove the MT penalty
hence the 11.7% improvement in Cinebench at same frequency
but other tests suggest an healthy IPC increase in integer
related tasks.

Wow man go slow with these guys. You will shatter their world if you explain them that IPC FP improvements are different from INT ones.

Let alone show them the importance of Int perf, when most code is integer heavy anyways.

 

[littlebluesmurf]

The main IPC improvement will be on the integer side of things,
it was known that FP wise they ll just remove the MT penalty
hence the 11.7% improvement in Cinebench at same frequency
but other tests suggest an healthy IPC increase in integer
related tasks.

In the screenshot of supposed Kaveri chip we have here the multithreaded score is lower than a lower clocked Trinity part a few posts down . Hence me questioning the validity of this screenshot.

 

[Abwx]

In the screenshot of supposed Kaveri chip we have here the multithreaded score is lower than a lower clocked Trinity part a few posts down . Hence me questioning the validity of this screenshot.

He questioned CB R15 validity since a 3.6GHz Trinity score
6.8% better than a.....3.8GHz Trinity , so his point hold.

But there s another CB FP test that is less variable generaly :

 

[JDG1980]

Not sure why hope springs eternal for AMD. They're finished. Will be interesting to see BYT-M eat into AMD's main PC revenue stream next year, dealing the final blow to this disappointment of a company.

AMD is selling every high-end GPU they can manufacture. They have long-term contracts to supply APUs for both of the biggest next-gen consoles. On top of that, they've got the Mac Pro GPU design win, which should help make multi-GPU OpenCL support mainstream, thus boosting the performance of their APUs across the board as well.

It's absurd to claim that the company is doomed because rumored Cinebench results on Kaveri aren't as good as some people wanted.