updatebitsandchipsAM4 Socket will be µOPGA and it will have 1331 pins

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

cbn

Lifer
Mar 27, 2009
12,968
221
106
AIO liquid cooler, judging by AMD's past experience.

That does make a lot of sense.

P.S. Since AIO does not cool the VRMs, maybe AMD also makes some kind of FX spec level or quasi FX spec level for motherboards as well?

See example of the following Asrock AM3+ board where top down blowing cooler is specified for the 125W CPUs (but not the 95W CPUs):

http://asrock.com/mb/AMD/970M Pro3/?cat=CPU
 
Last edited:

cbn

Lifer
Mar 27, 2009
12,968
221
106
Or maybe AMD plans to sell 140W AM4 chips with this type of packaging (ie, no cooler):

19-113-347-TS


But then still specify some kind of quasi FX type motherboard? (with VRMs appropriate for non top down blowing coolers such as towers (eg, Coolermaster Hyper 212 EVO or Raijintek Ereboss) and AIO liquid coolers)
 
Last edited:

BigDaveX

Senior member
Jun 12, 2014
440
216
116
Any confirmation that claim Zen will be fabbed at Glo Fo?

They're definitely producing something on 14nm for AMD, though no solid evidence yet as to whether it's Zen or a GPU of some description. Samsung's name has also been thrown around, but again, no firm confirmation so far.
 

mysticjbyrd

Golden Member
Oct 6, 2015
1,363
3
0
I thought the new wraith cooler was confirmed to be with zen? Why would they even make it, if it wasn't?
 

mysticjbyrd

Golden Member
Oct 6, 2015
1,363
3
0
I see going after the person is still easier than the ball when you lack arguments.
Since when was "herp derp bulldozer sucks amd bad" considered an argument? That's an unnecessary statement, that you repeat in every thread.

Since you didn't understand the cash flow part. Lets try again.

If Intel doesn't innovate, increase performance metrics, doesn't give any new incentive to upgrade and so on.

Who on earth will ever upgrade their CPU? They wouldn't even last a year before entering chapter 11.

It seems you think they sell tap water with a static demand. Hence dont understand why competition doesn't matter.

Oh, and we had what now, 10 years without competition? It have went quite well so far.

What? All they would have to do is make small improvements. Jack the price through the roof. Increase the rate of failure by skimping on materials, reduce the lifespan, while also dropping the warranty, etc...

Heck, intel's enthusiast series is the decent example of some of these things already. They have no competition in that market. The 5960x is easily overpriced by $400! It's an old architecture, and all they did was solder the ihs, and remove the igpu, which takes up about half the die space, and replaced it with more cores. If that cost them an extra $700, then my name is jack frost.
 
Last edited:

cbn

Lifer
Mar 27, 2009
12,968
221
106
Also another thing to consider is the lag time certain aftermarket heatsink companies will probably have developing AM4 compatible backplates. (This assuming AMD does not develop its own 140W cooler)

So if 140W does exist there are several reasons we might not see it during the initial launch.
 
Last edited:

ElFenix

Elite Member
Super Moderator
Mar 20, 2000
102,414
8,356
126
Also another thing to consider is the lag time certain aftermarket heatsink companies will probably have developing AM4 compatible backplates. (This assuming AMD does not develop its own 140W cooler)

So if 140W does exist there are several reasons we might not see it during the initial launch.
I'm going to assume the mechanical specs for the socket will be/are available in plenty of time for heatsink makers to have products. Otherwise OEMs won't be able to sell systems on day 1.
 
Aug 11, 2008
10,451
642
126
Well, if one thinks a poster is biased or posts useless information, they are certainly free to use the ignore button instead of making personal attacks.
 

JDG1980

Golden Member
Jul 18, 2013
1,663
570
136
Yes, and it was ridiculous to think they would release Bulldozer with worse IPC in a lot of workloads that Phenom, but they did it. And that was a "clean sheet" design. What I am saying is that it is really not that easy, and bad designs make it to market.

