Nehalem comments here

ilkhan

[H]F Junkie
Joined
Sep 23, 2002
Messages
13,791
Can we get all of the nehalem comments into one thread? Watching 5 conversations is getting annoying. So this is the unofficial nehalem commenting thread. (and remember, you can quote from other threads)

personally, I think the price concerns are stupid. DDR3 is NOT as expensive as people are saying. Boards...we'll see. The chips will fit into the current price structure, $300, $500, $1k.

performance, we know is damn good.

overclocking...we'll see.

Best resources:
http://en.wikipedia.org/wiki/Nehalem_(microarchitecture) (GREAT chart)
http://www.nehalemnews.com/

********************************
Its official- nvidia will support SLI on X58.
http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3395
Intel > nVidia, and they finally realized it. (Actually they realized that 4870 x2 is powerful enough to be a more than viable option for every person buying bloomfields for the next 6 months, and without SLI they would do so.
*********************************
 
adding to above
Overclockers.com's Ed takes a stab at what a potential Nehalem platform may cost:
http://overclockers.com/tips01362/

And that overclockers.com link is horrible. Newegg has a 1GB DDR-3 133 stick for $65 (3=195, already under their high price.) 6GB (3x2GB of DDR3-1600) is already under $375, $100 under their low price. Now, these are cheapest sticks, but the capacity is already there. I doubt that bloomfield wont have memory dividors availablel, if worse come to worst and the chips are far more overclockable than we imagine.

The people that are setting records cant bitch about price, those who want the best performance for their dollar aren't going to be buying the $800 memory kits.

Also remember that bloomfields will be running multiplyers in the 20-24x range, more than double what we are facing with yorkfields. Meaning that you need a LOT less QPI increase than we do today for the same CPU clocks.
 
DDR3 price concerns are valid to some people. Personally, I bought a 939 rig when DDR was pretty damn cheap and DDR2 was still absurdly expensive. I moved to DDR2 just recently, because it's now absurdly cheap compared to DDR3. I don't think anyone is arguing that DDR3 is any more expensive than DDR2 when it first came out. If you want to buy into DDR3 in its relative infancy, you'll have to pay the price - just like if you wanted to buy into DDR2 in its infancy.

High end boards have been getting more and more expensive for a while now, so I expect a few months later Gigabyte or someone else will come along to save us with a relatively inexpensive board. Remember, the northbridge no longer needs to include a memory controller, so this will make at least the midrange boards cheaper. One good thing about the platform is that we'll finally have SLi on Intel chipsets!

The Nehalem launch will likely be just like the Conroe launch - theoretically low prices, but high demand leading to scarce availibility for the first few weeks. How soon will the more affordable CPUs be readily availible? Who knows.

I predict that for the first few months building a Nehalem machine will carry a not at all insignificant premium over a C2D/C2Q PC. Prices of the latter will likely fall to make them an even more attractive proposition, but the performance superiorty of the former will still make it very attractive.

Now, is there any truth to these rumors that you won't be able to overclock Nehalem?
 
Everything Ive read says that LGA1366 (the bloomfields we are talking about) will be overclockable, but the LGA1160 chips (lynnfield, havendale) will be locked.

The other question is timing. If they ARE launched in september as is now rumored, then even 2-3 months of constrained supply allow supply to bloom out for the original launch date / x-mas / etc.
 
Those are the two big questions though now arent they guys. The OC issue and the motherboard prices at launch.

Personally I'm not buying into dr. dolittle "the sky is falling" non-sense with the motherboard prices. I'm sure we'll see more than enough 100-150 dollar boards at launch as well as the e-peen boards in the 250-300 dollar region.

Jury is still out and will probably stay out until the chips actually release with retail motherboards for the OC question.
 
The problem is I doubt we WILL see $150 boards at launch. X48 isnt that low, and X58 wont be either. On the other hand, the die size decrease of X58 chip should decrease cost by a decent amount (10-15?). Id expect the cheapest launch boards to be $200 or $250, but not below. Even today the cheapest X48 board on newegg is $189 after MIR, or $220 for the cheapest asus, which is still a DDR2 board.
 
If rumors are true and release is in late September I'm hoping that by November the initial shock will wear off and availability/pricing will be good. I want to get my folding heater started so I can leave the furnace off for an extra week or 2 this fall.
 
The big problem of Nehalem:

There is no P55.

I'll wager that even the non-SLIed X58s aren't cheaper than the most expensive P45s (discounting the Asus BS ROG series)
 
I am seriously excited about this release. I think that people will be extremely impressed with the Nehalem. Intel is gonna be the shiznit going forward. With Larrabee coming as well this should be a really cool cycle.
 
OP,

Your wikipedia link is somewhat broken (it's incomplete).

Thanks.
 
That pricing structure has been known for about a month (take a look at the date on the page you posted). It should also be noted that those prices are for batches of 1000 chips. The retail price for a low end Nehalem CPU will be slightly higher than $284.
 
That pricing structure has been known for about a month (take a look at the date on the page you posted). It should also be noted that those prices are for batches of 1000 chips. The retail price for a low end Nehalem CPU will be slightly higher than $284.

Regarding the price, I think they will be a lot of price gouging on release. These will be the hottest chips in town and demand will probably outstrip supply initially. I reckon we might see prices above $350 in the first couple of weeks, its generally how things pan out, anyone remember the 45nm C2D shortages? The E8400 was selling for $250 when its official price was $183.
 
