AMD - Piledriver chitchat

[CAT-THE-FIFTH]

More details about the Temash based Acer V5 series:

http://semiaccurate.com/forums/showp...&postcount=475



http://www.youtube.com/watch?v=Vet82boI8VM

[CAT-THE-FIFTH]

I wonder where they'll be made? GloFo may be possible at 20nm.

http://www.fudzilla.com/home/item/30...g-on-line-soon

http://www.xbitlabs.com/news/other/d...ap_Report.html

[DanceswithUnix]

Charlie says the new Atom uses a module architecture where 2 paired cores share a common L2 cache instead of the usual HT.

http://semiaccurate.com/2013/05/06/i...ont-atom-core/

[CAT-THE-FIFTH]

Charlie says the new Atom uses a module architecture where 2 paired cores share a common L2 cache instead of the usual HT.

http://semiaccurate.com/2013/05/06/i...ont-atom-core/

The same as Jaguar.

[DanceswithUnix]

The same as Jaguar.

Comparisons at the top end are going to happen though.

I liked Charlie's comment:

What you are seeing here is the difference between a “real core” and a HT thread, can you hear the champagne corks popping at AMD?

[scaryjim]

The same as Jaguar.

Plus you can add up to 4 modules on an SoC, like Jaguar (4 modules in the PS4 chip...)

Question is, can they use the huge marketing muscle that the Intel brand provides to catch up the technological delay behind AMD, who are, after all, currently fabbing chips based on exactly the same modular approach? After all, Intel are only talking about the new architecture, AMD are shipping it, and will have a reasonable amount of time to tweak and improve before the Intel products come to market...

[CAT-THE-FIFTH]

I can see some people on forums already predicting the end of the ARM due to 22NM Atom. LOL!

[kompukare]

I can see some people on forums already predicting the end of the ARM due to 22NM Atom. LOL!

Yep, they're at it again. Strange thing is, nobody mentions ARM #1 advantage in these threads: ARM will licence to anyone, Intel will not. ARM licensees range from 60%+ margin Qualcomm to tiny players happy to make 10%.

That is the main reason ARM is so successful - although there are other players with a similar approach like MIPS but they don't seem to have had as much success (AFAIK raw performance and perf/watt is/was actually better for MIPS).

The second reason is of course, that while Intel may have very deep pockets, ATM they don't have full package a SOC needs in terms of baseband modems etc.

We shall see but Intel coming in cherrypicking the high end with Silvermont might harm Qualcomm et all but Intel's much vaunted process lead does cost Intel a fair bit too. Their cap-ex has been very high for the last few years and even with their revenues they cannot use all their capacity. Idling $billion fabs is never a good thing even for the Intel cash machine.

[watercooled]

I can see some people on forums already predicting the end of the ARM due to 22NM Atom. LOL!

Usual morons who fail in so many ways to understand the market.

On the subject of MIPS, I wonder if we'll see continued development of their application processors? They would probably end up being produced by the likes of TI/Qualcomm so might not have a huge impact on the market/competition, but still. Also, it would be a fairly hard market to penetrate, especially with a different ISA.

MIPS currently don't really compete against ARM in terms of raw performance with their IP cores - the likes of proaptiv looked interesting but AFAIK haven't made it into production, and we know what on-paper specs can be like. I'm not so sure about efficiency, but in the mobile market, MIPS is only really found in a few cheaper devices so it's hard to gauge as they're older cores often built on older process nodes.

[CAT-THE-FIFTH]

The FX4350 and FX6350 are tested:

http://www.pcgameshardware.de/FX-635...-4350-1068215/

The FX6350 is faster and consumes less power than a Phenom II X6 1100T.

[Biscuit]

The FX4350 and FX6350 are tested:

http://www.pcgameshardware.de/FX-635...-4350-1068215/

The FX6350 is faster and consumes less power than a Phenom II X6 1100T.

Looking like very good value chips there for mid-range folk. I'm a bit disappointed AMD haven't managed to do anything about the power usage, would have been nice to see a significant watt per clock drop in some of the second round of PD chips.

