AMD CPU speculation... and expert conjecture

[juanrga]

i've been thinking about good enough...
the first thing popped into my mind was an hsa enabled arm apu with gcn 2.0/3.0 shaders made by amd on a problem-free process using the usual techs like rcm, hdl, fd-soi (as far as my imagination allows ) etc. with proper power monitoring and controlling logic and better power gating.
a little more thinking started to show these gaps in the fantasy.
amd doesn't have a working mainstream arm core. yet. but doesn't have it nonetheless. jaguar, otoh, seems perfectly suited despite being x86 arch. so if i imagine the same thing with arm replaced by a jaguar or jaguar-successor core, looks more realistic and less gap-y.

AMD already announced ARM APUs

http://www.theinquirer.net/inquirer/news/2224133/amd-will-bring-out-arm-opteron-apus-before-x86-version

kaveri exists. in silicon form. and it runs, games and stuff. haven't seen anything official/solid about specs yet. huma, hsa, gcn igpu are pretty much confirmed.
why won't amd just come out and say kaveri has steamroller cores? :S they didn't hide anything about kabini. kaveri's on track for 2013 launch, not sure if it's paper launch or real launch.

I think you missed this

This is a zoom

 

[sarinaide]

....
With haswell ... its designed for low power apparently (see idle power) so at high power, you get an inefficient design. This is intel's way of giving the finger to desktop computers and even more so to overclockers.

that's your beef with haswell? :lol:
wait till you see haswell u-series specs. no pcie 3.0 support. no discreet gfx support. <- this may trickle down to desktop broadwell or skylake.
the nuc/brix pcs use u-series cpus iirc.
bleak times ahead.
edit: at this point, i think sandy bridge was the last desktop-oriented(!) cpu lineup from intel.

That is devastating news, basically confirming Intel's "we give up" stance not that long ago.

On other news, couple people buying Haswells from me reporting terrible temperatures even at stock and with stock coolers on non k chips, I think Haswell is now appropriately termed a failure and as many reviewing retail chips have found, that haswell is slower than Ivy in many applications, Intels very own faildozer.

 

[gamerk316]

....
With haswell ... its designed for low power apparently (see idle power) so at high power, you get an inefficient design. This is intel's way of giving the finger to desktop computers and even more so to overclockers.

that's your beef with haswell? :lol:
wait till you see haswell u-series specs. no pcie 3.0 support. no discreet gfx support. <- this may trickle down to desktop broadwell or skylake.
the nuc/brix pcs use u-series cpus iirc.
bleak times ahead.
edit: at this point, i think sandy bridge was the last desktop-oriented(!) cpu lineup from intel.

That is devastating news, basically confirming Intel's "we give up" stance not that long ago.

On other news, couple people buying Haswells from me reporting terrible temperatures even at stock and with stock coolers on non k chips, I think Haswell is now appropriately termed a failure and as many reviewing retail chips have found, that haswell is slower than Ivy in many applications, Intels very own faildozer.

That would be true, if Intel were focused on the Desktop. They aren't. Haswell does EXACTLY what Intel wants it to do: Be optimised for lower form-factors.

 

[sarinaide]

....
With haswell ... its designed for low power apparently (see idle power) so at high power, you get an inefficient design. This is intel's way of giving the finger to desktop computers and even more so to overclockers.

that's your beef with haswell? :lol:
wait till you see haswell u-series specs. no pcie 3.0 support. no discreet gfx support. <- this may trickle down to desktop broadwell or skylake.
the nuc/brix pcs use u-series cpus iirc.
bleak times ahead.
edit: at this point, i think sandy bridge was the last desktop-oriented(!) cpu lineup from intel.

That is devastating news, basically confirming Intel's "we give up" stance not that long ago.

On other news, couple people buying Haswells from me reporting terrible temperatures even at stock and with stock coolers on non k chips, I think Haswell is now appropriately termed a failure and as many reviewing retail chips have found, that haswell is slower than Ivy in many applications, Intels very own faildozer.

That would be true, if Intel were focused on the Desktop. They aren't. Haswell does EXACTLY what Intel wants it to do: Be optimised for lower form-factors.