The MaxPC article posted in a seperate thread reminded me of something else important, current dual channel 2x2048MB memory chips WILL work just fine in dual channel mode. And I suspect the performance loss will be minor at best with higher speed memory vs slower tri-channel.
 
The MaxPC article posted in a seperate thread reminded me of something else important, current dual channel 2x2048MB memory chips WILL work just fine in dual channel mode. And I suspect the performance loss will be minor at best with higher speed memory vs slower tri-channel.

Indeed. The bandwith available from tri-channel DDR3 is immense, and I doubt even Nehalem will be able to use up all that bandwith effectively. What it does do is allow for future scalability with 8 core / 16 thread CPUs and so on.
 
Indeed. The bandwith available from tri-channel DDR3 is immense, and I doubt even Nehalem will be able to use up all that bandwith effectively. What it does do is allow for future scalability with 8 core / 16 thread CPUs and so on.
True. Im certain that the review sites will be doing comparative testing with 2x2GB and 3x1GB/3x2GB memory configs once they drop the NDAs.

Tri-channel DDR3-1600 will be rated at almost 40GB/s of bandwidth.
Dual channel DDR2-1066 is rated at 17GB/s. Slight difference there. ;)
 
DDR3 price concerns are valid to some people. Personally, I bought a 939 rig when DDR was pretty damn cheap and DDR2 was still absurdly expensive. I moved to DDR2 just recently, because it's now absurdly cheap compared to DDR3. I don't think anyone is arguing that DDR3 is any more expensive than DDR2 when it first came out. If you want to buy into DDR3 in its relative infancy, you'll have to pay the price - just like if you wanted to buy into DDR2 in its infancy.

High end boards have been getting more and more expensive for a while now, so I expect a few months later Gigabyte or someone else will come along to save us with a relatively inexpensive board. Remember, the northbridge no longer needs to include a memory controller, so this will make at least the midrange boards cheaper. One good thing about the platform is that we'll finally have SLi on Intel chipsets!

The Nehalem launch will likely be just like the Conroe launch - theoretically low prices, but high demand leading to scarce availibility for the first few weeks. How soon will the more affordable CPUs be readily availible? Who knows.

I predict that for the first few months building a Nehalem machine will carry a not at all insignificant premium over a C2D/C2Q PC. Prices of the latter will likely fall to make them an even more attractive proposition, but the performance superiorty of the former will still make it very attractive.

Now, is there any truth to these rumors that you won't be able to overclock Nehalem?

For the good DDR3 it is very expensive about $300 for 4GB so my 8GB Comp would be $600 just for the memory. I did the same and built a DDR PC even though DDR2 was out. This PC is still faster than my computer at work (Dell) with DDr2 and C2D instead of my opteron. This is because my DDR PC is using PC4000 and the work PC is using DDR2 PC5400 with crappy timings. Which is why at this time I would not hesitate to still buy DDR2. Besides look how the prices of DDR went up while DDR2 went down. Hopefully the samme will happen with DDR3, and even buying DDR3, we all know the good DDR3 will be out in a couple years time, not right now at which point the prices on everything else will also have dropped.

Look at AM2, it promised alot, but I never once regretted building my DDR PC except for the fact that it would have been expensive to upgrade it to 4Gb due to the older memory rocketing in price!

I would like to see some benchmarks for DDR3 1600 C7 vs DDR2 1000 C5 vs DDR 500 C3 or DDR400 C2

I think there would be some difference but I'll bet the difference isn't as much as most people think. The difference between DDR and SDR is likely much more significant. What we really need to to stop tacking numbers on the end and come out with QDR memory...

IMO as usual, if you want to build a super high end PC it's worth waiting forr, if your on a budget (less than 2k) it's probably not. Don't forget as soon as nehalem is released something better will be in the works. As far as sitting on the fence, no one is going to be able to say if it was worth the wait or not until it is released as we see some benchmarks. It's always a good idea to wait PC's don't go up in value...
 
IMO as usual, if you want to build a super high end PC it's worth waiting forr, if your on a budget (less than 2k) it's probably not. Don't forget as soon as nehalem is released something better will be in the works. As far as sitting on the fence, no one is going to be able to say if it was worth the wait or not until it is released as we see some benchmarks. It's always a good idea to wait PC's don't go up in value...
We already know whats after nehalem, westmere (32nm shrink) and then sandy bridge (new arch). Tick tock, tick tock.
 
Desktop Nehalem parts are going to be pointed directly at the highend market this year and probably Q1 of next.
 
:D :D :D Can't wait for a review!
At the rate they're going we'll see it in 2020.

AMD's first 45nm parts are now being talked about as 2009 parts, while intel will be pushing 32nm westmere parts out to market just 9 months later. Thats gonna hurt AMD.

Kyle- yup. I dont really understand why the onboard GPU mobile parts are going to take so long, though. Itd seem to me that pushing those cores as fast as possible would get system power consumption down the most in the area that needs it most.

At the moment I really wish AMD was as competitive in CPUs are the 48x0 radeon line is turning out to be against nvidia.
 
Pulled together the block diagrams that Ive seen.
So what exactly is the point of X58? The only thing it does is branch more PCI-E lanes than the PCH can do on its own and translate the QPI to intels standard DMI...thoughts?
 
Back
Top