Nehalem vs Q6600

[Sunin]

Is this upgrade path coming down the pipe going to be worth it?

Yes it splits a quad up into 8 threads, but it is still 4 cores... how much of a PPD do you think we will see from these?

I'm guessing only 20% or so... max..

 

[aldamon]

Is this upgrade path coming down the pipe going to be worth it?

Yes it splits a quad up into 8 threads, but it is still 4 cores... how much of a PPD do you think we will see from these?

I'm guessing only 20% or so... max..

I doubt Nehalem does much for folding unless memory bandwidth becomes a severe bottleneck.

 

[APOLLO]

It should be at least 20% better in performance than an equivalent Penryn core quad without taking HT into consideration. There was a recent thread with early Gainstown F@H benchmarks here: http://www.hardforum.com/showthread.php?t=1337038&highlight=nehalem

 

[Sunin]

This was ll before the current SMP correct? The new smp has some advancements in scaleability that should overall net the same results with just a single SMP running.

Or did I read the article incorrectly.

 

[APOLLO]

This was ll before the current SMP correct? The new smp has some advancements in scaleability that should overall net the same results with just a single SMP running..

TBH, I'm not sure, but scaling has definitely improved with i7 over all previous Intel microprocessor architectures, that much I do know. Scaling and especially memory subsystem performance are the two big advances with Nehalem. How this will translate to F@H is anyone's guess, but the early benchmarks do point in the general direction at where we're heading.

 

[Sunin]

I'm definitely more interested in the 16 core 32 virtual world. Where the process splits into 32 pieces and can be simultaneously computed and reassembled. I'd like to see one SMP running on that beast!

That is when I see the jump being truly worthy!

 

[APOLLO]

I'm definitely more interested in the 16 core 32 virtual world. Where the process splits into 32 pieces and can be simultaneously computed and reassembled. I'd like to see one SMP running on that beast!

That is when I see the jump being truly worthy!

Yes, I can see the appeal in a large scale multicore architecture. It will be quite a while before 16 logical cores are a available for the desktop, though. Even octal-core processors will likely be available only beyond 2009 at the earliest, if the roadmaps don't significantly change. To get that many cores sooner, you'll have to opt for a multi-socket solution and that's nearly always not worth purchasing strictly for F@H.

 

[Xilikon]

IMHO, let's way for some production numbers before seeing if it is worth the expense. Right now, it's still too early since the reports we saw is on early batches with early motherboards.

 

[BillR]

Every time we get a new processor the speculation begins. There is little doubt that the new batch will be faster than the last batch or they wouldn’t build it.

What no one thinks about is code optimization. Back in the P-4 days when AMD had the clear advantage Stanford decided to use a different Intel compiler, and Intel would not allow any compilation for AMD. Suddenly we had a whole crapload of units with better then a 4 to 1 edge biased toward the P-4 with HT.

The speed of processors will continue to expand exponentially that’s a fact.

Guessing, or betting when where and if that potential may or may not be utilized, especially for our purposes in my mind is a sucker bet until we learn more about how or if the advanced technology might be used.

 

[drgnfang]

At some point it be obvious, "Build a new machine using X". That will happen when the old hardware is no longer worth purchasing. Until then it is a crap shoot.

PS: Anyone know how to get the badge to work in Google Chrome?

 

[Mtnduey]

One thing to think about is the idea/fact that Nehalem can allocate far more power to a single thread than you can with any of the existing Quads today.

If this works as intended it could mean a single instance of F@[H] completing workloads much faster than any other Quad on the market.

We'll have to wait and see how it actually works out in the end.

Maybe there is some way when [H] reviews the Nehalem they could run some F@[H] workloads to see if the times are any better than previous Quads

 

[oraldlight]

One thing to think about is the idea/fact that Nehalem can allocate far more power to a single thread than you can with any of the existing Quads today.

If this works as intended it could mean a single instance of F@[H] completing workloads much faster than any other Quad on the market.

If that works, I'll definitely be questioning my AMD allegiance.

 

[APOLLO]

Maybe there is some way when [H] reviews the Nehalem they could run some F@[H] workloads to see if the times are any better than previous Quads

They already are even without optimizations. Look through the the thread I linked above. You'll see a post of mine where I make a rough comparison between the Gainstown system and a Harpertown system I have with only a 100MHz disparity. The Nehalem is much faster. Trust me the performance is there, but the value is not.

 

[BillR]

One thing to think about is the idea/fact that Nehalem can allocate far more power to a single thread than you can with any of the existing Quads today.

If this works as intended it could mean a single instance of F@[H] completing workloads much faster than any other Quad on the market.

We'll have to wait and see how it actually works out in the end.

Maybe there is some way when [H] reviews the Nehalem they could run some F@[H] workloads to see if the times are any better than previous Quads

That is going to depend on the code and compiler. Code compiled for HT isn’t nearly as fast on a non HT CPU

Under the right circumstances you could see a Q6600 become as worthless as tits on a daddy rabbit

 

[nomad8u]

Under the right circumstances you could see a Q6600 become as worthless as tits on a daddy rabbit

heh...heh... I think we just found us a new quote for the day F@H style!

Thanks Bill! I'll be grinnin' in my sleep on that one.

 

[BillR]

If that works, I'll definitely be questioning my AMD allegiance.