That excuse hasbeen D) used before, there is nowhere that stipulates expressly the focus on mobile and handhelds only, there may have been less focus on desktop but to say it was slapped together with a attitude of oh well disdain is even a further departure from the truth. Haswell has been somewhat anticlimatic, cue Haswell refreshes next year just another way for intel for flog market silicon and populate the Z87 chipset before they move over to the 90's and probably a new socket.

 

[The Q6660 Inside]

....
With haswell ... its designed for low power apparently (see idle power) so at high power, you get an inefficient design. This is intel's way of giving the finger to desktop computers and even more so to overclockers.

that's your beef with haswell? :lol:
wait till you see haswell u-series specs. no pcie 3.0 support. no discreet gfx support. <- this may trickle down to desktop broadwell or skylake.
the nuc/brix pcs use u-series cpus iirc.
bleak times ahead.
edit: at this point, i think sandy bridge was the last desktop-oriented(!) cpu lineup from intel.

That is devastating news, basically confirming Intel's "we give up" stance not that long ago.

On other news, couple people buying Haswells from me reporting terrible temperatures even at stock and with stock coolers on non k chips, I think Haswell is now appropriately termed a failure and as many reviewing retail chips have found, that haswell is slower than Ivy in many applications, Intels very own faildozer.

That would be true, if Intel were focused on the Desktop. They aren't. Haswell does EXACTLY what Intel wants it to do: Be optimised for lower form-factors.

That excuse hasbeen D) used before, there is nowhere that stipulates expressly the focus on mobile and handhelds only, there may have been less focus on desktop but to say it was slapped together with a attitude of oh well disdain is even a further departure from the truth. Haswell has been somewhat anticlimatic, cue Haswell refreshes next year just another way for intel for flog market silicon and populate the Z87 chipset before they move over to the 90's and probably a new socket.

HasBurst™ (Desktop) reminds me of -
http://upload.wikimedia.org/wikipedia/de/0/0d/Intel_pentium4_logo_original.jpg

 

[Cazalan]

I trust Nvidia on this one. Apart from those from Apple how many notebooks will use Haswell GT3e? Gigabyte presented a new line of Hasfail gaming notebooks at Computex and all them use a discrete GPU

http://www.pcmag.com/article2/0,2817,2419859,00.asp

AT is saying the 1 chip Haswell (2 core) solutions don't have PCIe. That's going to put a damper on mobile discrete sales.

SA said this a while ago too. About how Intel is going to kill AMD/NVidia by limiting PCIe. Lower end mobile PCs/tablets will just have USB/HDMI/Thunderbolt.

 

[de5_Roy]

AT is saying the 1 chip Haswell (2 core) solutions don't have PCIe. That's going to put a damper on mobile discrete sales.

SA said this a while ago too. About how Intel is going to kill AMD/NVidia by limiting PCIe. Lower end mobile PCs/tablets will just have USB/HDMI/Thunderbolt.

my speculation: in the future we might see 3 distinct types-
intel with 'near-proprietary' x86 cpu, gpu, soc,
nvidia with arm soc, gpu,
amd with both x86 and arm soc(unless intel does 'something'), apu, gpus.
intel and amd have both started with their bga platforms.
nvidia is the only one behind (i.e. without a 'cpu'), might eventually go that way when intel shuts them out of x86.
however, i think high end desktop(from both amd and intel) might retain lga cpus and discreet gfx. entry level and midrange gfx might take a large hit while upper-mid to high end might survive. amd will likely see no reason to offer cheap prices once intel fully closes it's door to mainstream d.i.y. that could still be 2 years away...
edit: if pcie dudes come up with a power efficient pcie standard, intel's excuse would be invalidated.

 

[-Fran-]

I read Anandtech's Haswell ULT rushed-review with the new Acer S7 and the numbers are mixed for Notebook users.

For average Joe, it's an amazing improvement in Battery life, but (according to comments) Haswell based Ultrabooks are heating up more than their IB siblings. Looks like moving so much hardware into a tiny place was good and bad call at the same time; go figure! haha

And for us, "enthusiasts", who would use our lappies attached to a wall socket, there's really nothing to see or rejoice over. The "IPC" is basically the same when plugged in and it shouldn't be different in the non-ULV parts running against IB parts.

