Nehalem....but how does it fold?

[HighYield]

Here is some linkage to wet your appetite for futuristic folding power....

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3326

If you could build just one more computer this year would you wait for Nehalem?

 

[digital_exhaust]

My only issue would be cost.....I think.

Any idea how much these things are going to run?

Evil and I were talking about these the other day, and the HT bit does look promising, but you have to wonder how it will work with F@H, time will tell I suppose....

 

[Catboxer]

This is my problem. I'd love to do an overhaul of my current rig which a Q6600, P35 chipset, ATI 3870, 4GB PC6400. That system would go into my MAME/F@H cabinet and I'd build a new workstation.

What I'm not willing to do is wait for the end of Q4 only to pay 500-900 just for the CPU. I thought about sticking with the current setup and getting a C2Duo for the MAME cabinet, but that's going to run 350-400 dollars right there and wouldn't that money be better off put to use upgrading and then trickling systems down? Tough call for me.

 

[BassTek]

Well you'll need the chip, a new motherboard since it's a new socket and new DDR3 memory. I probably won't get them until a while after launch, I can't justify it as a pure folding box so I'll just wait until my main rig needs an upgrade.

Most likely my next folding related purchase will be a new Nvidia (possibly ATI) video card for my main rig and then I can shift the 8800 GTS to pure folding duties in another system.

 

[EvilAlchemist]

I suspect it will be just like any other upgrade.

Give it a while and the price will come down.
There will always be something new coming out .. that is faster and better.

The C2Q 9XXX series is pretty good.
I like my Q6600's but the lower power draw & heat make the news one look better.

Only time will tell if the new CPU are better at F@H ..

 

[APOLLO]

There have been a few benchmarks leaked recently such as the link by the OP that indicated some amazing performance possibilities for Nehalem, but how can we estimate F@H production upon benchmarks alone, and which benchmarks would best correlate with tentative F@H performance?

What we know thus far is a processor based loosely on the Core architecture that has a far superior memory subsystem with IMC, superior cache scheme including an efficient large L3 and Hyper Threading. Eight logical processors will definitely not harm F@H production, and all that from a single chip! That would translate to 16 logical processors on dual sockets...

More to the point, scaling should improve markedly with the IMC and overall architectural improvements. Unless I'm reading the previews on hardware sites completely wrong, I foresee only vast improvements over what we have from any manufacturer today.

 

[Xilikon]

Yes, the thing we are 100% sure is that it won't be worse than Penryn. With the IMC and tweaks, we can foresee up to 20% folding performance improvement vs a Penryn. The only question is how much it will cost and if there is a locked overclocking possibility (with only the Bloomfield line being unlocked, possibly in the QX form and the 4 digits price tag).

 

[SmokeRngs]

It will probably be closer to a year before I go with Nehalem. At this point it will mainly be due to cost more than anything. New Intel platforms are always expensive and the requirement of still very pricey DDR3 will not help at all. I'm thinking of building another system before the end of the year but I'm not going to wait on Nehalem because I don't think it would be worth the price. I'll probably just throw together another Q6600 (or Q9xxx if the price and multiplier are right) system and not worry about things. Then again, this is also the guy that still has P3s and AthlonXPs up and running 24/7.

While performance with Nehalem may be really good, I don't think it's going to be worth the money for a while. Also keep in mind the problems that plague practically every new platform and architecture change like this. The Core2 line was by no means bug free when it was introduced and it was based off of the Core line. The processors were mostly mature but the motherboards were crap and every expensive for a while. Until the problems with the motherboards (and supply of them) were worked out, the Core2 platform wasn't exactly the best thing to go with. It will more than likely be the same and possibly worse with Nehalem as everything but DDR3 will be new. Even then, it's still somewhat new as it will be Intel's first production IMC and it may not play very well with DDR3 at first.

 

[Celerator]

Nehalem looks promising with a nice boost in benchmarks and apps.
But, wouldn't it be nice to see a FAH benchmark?

I'm imagining a folding benchmark program that would cycle through
one step of various WU types to estimate folding power. It would be great
advertising for Stanford. At the end of the benchmark, it might ask,
"Would you like to install FAH on your system and join the folding
community?"

Maybe the [H]orde should do it...

 

[Parja]

...and the requirement of still very pricey DDR3 will not help at all.

Well, you can bet that with Intel basically locking themselves into DDR3 with Nehalem, thus guaranteeing the memory manufacturers a crapton of DDR3 sales, the prices should drop pretty quickly.

 

[SmokeRngs]

Well, you can bet that with Intel basically locking themselves into DDR3 with Nehalem, thus guaranteeing the memory manufacturers a crapton of DDR3 sales, the prices should drop pretty quickly.

Well, I thinking about what happened with DDR2. It was actually quite cheap prior to the Core2 line being released. Not long after that, the prices for it went sky high and stayed that way for a while. There were reasons other than the Core2 line being released for the prices to jump up like that, but it was a factor. I'm thinking we'll probably see something similar in this case especially since DDR3 prices aren't exactly low (at least the last time I looked which was a while ago, I admit).

 

[Sunin]

Here is some linkage to wet your appetite for futuristic folding power....

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3326

If you could build just one more computer this year would you wait for Nehalem?

I would not wait.. in the cost of a single nehalem system I think you will be able to build 4-5 Q6600 machines and you would end up with more power overall. It will be a while before we see 16 cores...

If they price it for 2k for a 16 core Nehalem system with everything it might be interesting, but I highly doubt it will even be that cheap!

 

[APOLLO]

I would not wait.. in the cost of a single nehalem system I think you will be able to build 4-5 Q6600 machines and you would end up with more power overall. It will be a while before we see 16 cores...

Do you mean on one or two sockets? Logical processors or physical cores?

If they price it for 2k for a 16 core Nehalem system with everything it might be interesting, but I highly doubt it will even be that cheap!

A 16-core Nehalem system would be two Beckton revision parts, which are octal-core processors with HT (16 threads each). However, I read that Beckton won't be released this year at all and that it's slated for the MP space (4+sockets) which means megabucks (~$10K+ servers).

Unless AMD releases an octal-core processor before Intel, it will be over a year before we see 8 true core processors and probably close to two years before we see them on the desktop. Roadmaps can change drastically, though.

 

[Sunin]

Do you mean on one or two sockets? Logical processors or physical cores?

A 16-core Nehalem system would be two Beckton revision parts, which are octal-core processors with HT (16 threads each). However, I read that Beckton won't be released this year at all and that it's slated for the MP space (4+sockets) which means megabucks (~$10K+ servers).

Unless AMD releases an octal-core processor before Intel, it will be over a year before we see 8 true core processors and probably close to two years before we see them on the desktop. Roadmaps can change drastically, though.

Well I assume single socket... a single 4 core 8 with HT will only do about the equiv of a Q6600 and a half... but at probably a price tag of 4-5 Q6600 machines...

I agree it will be probably a year and a half before we see 16 physical core versions of Nehalem even if they hit under this architecture it might not hit until the next one in 2yrs... I would love to see a 32nm 32 core chip that is basically faster than the 9 quads I have right now... /drool! and take 1/2 the energy!

mmmm power!!!