Bulldozer Chips will most likely be a drop-in for AM3 boards

[ragejg]

Well, FWIW AMD's socket AM series has stayed mostly pin compatible and has seen consumers through quite a few feature upgrades as well... from an HT frequency of 2000 to 3200 mhz, from dual core to Octa-core compatibility, from DDR2-667 to today's fastest DDR3 compatibility.... along with some other noteworthy features I can't remember at the moment. I'm not arguing for argument's sake here, but it is the facts... AMD created a socket design that was flexible enough to grow with the technological leaps.

I think it's pretty neat right now that you could buy an AM2+ motherboard and either slap a single-core Sempron and some DDR2-667 in it, OR an x6 1090t and some fast DDR2-1200 (or maybe faster?) ... and with there being a good possibility that you may be able to do the same with a Bulldozer x8, it really makes the range of the socket quite large.

I'm all for new tech being pushed through the channels via platform upgrades, but as a constantly bang-for-the-buck hardware enthusiast, AMD's route makes more sense for me and for hardware buyers like me.

[Ninja Prime]

You have no clue what you're talking about slawter.

[Maverick123w]

Originally Posted by Bearclaw

This. While I like that sockets don't change simply for easier upgrade, matter of the fact is - new sockets bring new tech.

New chipsets bring new tech :P

Same socket has seen usb 3 and sata 6

Originally Posted by Ninja Prime

You have no clue what you're talking about slawter.

/agree

We all know you like intel Slawter it's OK.

[Maverick123w]

Originally Posted by slaWter

Whatever. All I'm saying is that current Intel CPUs wouldn't be possible in the old 775 socket. CPU aren't just CPUs anymore. They include parts of the northbridge now as well, which reduces latency.

AMD didn't design a future proof socket back then. All they did were minor changes to the CPU, stuff that doesn't change the interface. AMD had to change sockets in the server world too in order to provide a better memory interface, which is good.

I'm saying that new sockets are a good thing, not a bad one. This is not AMD vs Intel or some fanboy talk.

But Maverick123w is right, chipsets also bring new tech to the end user. So a mainboard change is necessary anyway. Why not time that good and introduce a new socket at the same time?

BTW I want to make it clear that when agreeing with ninja prime I'm just playing around. In all serious I know you know what you're talking about

I hope intel continues to do well. It allows me to buy the AMD equivalent that provides 80-90% of the performance for 70% of the cost.

[Lyme]

Originally Posted by slaWter

For example, AMD talked about integrated GPUs for years. They don't have them yet. Intel does.

Now only if Intel in it's divine greatness could come up with a gpu, integrated or not, that wasn't utter Sh1t

[Viral]

Originally Posted by slaWter

Whatever. All I'm saying is that current Intel CPUs wouldn't be possible in the old 775 socket. CPU aren't just CPUs anymore. They include parts of the northbridge now as well, which reduces latency.

The same way Athlon 64's weren't possible in the old Socket A. Integrating the memory controller will do that...

Just look at AMDs highest end CPU offering. It still uses the old base socket. Now compare it to current Intel CPUs, Intel is faster and has more features.

Faster and has more features has nothing to do with the socket in this case since they've used the same sockets since they first intergrated the memory controller.

For example, AMD talked about integrated GPUs for years. They don't have them yet. Intel does.
Intel is not afraid to change the socket if necessary, if it helps bring new tech to the market.

How's integrating a GPU a good example of why changing sockets is a good thing for moving technology forward when intel didn't change the socket when they integrated the GPU? That makes no sense and only serves to disprove your argument.

Just look at your list of features there. Nothing special, just improvements of their tech over the years. Faster bus speeds, more cores, faster memory.
...
New tech is like integrated GPU, integrated PCIe controller, new bus standards and so on. There is no way around another socket because you're changing the interface to the outside. But it brings great benefits to certain market segments and end users.

It's a faster bus speed because it's a new standard. So by your definition, it is new tech. You have crazy double standards here. When Intel change their bus it's a new standard and that's okay, when AMD do it, it's just a speed bump? Wake up.

As for faster memory, they moved to DDR3 and have made no changes since. Of course the socket is going to have to change with the move to DDR3, that's a different thing to just "faster memory".

AMD didn't design a future proof socket back then. All they did were minor changes to the CPU, stuff that doesn't change the interface. AMD had to change sockets in the server world too in order to provide a better memory interface, which is good.

There's much, much more behind the shift to G34 than the memory interface. You're oversimplifying things.

How exactly is moving from HT2.0(AM2) to HT3.0(Am3) "not changing the interface? How exactly does moving to DDR3 "not change the interface". You keep saying that they only made minor changes, I don't think you actually know anything about the changes.

Here,


Can we stop the double standards now? The interfaces changed, thus a new socket was needed. Same thing Intel do.

I'm saying that new sockets are a good thing, not a bad one. This is not AMD vs Intel or some fanboy talk.

