AMD - Bulldozer Chitchat

[CAT-THE-FIFTH]

Thanks to Jimbo75 for the link to the following thread:

http://www.xtremesystems.org/forums/...90#post4875190

This post is quite interesting:

http://www.xtremesystems.org/forums/...0&postcount=16

It seems Llano has slightly better IPC than Deneb even though the former lacks L3 cache.

Here are some HWiNFO screenshots showing the correct clockspeeds:

http://forum.coolaler.com/showpost.p...7&postcount=81

http://forum.coolaler.com/showthread...=266652&page=9

[EvanJackPenn]

Hmm. I assume it will be the i7 it will be comparable to then? The Phenom II x6 was barely i5-2500 standard. I want something to kill the i5-2500K's reign of the midpoint processor market.

[badass]

Maybe that's what they mean by aggressive pricing - faster for a much lower price? Don't forget Intel are only charging that because of a lack of competition, they could charge much less if they wanted/needed to.

AMD used to charge $1000 for their old FX processors.
Why would they charge less than $400 if they can beat a $1000 Intel CPU?

[scaryjim]

AMD used to charge $1000 for their old FX processors.
Why would they charge less than $400 if they can beat a $1000 Intel CPU?

Because they had a reputation for selling the best performing chips back then. Today they are coming from a *long* way behind in the performance stakes, so they need to give the enthusiasts who demand the peak of performance a reason to abandon their Intel rigs. I think we can safely assume that Bulldozer isn't going to do to the i7 9xx what the Core 2 did to the Athlon FX, and Intel have a new generation of range topping processors just round the corner anyway, so AMD will not want to appear too ambitious or over-confident about their new processors.

OR

Because they have a new flagship processor that will come out just after the first wave of BDs that *will* wipe the floor with everything else. Don't forget that when Core 2 came out you could get a $200 Core 2 that would outperform the Althon X2 FX - a $1000 processor. But intel couldn't realistically charge $1000 for the *cheapest* Core 2 processor because it simply wouldn't shift enough volume: people would continue to buy the cheaper, if slower, Athlons. So despite their slowest Core 2 being faster than Athlon FX, they sold it a lot cheaper. AMD might be in the same position now, knowing that they will have a new processor entering the top end in the future, and are pragmatically deciding to price their first wave of BD based on that.

OR

Because AMD are a long way behind in market share for x86 desktop and server processors. So if they can launch a processor that performs like an i7 990X, but charge $400 for it and still make a decent profit, they stand a chance of making big market-share wins that will strengthen their overall position - after all, if you were offered two processor with similar performance, but with one costing $1000 and one costing $400, which would you buy? I think AMD have learned a lot about the processor market from losing their position of strength so dramatically, and one of those lesson is that you can be profitable by selling good processors at a reasonable price. So they could be deliberately contracting the market under Intel's feet - saying "You don't need to spend $1000 on a flagship processor - that's old-fashioned thinking".


Or the most plausible option; BD will actually be slower than the i7 990X in most workloads. Quite probably really, since it'll be an 8 thread processor going up against a 12 thread processor - in heavily threaded tests it's going to struggle. It may well beat out Intel's 4 core 8 thread i7s handily (in fact it should, since the Phenom II X6s do), but they're considerably cheaper than $1000...

[CAT-THE-FIFTH]

This thread has an early B0 4 module Bulldozer engineering CPU in a 990FX motherboard:

http://www.chiphell.com/thread-210890-1-1.html

Supposedly,some benchmarks will be posted soon:

http://www.xtremesystems.org/forums/...4&postcount=16

There are also some results from another engineering CPU:

http://www.chiphell.com/thread-210024-1-1.html

However,the system had a few BSODs. This probably means the CPU is a very early buggy engineering sample and hence the results cannot be fully trusted IMHO.

TBH,I don't see the point of these benchmarks if they are early engineering samples with buggy microcode.

[Terbinator]

FWIW, a Core i7 860 scores 7.5 on WIE as well.

[CAT-THE-FIFTH]

Some more scores have appeared in the first thread. It looks like a Phenom II X6 1055T is faster than the B0 engineering sample tested in both SuperPi and Cinebench 11.5! It seems another pointless test. The current CPUs are meant to B1 and B2 steppings.

[watercooled]

Yeah the retail SNB chips performed a fair bit better than the samples Anandtech got IIRC, and they were pretty late samples, not early, first stepping engineering samples.

[Terbinator]

Yeah the retail SNB chips performed a fair bit better than the samples Anandtech got IIRC, and they were pretty late samples, not early, first stepping engineering samples.

Yeah, they were locked at a given frequency <3GHz and no/little Turbo Boost IIRC.

[Blackmage]

I wouldn't even bother with these benches, I don't think performace could be that bad. These are early ES chips!!

[CAT-THE-FIFTH]

Another test of the crippled B0 engineering sample:

http://www.xtremesystems.org/forums/...d.php?t=272022

It seems that the B0 engineering samples are actually running at half their stated clockspeeds like with the Llano engineering samples. The L2 and L3 cache may also not be fully functional in the B0 engineering samples.

[CAT-THE-FIFTH]

It seems Llano is launching soon:

http://www.youtube.com/watch?v=Lkbdy...embedded#at=17

The battery life tests compare an i3-2310M with an A8-3500M. The A8 CPU has a 400 shader IGP.

Both are 35W CPUs:

http://ark.intel.com/Product.aspx?id=52220

http://www.cpu-world.com/CPUs/K10/AM...0A8-3500M.html

[Biscuit]

It seems Llano is launching soon:

http://www.youtube.com/watch?v=Lkbdy...embedded#at=17

The battery life tests compare an i3-2310M with an A8-3500M. The A8 CPU has a 400 shader IGP.

Both are 35W CPUs:

http://ark.intel.com/Product.aspx?id=52220

http://www.cpu-world.com/CPUs/K10/AM...0A8-3500M.html

Marketing gumph, means nothing until real products come out IMO, that wasnt even real laptops it was just 2 images/videos.

[CAT-THE-FIFTH]

Marketing gumph, means nothing until real products come out IMO, that wasnt even real laptops it was just 2 images/videos.

What is interesting is that the AMD system also had a discrete graphics card too and it seems that this was switched off. This is actually quite an interesting feature for an AMD system. Already laptops with the switchable graphics are starting to appear on pre-order:

http://www.passiontec.de/_fenster.ph...&affId=1376965

AFAIK,the A6 and A8 IGPs can also be used in Crossfire with the HD6000 and HD6700 series discrete graphics cards. The HD6690G2 moniker is the name given to the Crossfire combination.

The only actual picture of an A8-3500M in an actual computer is in this Asus AIW computer:

http://news.softpedia.com/newsImage/...ex-2011-4.jpg/

[pp05]

Change of strategy?

Well that's what AMD says, and of course any decision could be labeled as strategic but the Bulldozer processors (Zambezi FX series) have been delayed alright. guru3D was able to confirm that it will take 60 to 90 days from this week before the new processors will see the light of day. Meanwhile there of course will be Llano.

Guido Lohmann, AMD PR Manager Northern Europe, "The very gratifying success of the C-and E-series APUs has led to the decision to go ahead first in the desktop area with an APU, and then to follow with the Bulldozer processors,” concluded the company's rep."

link

[Blackmage]

How much additional performance do you think we will see from the b2 stepping chips? In the area of 30%?