Man, after reading that, I'm really eager to see Kaveri leaks of some sort. Any Chinese site photoshopping slides or something? haha.

Cheers!

 

[noob2222]

Spoiler

....
With haswell ... its designed for low power apparently (see idle power) so at high power, you get an inefficient design. This is intel's way of giving the finger to desktop computers and even more so to overclockers.

that's your beef with haswell? :lol:
wait till you see haswell u-series specs. no pcie 3.0 support. no discreet gfx support. <- this may trickle down to desktop broadwell or skylake.
the nuc/brix pcs use u-series cpus iirc.
bleak times ahead.
edit: at this point, i think sandy bridge was the last desktop-oriented(!) cpu lineup from intel.

That is devastating news, basically confirming Intel's "we give up" stance not that long ago.

On other news, couple people buying Haswells from me reporting terrible temperatures even at stock and with stock coolers on non k chips, I think Haswell is now appropriately termed a failure and as many reviewing retail chips have found, that haswell is slower than Ivy in many applications, Intels very own faildozer.

That would be true, if Intel were focused on the Desktop. They aren't. Haswell does EXACTLY what Intel wants it to do: Be optimised for lower form-factors.

That excuse hasbeen D) used before, there is nowhere that stipulates expressly the focus on mobile and handhelds only, there may have been less focus on desktop but to say it was slapped together with a attitude of oh well disdain is even a further departure from the truth. Haswell has been somewhat anticlimatic, cue Haswell refreshes next year just another way for intel for flog market silicon and populate the Z87 chipset before they move over to the 90's and probably a new socket.

probably a no socket bga broadwell. There have only been rumors that broadwell might have an LGA variant. The only thing Intel said is that there will be a new cpu in 2014 for 1150, low and behold, our haswell refresh cpus.

The speculation that Intel meant broadwell was just put into orbit of the unknown. What is known is that Intel wants to say farewell to desktop so they can market their ever-srhinking tech.

 

[sarinaide]

I trust Nvidia on this one. Apart from those from Apple how many notebooks will use Haswell GT3e? Gigabyte presented a new line of Hasfail gaming notebooks at Computex and all them use a discrete GPU

http://www.pcmag.com/article2/0,2817,2419859,00.asp

AT is saying the 1 chip Haswell (2 core) solutions don't have PCIe. That's going to put a damper on mobile discrete sales.

SA said this a while ago too. About how Intel is going to kill AMD/NVidia by limiting PCIe. Lower end mobile PCs/tablets will just have USB/HDMI/Thunderbolt.

Now that is anti-competitive.

 

[noob2222]

....
With haswell ... its designed for low power apparently (see idle power) so at high power, you get an inefficient design. This is intel's way of giving the finger to desktop computers and even more so to overclockers.

that's your beef with haswell? :lol:
wait till you see haswell u-series specs. no pcie 3.0 support. no discreet gfx support. <- this may trickle down to desktop broadwell or skylake.
the nuc/brix pcs use u-series cpus iirc.
bleak times ahead.
edit: at this point, i think sandy bridge was the last desktop-oriented(!) cpu lineup from intel.

That would be quite comical to see Intel turn IGP only about the time that 4k takes off. Good luck gaming on that with an IGP.

Imagine the opportunity for AMD to become the only "desktop cpu" manufacturer. Unfortunately, I don't see it happening. Intel would never leave AMD an opening like that to make a profit.

 

[sarinaide]

Spoiler

....
With haswell ... its designed for low power apparently (see idle power) so at high power, you get an inefficient design. This is intel's way of giving the finger to desktop computers and even more so to overclockers.

that's your beef with haswell? :lol:
wait till you see haswell u-series specs. no pcie 3.0 support. no discreet gfx support. <- this may trickle down to desktop broadwell or skylake.
the nuc/brix pcs use u-series cpus iirc.
bleak times ahead.
edit: at this point, i think sandy bridge was the last desktop-oriented(!) cpu lineup from intel.

That is devastating news, basically confirming Intel's "we give up" stance not that long ago.

On other news, couple people buying Haswells from me reporting terrible temperatures even at stock and with stock coolers on non k chips, I think Haswell is now appropriately termed a failure and as many reviewing retail chips have found, that haswell is slower than Ivy in many applications, Intels very own faildozer.