And I agree, but how is it suddenly not AMD vs. Intel when you've used the subject to point out AMD socket changes = pointless, while Intel socket changes = necessary.

If you like it or not, that's personal taste.

Or... it's technology - personal taste has nothing to do with it unless someone has made it personal for some strange reason. I think we all know what that makes them...

[Viral]

Originally Posted by ragejg

It's still gonna be the same basic socket design, slawter. That's why AM3 CPUs will work in AM2+ and AM2 boards, so it's likely that AM3+ will also work in AM2 (with a BIOS upgrade).

It won't.

Posted by JF-AMD
AM3: AM3 processors - yes; Bulldozer no
AM3+: AM3 processors - yes; Bulldozer yes

There is no way that BD will work in an AM3, there is nothing to "shut off"

K10 wasn't a radical departure from K8, just an evolution. Bulldozer is a complete new design from the ground up. This is the first time AMD have done that since K7.

AMD have decided to cut ties with any socket limitations that may hold back their new architecture. Most likely the main feature that is the reason behind this is per module power gating.

Slawter is right in this case, AMD have done this to enable new features on their new design. Of course, that doesn't make his anti-AMD spin on previous socket history any more justified.

[Viral]

Originally Posted by slaWter

AMD introduced the IMC years ago. A Socket change was necessary.

Intel finally did the same with Nehalem. They also moved the PCIe controller to the CPU and introduced QPI and so on. New socket right there.

AMD also introduced Hyper Transport, Integrated the NB, introduced features like CoolnQuiet and eventually introduced dual core chips. All of which required the new socket.

Changes like those move the products forward, those are new features.

Indeed, integrating the memory controller, NB, adding Hyper Transport, power saving features and 64-bit capability were great features. AMD did it all in 2003 and you give them no credit besides the memory controller. Intel do the same stuff years later and all of a sudden they're great changes that move products forward? Yes, I'm afraid you do appear to have double standards.

I'm also confused regarding what you define as a feature, which is why I italicised the word above. How exactly is QPI/HTT, integrated memory controller, adding another memory channel and moving the PCIe controller to the CPU a feature for the end user? These things effect performance, the user doesn't gain anything besides that so I don't see how you can call them features and as such they are lumped in with improvements in bus speeds and memory support. Sure, they may add more performance, but that's all they are for the end user, performance enhancers.

Improvements like faster bus speeds or memory support is nice but not as important.
Intel did that too, like faster FSB speeds while keeping the socket.

You can keep the same socket pretty easily when you're still using an FSB and have the NB on the motherboard... back then, changing chipsets was all the rage for Intel. Everyone critisized AMD when they went from S754 and S940 to S939 then later to AM2 etc. Intel have done the same thing... bam.. three desktop sockets already with two more to come.

Unfortunately AMD did this for years, that's why their CPUs are kinda stuck. Hopefully that'll change soon.

So according to you AMD did nothing but adding support for faster memory and HT speeds and that's why they're "stuck"? That's flat out wrong.

Like I pointed out, AMD did everything they needed to do in 2003 to stay ahead of intel as a platform for 5 years. They also did a good job of keeping these changes updated(CnQ, HTT, memory controller). All they then needed was to change/update is their chip architecture, this is where they have fallen down, their platform has remained excellent. Granted, no longer having this as an advantage will take its toll.

The only things they were a bit slow on were power management features, turbo and integrating the GPU... I can't see how any of those features would effect you at all? Granted though, they are likely to matter to other users.

AMD have changed sockets often enough, maintaining backwards compatibility with the last 2 sockets has in no way held them back. I can't get my head around your idea that where they are going wrong is that they have stuck to the same socket. Sure, now they are changing sockets because it makes sense to do so. Doing so with K10 wouldn't have fixed anything. Can you try to explain how changing sockets this year, say with the Phenom x6 launch, could have helped AMD?

[Viral]

Fair enough. To me what AMD have done right in the last 3 years is that they've been able to introduce new sockets that both brought about improvements AND allowed for backwards and forwards compatibility (although with disabled features when a newer proc was used in an old socket). My argument is that besides power gating (being able to completely shut off cores) and adding a GPU earlier, AMD havent missed out on anything by sticking with the sockets they have. The GPU thing is arguable... I don't think they felt they were ready regardless. Ontario required a new design, and Llano has been waiting for 32nm to be ready. I think it is moreso Intels process advantage coming into play here.

Clearly they can only do maintain compatibility for so long. Bulldozer, being a new arch, seems like the perfect time to make the change. Same deal with Llano and Ontario.

Personally, I just hope they launch the AM3+ platform sometime before Bulldozer launches. For me, neither company offers a platform that interest me with a reasonable life span.

[jcrox]

Originally Posted by Ninja Prime

You have no clue what you're talking about slawter.

This

[noko]

Now when will the AM3+ sockets be available? Why are they not called AM4 socket if they are radically enhanced or different? For backwards compatibility of AM3 processors? May just have to hold off my upgrade plans for awhile now as in 6-12months.