PCI Express 3.0

[NeoGen]

I just saw that a few months ago MSI launched an Intel motherboard (reportedly World's first?) that implements the brand new PCI Express 3.0, through the Intel Z68 chipset, which doubles the bandwidth of the current PCI Express implementations.

Does anyone know when will AMD put out a chipset to support it too?

Anyway... trolling (err.. read "searching", trolling is a bad term I hear) around the internet I found rumours that the next generation of AMD GPUs (HD 7000?) will be "PCI-E 3.0 Compliant"... I wonder if that means that they will be just standard PCI-E 2.x running on a 3.0 slot or will they really make use of the full potential (and bandwidth) of PCI-E 3.0?

Let the discussion begin! :icon_mrgreen:

[Brucifer]

You know, over the last few years the economy has tanked. The standard old game of keep increasing the requirements has slowed way down... People just aren't purchasing all this rassmatazz anymore. Why? Cause they flat don't have the bucks. The old game just isn't sustainable in this present economy. The companies are doing good just to hold even with what they have. Now they want to change the game again, and start making more stuff obsolete. Fine.... if they think that the out of work folks, or the folks that have not received pay increases for a few years are just going to rush out the door and buy all this stuff, they have a larger hole in their head than I thought. I'll keep crunching what I have but I'm not going to run out and re-do a small farm. It's easier to just find a new hobby, or just stick with projects that don't require playing the constant upgrade game. I'm just not impressed with the PCI-E 3.0 game. The present one works fine.

[Dirk Broer]

While I agree with most of what Brucifer says -did receive just one Pay increase since 2000, and presently being unemployed myself-, there is something to say against keeping on crunching with the same old hardware till it falls over. In last month's (Dutch) PCM magazine there was a test comparing generation P4 systems and assorted hardware with a new setup. Conclusion was that a new system will always perform much better, and can do so at a significantly lower cost per year, energy-wise. When I look at the efficiency of the hand-me-down P4 machines I have running, the socket 478 machines are fit to be junked and replaced with something new. New mobos come from around $40, DDR3-1333 memory is $45 for 8Gb, and the only thing left to choose is what CPU do you want with your new system. Tomorrow the launch of Bulldozer will prove whether it was wise to postpone that choice.

[AMDave]

Does anyone know when will AMD put out a chipset to support it too?
AMD scheduled Terramar (20 core) and Sepang (10 core) cpus with PCIe3 for sometime in 2012 so I would anticipate their PCIe3 chipsets to be rolling in late 2011.
There was some hype in 2008 that AMD would bring out PCIe3 with the SB800 controller but they appeared to have concerns and rolled out an accelerated PCIe2 instead.
In 2008 and 2009 HP and AMD contributed changes to add extensions to the specs for PCIe3 for finalisation in 2010.
At that time HP's Michael Krause said they expected the tech to hit the market in 2012 - 13 (http://www.eetimes.com/electronics-news/4084075/AMD-HP-propose-extensions-for-PCI-Express-3-0).
Among the extensions that were forwarded included features that would allow the CPU to offload tasks to GPGPUs where available and another feature to help the increasing load of control functions for managing processing across many cores.
Now you know why the development of Fusion has been so compelling to AMD.
Expect some interesting things to come in 2012.

[NeoGen]

I am pretty sure that right now, for 99.9% of the population, PCIe 3.0 will make no difference from what they have now, because they don't even fully use the bandwidth and potential of PCIe 2.0

Now for the remaining 0.1% which includes mostly hardcore gamers and crunchers like ourselves I am curious to know what the impact will be... I've never investigated if the PCI Express bus was a bottleneck for the current GPUs when doing heavy crunching, (maybe depending on the project?) but if it is, then imagine if you would potentially double your credits by making a motherboard switch. That would be cool. :)

[Brucifer]

As your opening sentence said neo, they don't even use the potential of 2.0. And even with the optimized coding, etc., the present cards still don't. So we are talking a complete rehash of the world here, new chips, new m/b, new graphics cards, etc. lol

As crunchers, we aren't really hitting the walls with this stuff. Look at the commercial world with the super high powered graphics and scientific computing requirements, and of course too, look at the prices they pay for that stuff. Way above and beyond what the high volume everyday consumer would pay for the stuff.

I think with a lot of this hardware stuff it is purely pursued to try and maximize the ability to incrementaly withdraw money from people's wallets. It isn't directed towards the perfection of the technical abilities of the product. Look at the fundamental difference between windows and the UNIX world............. Windows brings out a new version, and with it a whole new batch of problems that goes through the process of being fixed, and about the time they get it working worth a hoot, then bingo it's time for a new version and we go through all the stuff again. What is it directed at? MS's incoming cash flow is the primary and real world goal. In the UNIX world, the process is to fix a problem which normally doesn't entail a whole new release of the product that revamps the game as with windows. Linux got bastardized a bit in this process as some of the commercial software houses started trying to play the M$ business model, which as we have seen hasn't worked out well in the linux/unix world.