[DanceswithUnix]

Usual morons who fail in so many ways to understand the market.

On the subject of MIPS, I wonder if we'll see continued development of their application processors? They would probably end up being produced by the likes of TI/Qualcomm so might not have a huge impact on the market/competition, but still. Also, it would be a fairly hard market to penetrate, especially with a different ISA.

MIPS currently don't really compete against ARM in terms of raw performance with their IP cores - the likes of proaptiv looked interesting but AFAIK haven't made it into production, and we know what on-paper specs can be like. I'm not so sure about efficiency, but in the mobile market, MIPS is only really found in a few cheaper devices so it's hard to gauge as they're older cores often built on older process nodes.

MIPS seems to be the ISA of choice of the Chinese government, with IIRC two homegrown implementations. ARM has made inroads, but I think MIPS is still the norm in broadband/adsl routers and there are a lot of those around. Will be interesting to see what Imagination do with the company, but overall I don't see them going away.

Just had a google for Chinese mips stuff, seems to be a few laptops/netbooks made and they have added instructions to help emulate x86. Guess that was inevitable, but it seems a shame to taint a proper cpu with x86 related cruft

[HalloweenJack]

The FX4350 and FX6350 are tested:

http://www.pcgameshardware.de/FX-635...-4350-1068215/

The FX6350 is faster and consumes less power than a Phenom II X6 1100T.

finally - I mean that . Had my 1090T @ 3.7ghz for nearly 18 months now!

[DanceswithUnix]

The quad core, presumably thanks to its higher turbo, is still beating the 6 core version though albeit not by much now.

Interesting to compare the FM2 socket X4 750K results with the 4350. The 4350 gets between 25% and 40% better performance. Either the X4 doesn't turbo much, or the lack of cache is really hurting it (or both).

[scaryjim]

... Either the X4 doesn't turbo much, or the lack of cache is really hurting it (or both).

I'd guess both. The 4GHz Turbo for the Athlon is single-thread peak turbo, there'll be a lower peak turbo (probably 3.7GHz, since it appears to all intents and purposes to be an A10-5700 without the GPU) for all cores. So in multi-threaded tests, it'll (proabbly ) be peaking at 3.7GHz while the 4350 will be sat at at least 4.2GHz all the time. That's a 13% clock speed deficit on heavily threaded tests.

[CAT-THE-FIFTH]

Yep, they're at it again. Strange thing is, nobody mentions ARM #1 advantage in these threads: ARM will licence to anyone, Intel will not. ARM licensees range from 60%+ margin Qualcomm to tiny players happy to make 10%.

That is the main reason ARM is so successful - although there are other players with a similar approach like MIPS but they don't seem to have had as much success (AFAIK raw performance and perf/watt is/was actually better for MIPS).

The second reason is of course, that while Intel may have very deep pockets, ATM they don't have full package a SOC needs in terms of baseband modems etc.

We shall see but Intel coming in cherrypicking the high end with Silvermont might harm Qualcomm et all but Intel's much vaunted process lead does cost Intel a fair bit too. Their cap-ex has been very high for the last few years and even with their revenues they cannot use all their capacity. Idling $billion fabs is never a good thing even for the Intel cash machine.

Usual morons who fail in so many ways to understand the market.

On the subject of MIPS, I wonder if we'll see continued development of their application processors? They would probably end up being produced by the likes of TI/Qualcomm so might not have a huge impact on the market/competition, but still. Also, it would be a fairly hard market to penetrate, especially with a different ISA.

MIPS currently don't really compete against ARM in terms of raw performance with their IP cores - the likes of proaptiv looked interesting but AFAIK haven't made it into production, and we know what on-paper specs can be like. I'm not so sure about efficiency, but in the mobile market, MIPS is only really found in a few cheaper devices so it's hard to gauge as they're older cores often built on older process nodes.

People seem to be very gullible with Intel marketing it seems.