No one doubts Bulldozer was an epic fail. From what I can tell, it was fundamentally flawed both at a conceptual level (CMT, emphasizing speed at the cost of IPC) and in execution (poor cache performance, inefficient and badly laid-out design due to reliance on automated tools). But AMD knows that they can't afford another CPU flop, and they have learned some lessons from BD, just as Intel learned some lessons from Netburst in the design of Conroe and its successors. No, it's not "easy" by any means, but they had one of the best designers in the business (Jim Keller) and they're shooting for a goal that is known to be attainable, with a far more mainstream and traditional design than Bulldozer was. Is it possible they could still lay a turd? Yes, anything is possible. But I'm optimistic that Zen will deliver to reasonable expectations (which I would define as Sandy Bridge or greater IPC, with competitive perf/watt to modern Intel designs and competitive clock speeds based on the usage case and number of cores).
 

JDG1980

Golden Member
Jul 18, 2013
1,663
570
136
With what, 1/10th the budget on a 3rd party node?

In the past (Netburst era), AMD has flat-out beaten Intel in performance, despite a lower R&D budget. Despite notable missteps, they seem to be pretty good at getting the most out of their R&D dollar. No, they won't beat Intel outright this time (their last victory was assisted by Intel's own stupidity in pushing Netburst and Itanic, a mistake they won't make again), but I do expect them to return to a level where they can actually compete in the high-end laptop/AIO, HEDT and server markets, rather than just bottom-feed as they are doing now with the construction cores.

And let's not forget that they are trying to match an IPC target that has only barely budged since 2011. It's easier to catch up to a competitor than to blaze new ground. AMD also had the advantage of being able to pick and choose the best aspects of three different ground-up architectures (K10, cat cores, and construction cores) when designing Zen. They probably knew quite a bit from the start about what should be done and what should be avoided.

So you expect 40% IPC increase and 4Ghz+? :)

On the top quad-core bin, absolutely. If they stick with a maximum 95W TDP as leaks indicate (was this ever put in an official document), then an 8-core CPU won't be going that high at stock. But it should still overclock to that level just fine. Intel's 22nm process is inferior to Samsung/GloFo 14LPP in every metric, but 8-core Haswell-E clocks to 4.0-4.5 GHz just fine if you're willing to tolerate 140-160W power consumption. And that chip has a beefier AVX unit than Zen (at least if Dresdenboy's inferences are accurate), and we know AVX is one of the most power-hungry features of a modern CPU.
 

MiddleOfTheRoad

Golden Member
Aug 6, 2014
1,123
5
0
Quite an overstatement actually. I agree we need competition, but all the hysteria and hand-wringing about the lack of competition is crazy as well. Just look at the server market, Intel *is* essentially a monopoly there, yet advancements are made each generation, probably more than in the consumer market. Why? Because Intel needs to bring out an improved product to drive new sales and especially upgrades.

What? How do you figure? Intel has amazing marketshare among x86 servers -- but x86 is only a portion of the market. There are many other architectures like Power and ARM that are grabbing share too. It's a very competitive environment -- actually, servers have arguably more players fighting for sales than consumer CPU's.

Granted, x86 has about 60% of the market because of their popularity in data centers. But around $20 billion is still being spent on non-x86 servers. That's a pretty sizable figure going to Intel's competitors.
 

JDG1980

Golden Member
Jul 18, 2013
1,663
570
136
So what do you base this claim on that 14 nm LPP is "superior in every way" to intel 22nm?

http://www.extremetech.com/wp-content/uploads/2016/01/FeatureSizes1.png

While 14LPP isn't quite as dense as Intel's 14nm process, every feature size is smaller than Intel 22nm. Do you know of any metrics where Intel 22nm is superior? I've seen no credible reports of any.

Based on current information, I'd put Intel's advantage over Samsung/GloFo on cutting-edge processes at about a half-node. That's a much slimmer lead than we've seen in the past couple of years.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Also another thing to consider is the lag time certain aftermarket heatsink companies will probably have developing AM4 compatible backplates. (This assuming AMD does not develop its own 140W cooler)

So if 140W does exist there are several reasons we might not see it during the initial launch.

I'm going to assume the mechanical specs for the socket will be/are available in plenty of time for heatsink makers to have products. Otherwise OEMs won't be able to sell systems on day 1.

