Ahh, my thread was moved ... wish mods would let people know when they do that since the link no longer works. Plus, why wait until the 4th page after 4 months to move it at all?

Anyway, I am wondering if AMD's Piledriver would be worth waiting for (if my computer from 2004 can hold out that long - unlikely). How does the Piledriver compare to Intel's Ivy Bridge? I am assuming that Piledriver would've fixed all those issues between Bulldozer and Windows 7, RIGHT???

When threads that you started or are subscribed to are moved it sends an email out to the email account that you registered with (see "profile" at the top right of your screen).

You won't see any performance information on Piledriver until it is released. That is AMD's business model (most of the time). The issues with Bulldozer's scheduler and Win 7 are still in beta phase so we will not know.

-------------------------
The opinions expressed above do not represent those of Advanced Micro Devices or any of their affiliates.

Interesting thread and this in my opinion is going to make or break AMD in the next year if they dont sort it.

Here is an interesting story for you, i can understand the concerns why many want PCIe 3.0 support and its not really the performance update.

In 2001 high end workstation were breaking over to PCIe leaving AGP behind as well as moving to 64bit computing. The high end market was embeding superfast AGP slots that had not yet been saturated in fact they were not saturated until about 2004. Why does this matter? well because those that had purchased a workstation at costs 3X the standard top PC gaming system ended up royaly screwed. The chipsets throttled you to upgrading to 64bit and the PCIe generation crushed out all future options for AGP. Yes we seen the gaming market change there boards like they change there underpants but when that mainboard, CPU(S) cost as much as they do its not that simple. Your pretty much forced to chuck it and replace in fact expandible options in the HPC workstation server market is a massive problem as it is not cheap tat that can be changed like your underpants.

With the above in mind on buying a mainboard processor solution you need to think 4 years from now what you are going to need, remeber were talking $5000.00++ investment here that you might have to throw away after 2 years urm NOT an option.

You spend that kind of cash you really need a future proof solution so its not actually about performance now its about getting that extra bit of poke for minimal upgrade to your investment. By scrapping and buying new it is clearly a bad move. If AMD dont move fast or change something soon in the HPC market people are not going to concider AMD solutions to there investments for the next few years.

I not going to blat lots of replies to different people so ill put it all here. Someone had mentioned changing the socket and having PCIe 3.0 on the CPU die. Yes intel have done this, there maybe a issue for AMD to do this its called Patents lol, Intel might have the IP on it. At present if you take a look at the mainboard plans for C32/G34 sockets you will see all PCIe is handled by the southbridge SR5xxx chipset. HT3 is not 2 x X16 its actually x16 + x8 take a look at the Tyan mainboard documentation to see this. PCIe gen 2.0 X16 is a 5GT/s lane where 3.0 ups the speed to 8GT/s PCIe gen 4.0 due 2015 is PCIe gen 3.0 in paralel making X32 lane @ 8gt/s so if AMD dont move fast on this there in trouble with PCIe 4.0 just 3 years away you can bet your bottom doller intels E7 chips are going to be all over it in 2 years. I am expecting Intel to be all over PCIe 3.0 for there E7 DP/MP before the end of this year. There is a market lead AMD can take here with there opterons 4000/6000 and FX series.

AMD memory controller is already set for delivering 2ghz ddr3 memory. There HT interconects need to be overclocked as mentioned by some posts in this thread and that is mostly i say mostly down to mainboard manufacturers. The PCB tracking has to be made and tuned to handle these speeds. Check out PCIsig for the specs they hawe to adhere to. Crosstalk, inductance are a big issue when dealing with these speeds a slight mistake and it can render the entire board useless. Remeber you cant just overclock you need to balance the PCB interconects correctly to get it working. Intel have managed it to a certain degree of success, proves it can be done.

To those who thing software does not use it, well that is in your Windows world and this is why the Linux server dominates the planet at present. It allows developers/administrators to really open it up, thats the meaning of free software after all. Thats why we need PCIe 3.0 this year so AMD needs get busy squeezing there chips to blow the world away with the best HPC or they might just loose the race.

Thanks for reading and i hope AMD read this too because after all its there customers that buy which in turn pay there wages. Sorry for spelling errors if any peace and fingers crossed we get what we need so we can invest and keep our inovation going. I say that because compiling allot of source code takes a loooooong time and openCL really needs a faster PCIe slot.

-------------------------
I see a rainbow of colours between black and white.