That would be true, if Intel were focused on the Desktop. They aren't. Haswell does EXACTLY what Intel wants it to do: Be optimised for lower form-factors.

That excuse hasbeen D) used before, there is nowhere that stipulates expressly the focus on mobile and handhelds only, there may have been less focus on desktop but to say it was slapped together with a attitude of oh well disdain is even a further departure from the truth. Haswell has been somewhat anticlimatic, cue Haswell refreshes next year just another way for intel for flog market silicon and populate the Z87 chipset before they move over to the 90's and probably a new socket.

probably a no socket bga broadwell. There have only been rumors that broadwell might have an LGA variant. The only thing Intel said is that there will be a new cpu in 2014 for 1150, low and behold, our haswell refresh cpus.

The speculation that Intel meant broadwell was just put into orbit of the unknown. What is known is that Intel wants to say farewell to desktop so they can market their ever-srhinking tech.

They will leave DT then be running back with their tail between their legs as the gaming and enthusiast market is still big.

 

[Cazalan]

Here is part of the problem with hasbeen.

its one chip, one design for low power. look at the vrm efficiency. If you design it for low power, it sucks at high power. If its designed for high power, its not quite as good at low power. MB designers could hand-design their boards for the end-users purpose.

This chart i think was from ~2006. Today the efficency peaks ~95%. If you hand desing your entire system for the load that will be drawn (easier and commonly done on servers), you can hand-pick your vrm module such that its always or almost always at 95%. The curve difference is still the same. if designed for low power, it peaks really fast and drops off fairly quick. If its designed for high power, it peaks slower, but doesn't drop off nearly as fast (wich is why its essential for overclocking to have 6 or more phase vrm)

With haswell ... its designed for low power apparently (see idle power) so at high power, you get an inefficient design. This is intel's way of giving the finger to desktop computers and even more so to overclockers.

Those are ancient 4-6 phase VRMs. The one in Haswell is 320 phase, or 16 phase per power core x20.

Motherboards are using 16-32 phase VRMs which are good but still not as high quality as in Haswell.

AT is showing up to 100% better battery life on the Haswell ultrabook. That is significant for 1 generation.

 

[Cazalan]

That would be quite comical to see Intel turn IGP only about the time that 4k takes off. Good luck gaming on that with an IGP.

4k gaming wont take off anytime soon. The panels are too expensive.

 

[8350rocks]

Here is part of the problem with hasbeen.

its one chip, one design for low power. look at the vrm efficiency. If you design it for low power, it sucks at high power. If its designed for high power, its not quite as good at low power. MB designers could hand-design their boards for the end-users purpose.

This chart i think was from ~2006. Today the efficency peaks ~95%. If you hand desing your entire system for the load that will be drawn (easier and commonly done on servers), you can hand-pick your vrm module such that its always or almost always at 95%. The curve difference is still the same. if designed for low power, it peaks really fast and drops off fairly quick. If its designed for high power, it peaks slower, but doesn't drop off nearly as fast (wich is why its essential for overclocking to have 6 or more phase vrm)

With haswell ... its designed for low power apparently (see idle power) so at high power, you get an inefficient design. This is intel's way of giving the finger to desktop computers and even more so to overclockers.

Those are ancient 4-6 phase VRMs. The one in Haswell is 320 phase, or 16 phase per power core x20.

Motherboards are using 16-32 phase VRMs which are good but still not as high quality as in Haswell.

AT is showing up to 100% better battery life on the Haswell ultrabook at idle. That is significant for 1 generation.

Fixed! They always quote the longest battery life, which is at idle, when they compare battery power consumption.

Additionally...Acer added 31% more Wh to the battery of the S7 in that review...in case you guys missed that, I will quote it here:

http://www.anandtech.com/show/7047/the-haswell-ultrabook-review-core-i74500u-tested/4

Acer dramatically increased the battery capacity of the Aspire S7 from 35Wh up to 46Wh. Despite the higher capacity battery there’s no increase in weight. The 13.3-inch S7 still weighs around 2.86 lbs.

The increases in hasfail for ultrabooks remain around 16-40% additional battery life on average, with edge cases seeing close to 50%.