Maybe AMD would want two tiers of AM4 motherboards then (assuming 140W CPUs come):

1. 95W Spec with VRMs geared for top down blowing coolers.

2. 140W Spec with enhanced VRMs meant to not only handle increased power draw, but also the the low VRM airflow of AIO liquid coolers and classic tower coolers.

LATE EDIT: Also consider the Bristol Ridge APUs and CPUs (which share AM4 socket) are only 65W.
 
Last edited:

ElFenix

Elite Member
Super Moderator
Mar 20, 2000
102,414
8,356
126
While different power ratings would no doubt be the preference of OEMs (save some pennies!), IMHO it was horrible on AM3.
 
Aug 11, 2008
10,451
642
126
I am talking about real world performance at high frequencies. We have no proof that 14nm LPP will even clock to 4ghz. Just look at how far superior intel 14 nm is to 22 nm by all the metrics in that chart, but it brought very little to no gain at high frequencies.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
While different power ratings would no doubt be the preference of OEMs (save some pennies!)

Also I think 140W would be too much for Mini-ITX.

EDIT: Come to think of it there is LGA 2011-3 Mini-ITX. (There was also a AM3 Mini-ITX but I checked and it only supported 95W processors).
 
Last edited:

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
In the past (Netburst era), AMD has flat-out beaten Intel in performance, despite a lower R&D budget. Despite notable missteps, they seem to be pretty good at getting the most out of their R&D dollar. No, they won't beat Intel outright this time (their last victory was assisted by Intel's own stupidity in pushing Netburst and Itanic, a mistake they won't make again), but I do expect them to return to a level where they can actually compete in the high-end laptop/AIO, HEDT and server markets, rather than just bottom-feed as they are doing now with the construction cores.

And let's not forget that they are trying to match an IPC target that has only barely budged since 2011. It's easier to catch up to a competitor than to blaze new ground. AMD also had the advantage of being able to pick and choose the best aspects of three different ground-up architectures (K10, cat cores, and construction cores) when designing Zen. They probably knew quite a bit from the start about what should be done and what should be avoided.



On the top quad-core bin, absolutely. If they stick with a maximum 95W TDP as leaks indicate (was this ever put in an official document), then an 8-core CPU won't be going that high at stock. But it should still overclock to that level just fine. Intel's 22nm process is inferior to Samsung/GloFo 14LPP in every metric, but 8-core Haswell-E clocks to 4.0-4.5 GHz just fine if you're willing to tolerate 140-160W power consumption. And that chip has a beefier AVX unit than Zen (at least if Dresdenboy's inferences are accurate), and we know AVX is one of the most power-hungry features of a modern CPU.

So you say that Intel needs to make all the mistakes in the book, AMD make everything right. And even then they couldn't beat the Pentium-M. And back then their R&D was much higher than today and they had a fab that was in the game. Are you saying Skylake is a P4 type disaster?

If you already need to rely on overclocking its lost. You should know that.

Good luck with your hyperbole.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
http://www.extremetech.com/wp-content/uploads/2016/01/FeatureSizes1.png

While 14LPP isn't quite as dense as Intel's 14nm process, every feature size is smaller than Intel 22nm. Do you know of any metrics where Intel 22nm is superior? I've seen no credible reports of any.

Based on current information, I'd put Intel's advantage over Samsung/GloFo on cutting-edge processes at about a half-node. That's a much slimmer lead than we've seen in the past couple of years.

What 4Ghz MPUs have Samsung made on their 14lpp?
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Also I think 140W would be too much for Mini-ITX.

Ye, its not pretty. And 140W in a world where 65W is the norm now for non K chips. That's not going to sell well. People want smaller packages and low power. Only the "dinosaur" people accepts 140W or higher.
 

Dresdenboy

Golden Member
Jul 28, 2003
1,730
554
136
citavia.blog.de
Now that we know AM4 is 140W TDP capable, do we expect more than 95W TDP CPUs ??
I somehow missed the confirmation. Is there a link?

If voltages drop from node to node, and Bitlife slides listed 95W SR, are there enough pins left to provide the needed currents or is the package fed at higher voltages, which gets further stepped down for the cores?