Well, once again let us speculate for a moment. Suppose in the next few months it becomes very apparent that folding on a standard CPU is just far too inefficient but some break through code for GPU folding makes the choice of hardware almost a non choice.

You aquire a few boards that hold lets say 4 PCIE slots and you are getting 10k a day per board per card. That would be cool yes? The issue becomes you don’t need anything but the most simple Quad core CPUs available. In that case a low end Q whatever or AMD Quad is all you will need to sustain those high out puts and a Nelham core becomes over kill, at least for folding. No need to over clock our main CPU anymore.

For once your old hardware has value again and naturally you have less to purchase on the basic side of things.

Just some additional thoughts.

 

[drgnfang]

Both of my dedicated GPU boxen are running dual core CPU's @ stock speeds (E6600, E2180). It was one of the main reasons why I was able to get these up so fast, all I had to do was buy the GPU's. Every other piece of hardware in both boxen are "old". Thankfully I had a couple of PSU's that could handle the workload. Even when I build my tri-98gx2 I'll just pull the Q6600 from my shuttle and use it @ stock.

 

[BillR]

Both of my dedicated GPU boxen are running dual core CPU's @ stock speeds (E6600, E2180). It was one of the main reasons why I was able to get these up so fast, all I had to do was buy the GPU's. Every other piece of hardware in both boxen are "old". Thankfully I had a couple of PSU's that could handle the workload. Even when I build my tri-98gx2 I'll just pull the Q6600 from my shuttle and use it @ stock.

Yup, for the price of getting them up...........so fast you saved enough to buy more viagra huh?

 

[aldamon]

Well, once again let us speculate for a moment. Suppose in the next few months it becomes very apparent that folding on a standard CPU is just far too inefficient but some break through code for GPU folding makes the choice of hardware almost a non choice.

We don't need to wait a few months. I'm pretty sure we know that already. That's why I don't understand the interest in Nehalem. You could get 10-15 video cards for the price of a Nehalem setup.

 

[jws2346]

IMHO, let's way for some production numbers before seeing if it is worth the expense. Right now, it's still too early since the reports we saw is on early batches with early motherboards.

Yeah, I've got to agree with Xilikon on the waiting before we make any decisions. Decisions on whether it'd be worth the expense to upgrade (up to date mobos, probably DDR3, etc) If it is worth it, even if I have to sell my sole and hock my cat (only kiddin' ), and these new CPU's are in fact more efficient for folding there's a good chance I'll give them a try.

At the moment everything I've read is merely speculation and I can't pay my rent on speculation

FOLD ON!

 

[Kendrak]

Unless we get some sort of uber WU that needs 8 threads and the points to go with it I can't see this as effective on a cost basis. A 25-50 precent improvement over Q6600 but a 600 percent increese in cost (just for the CPU, dont forget the RAM and mobo will be more expencive as well.)
Unless you a CAD or photoshop nut I think these chips will be a non event for atleast a year and untill we get under $400 for a HT Quad.

 

[Xilikon]

There are 399$ Bloomfield I think. However, there are rumors those CPU are truly overclocking locked so that might make it uninteresting. Add the cost of a new motherboard and 3 sticks of DDR3 and we are talkiong about 1000$ at least to upgrade for just 500-1000 ppd more (assuming the overclocking lock is bogus). 10x8800GS can bring 45,000 PPD

The only time it would be interesting if you are long overdue for a upgrade for your main PC. In that case, it's wiser to jump now on the new bandwagon instead of investing in a dead-end technology.

 

[nomad8u]

even if I have to sell my sole and hock my cat (only kiddin' ),

Dibs on the cat... I'm running out of chickens...

(jk JWS..jk)

 

[jws2346]

Hey my man (nomad8u), you almost got me evicted (or almost made my life miserable with her whinin'all the time, cats got no sense of humor ) It's a good thing the cat can't read . She just lays in front of the monitor and I have to peak around her to "surf" the Net and post stuff.

My lady friend suggested I get a small radio to keep her company when I'm not home. I guess that's cool, but what kinda' music do cats like? She'd literally destroy a little radio if she didn't like the music and I ain't takin' no chances with the "boob tube" I'd leave the window open, but it's a long drop from the 7th floor to the ground.

FOLD ON!

 

[Xilikon]

Hey my man (nomad8u), you almost got me evicted (or almost made my life miserable with her whinin'all the time, cats got no sense of humor ) It's a good thing the cat can't read . She just lays in front of the monitor and I have to peak around her to "surf" the Net and post stuff.

My lady friend suggested I get a small radio to keep her company when I'm not home. I guess that's cool, but what kinda' music do cats like? She'd literally destroy a little radio if she didn't like the music and I ain't takin' no chances with the "boob tube" I'd leave the window open, but it's a long drop from the 7th floor to the ground.

FOLD ON!

If only your cat can browse for lolcatz pics

 

[aldamon]

I guess that's cool, but what kinda' music do cats like?

Pussycat Dolls
John Cougar Mellencamp
The White Stripes

 

[jws2346]

Pussycat Dolls
John Cougar Mellencamp
The White Stripes

Hey that's cool brother aldamon. What I'll do is use some of my precious "duct" tape, tape ear buds to her ears and plug it into one of those iPod things and she should be copacetic (how long will the iPod deal play )

FOLD ON!