Meaning intel got a good boost from the additional C-states that power down at idle.

 

[hcl123]

Their issue is that they don't do so well in poorly-threaded games and benchmarks, which to add insult to injury, are often compiled with Intel's ICC as well

Do not make me start analyzing compiler signatures again; didn't I debunk the "Things use ICC" myth already?

I know of exactly ONE benchmark that uses ICC, everything else uses MSVC, because it is the best development studio on gods green earth, and nothing else comes close.

Nevertheless independent of compiler all pieces of software can be tweaked better this way than that, even the compiler can be tweaked more this way than another if access to source... and the bench software can change a lot(not only bug hunt) yet maintain exactly the same denomination

So excuse if i ask, but where is that debunk analyses ?

... i never saw one, SPEC is above the contend because it publishes the compiler used (VERY WISE)... yet it misses to deliver the md5 of all those pieces...

The only think that can debunk everything my friend, are md5 signatures for the blobs uses, either the source code is available or not( of course if source is available, then all the exact flags for a determined compiler must be knowned... meaning you can have the same source and compiler, yet use different flags, then blobs will carry the same exact name yet are quite different.

UPDATE:

See the difficulty ?

You can have everything on the open, code compiler, yet vary the flags on this last one to distort the results a lot, some flags can be favorable to some uarch features others to others, results can vary from almost nothing to 10 to 15% or more, what with some version of that code and compiler can win, with a diffrent version derived from different flags it can lose...

Yet what we are measuring is hardware not software, so the only thing essential is "guaranties" the blob (software) is the same independent of machine tested, and so md5 is what is essential, it tells origin which is always nice and most important it tells if there is differences from possible other blobs circulating around...

If i were an IDM controlling this paradigma, i would always tweak the bench software instead of the hardware... no md5 none the wiser... and its a loooot less expensive lol...

 

[noob2222]

Here is part of the problem with hasbeen.

its one chip, one design for low power. look at the vrm efficiency. If you design it for low power, it sucks at high power. If its designed for high power, its not quite as good at low power. MB designers could hand-design their boards for the end-users purpose.

This chart i think was from ~2006. Today the efficency peaks ~95%. If you hand desing your entire system for the load that will be drawn (easier and commonly done on servers), you can hand-pick your vrm module such that its always or almost always at 95%. The curve difference is still the same. if designed for low power, it peaks really fast and drops off fairly quick. If its designed for high power, it peaks slower, but doesn't drop off nearly as fast (wich is why its essential for overclocking to have 6 or more phase vrm)

With haswell ... its designed for low power apparently (see idle power) so at high power, you get an inefficient design. This is intel's way of giving the finger to desktop computers and even more so to overclockers.

Those are ancient 4-6 phase VRMs. The one in Haswell is 320 phase, or 16 phase per power core x20.

Motherboards are using 16-32 phase VRMs which are good but still not as high quality as in Haswell.

AT is showing up to 100% better battery life on the Haswell ultrabook. That is significant for 1 generation.

Electronics 101: To put it simply as possible, the smaller the component, the less power it can handle. A single 22nm transistor isn't going to be able to handle 84W of power draw through it. So why does HW have 320 "phases"? At any given "phase" you have to run it through multiple parallel transistors to reduce the amperage draw across each individual one.

Most motherboards z77, am3+, ect are a 8-phase setup, some of those not even a true 8-phase but rather a 4+4 phase running parallel. gigabyte's z77x up7 has a 32-phase vrm, but not sure how its setup (8+8+8+8 or 16+16, or true 32). Either way, they are all doing the same thing, splitting the power to increase the overall available power. Sure, HW has 320 individual ones, but how many are running parallel? That answer gives you the true (?) phase design. Is it a 4-phase design running 80 vrm parallel or 8-phase running 40 parallel, ect ..? also, how many are dedicated to the igp, removing x# for the cpu. ie, gigabytes up7 is 32 (cpu) +3 (igp) +2 (vtt).

as I showed before, the more true phases you use, the less efficient at low power.

As far as HW battery life ... lol ... 100% ...

Acer dramatically increased the battery capacity of the Aspire S7 from 35Wh up to 46Wh.
...
The new S7 delivers over 2x the battery life of the old model. Normalizing for battery capacity, the improvement due to Haswell is 57.5%

so for web browsing, its 57.5% better.

 

[Cazalan]

so for web browsing, its 57.5% better.

Which is about as demanding as the general user gets.

That's a huge savings for the same process node.

 

[Cazalan]

Most motherboards z77, am3+, ect are a 8-phase setup, some of those not even a true 8-phase but rather a 4+4 phase running parallel. gigabyte's z77x up7 has a 32-phase vrm, but not sure how its setup (8+8+8+8 or 16+16, or true 32). Either way, they are all doing the same thing, splitting the power to increase the overall available power. Sure, HW has 320 individual ones, but how many are running parallel? That answer gives you the true (?) phase design. Is it a 4-phase design running 80 vrm parallel or 8-phase running 40 parallel, ect ..? also, how many are dedicated to the igp, removing x# for the cpu. ie, gigabytes up7 is 32 (cpu) +3 (igp) +2 (vtt).

Intel specified 20 power "cells" with 16 phase each. Individually programmable so it can be tweaked for different power envelopes.

There have no doubt been some growing pains with this architecture (extra heat) but it's rather key going down to smaller nodes. They'll likely go to more than 20 cells and when integrated in the die be spread out more to get even closer to the loads.

 

[8350rocks]

As they slowly creep toward a BGA desktop setup across the board except for LGA 2011 and it's descendants.

 

[hcl123]

That would be quite comical to see Intel turn IGP only about the time that 4k takes off. Good luck gaming on that with an IGP.

Imagine the opportunity for AMD to become the only "desktop cpu" manufacturer. Unfortunately, I don't see it happening. Intel would never leave AMD an opening like that to make a profit.

Don't sing a requiem yet, intel still has the IB-E and by 2H 14 the HSW-E. I think it will be more appropriate for the "enthusiast" sector (the OCers), and it will be gran if they endorse with it better multithreading software, after all we are talking 12 threads and with HSW-E probably 16.

Only intel have to lower the price of this pieces a good deal to have success, and on top of that for competition, and if its confirmed 28nm PD-SOI, is AMD launching G34 with RD1000 chipsets for this markets... it could have the same 4 channel DDR, it could have the same number of threads... and it could even have better clocks.

Always good for end users. Nevertheless for this i believe intel process will be tweaked quite differently of this iGPs.

 

[hcl123]

as I showed before, the more true phases you use, the less efficient at low power.

As far as HW battery life ... lol ... 100% ...

Acer dramatically increased the battery capacity of the Aspire S7 from 35Wh up to 46Wh.
...
The new S7 delivers over 2x the battery life of the old model. Normalizing for battery capacity, the improvement due to Haswell is 57.5%

so for web browsing, its 57.5% better.

Battery or circuits ?

there is possible now a system quite worst than intel at low power yet have quite an improved battery capacity, and so have the same life span over charges( the reverse is also possible)

Yet none of it tells anything about performance, where those 57% comes from ?

 

[Cazalan]

Yet none of it tells anything about performance, where those 57% comes from ?

UltraBook Review

http://www.anandtech.com/show/7047/the-haswell-ultrabook-review-core-i74500u-tested/3

And it should improve in Win 8.1 due to Connected Standby implementation.

"Although you should be able to realize some of the benefits from S0ix with the first Haswell Ultrabooks, Connected Standby (periodic content refresh while in a sleep state) requires OS support. For Haswell Ultrabooks, that means waiting for Windows 8.1."

 

[Cazalan]

The speculation that Intel meant broadwell was just put into orbit of the unknown. What is known is that Intel wants to say farewell to desktop so they can market their ever-srhinking tech.

Is it Intel or the market saying farewell to desktop?

The market seems perfectly happy buying up expensive tablets with 0 upgradability. Apple was made the richest company in the world briefly because of the iPad, and they didn't even offer a microSD slot.

I think the tablet phase will eventually die and be replaced with the "2 in 1" or convertibles. Which both Intel and AMD are poised to get into.

 

[Cazalan]

As they slowly creep toward a BGA desktop setup across the board except for LGA 2011 and it's descendants.

The traditional desktop will likely disappear to be replaced as a workstation class CPU. A single server processor like the E3. They're actually not that expensive anymore.