Intel Nehalem?

[VulcanX]

I have reading a bit bout the flavours of the Nahalems, and the differing codenames etc, but which would you guys say is the best p/p considering they are releasing LGA 1366 of the Duo, Quads and even Octa, and with that said, which one would actually be able to be OCed without generating WAY too much heat, as i want my next pc to be an OC pc, and am willing to put some dosh into it, but im gonna wait it out for this gen, they seem to be best value for money
what your guys 2 cents on this architecture?

 

[boomhower]

There is really no way to even guess at this point, it is still to early. With there being talk of multiple sockets and the possibility of both multi and FSB locked chips we will just have to wait and see. Even though Nehalem is coming in the fall you must realize that it is just the extreme chips and it is going to be late winter before anything that us mere mortals will be able to afford will be out. So it is still way to early to have much concrete information.

 

[myocardia]

Actually, it's going to be sometime in 2009 before the mainstream chips are released, possibly even mid or late 2009, if we're to believe the latest rumors.

 

[VulcanX]

Hmmm so its nothing that anyone really knows about it, i would love to know more bout those bcoz i rather save for an octa core than spend all the time slowly upgrading and having to constantly buy stuff, so im waiting for the price of the 8800GTS to come down, and buy that, or wait even longer for the GT200 and save that lil extra in the long term

 

[Shortass]

Well if you want a Nehalem, as myo mentioned, you might be waiting until potentially Q3 or Q4 2009 to pick up a reasonably priced rig. The 8800gts will CERTAINLY be cheap then, but honestly if you're looking to upgrade in any reasonable amount of time (and you don't have $3000 to spend on an extremely overpriced rig) you might want to just go for it now or later this summer and wait for the prices to become decent in a year or so.

 

[bryanW1995]

intel is under no pressure from amd right now. we should consider ourselves lucky if we have sub-$1000 bloomfields within 18 mos at this point :|

 

[Nemesis 1]

Your question is interesting. I have given it some thought myself. With the 1366 socket its going tp be expensive . But thats not what I am struggling with. Its the very queation you posed. Buy 1366 at releases your paying big bucks for a 4 core that can handle 8 threads , you also get QPlink to gpu. The part that bothers me is this. IF intel releases a 2core nehalem that runs on the 1366 socket much latter . Is this possiably the biggest sleeper die of all time. It should O/C better than 4 core die. It still have 4 threads per die. So its future proof.

I probably spend the big bucks after all the chipsets are installed on everyones 1st gen. M/B and than buy the expensive 4 core . But the 2core if intel produces it should be great if it does SMT.and its available on socket 1366. THIS could be the sleeper!

 

[taltamir]

from what i hear, intel feels so little pressure that it is making an entirely new anti overclocking capability to integrate into all non "Extreme" version chips...

 

[Idontcare]

Originally posted by: myocardia
Actually, it's going to be sometime in 2009 before the mainstream chips are released, possibly even mid or late 2009, if we're to believe the latest rumors.

The lead at this point is so damning that I 100% believe Intel's merely hedging against a dark horse coming out with Shanghai when they roadmap in anything Nehalem mainstream on the desktop coming from 45nm.

As soon as Intel has confirmation (i.e. retail shanghai chips in hand) of whether or not Penryn will be sufficient to fend off Shanghai you can bet everything mainstream relating to Nehalem gets slipped to Westmere and 32nm timeline for cost management and yield reasons.

 

[Cheex]

Originally posted by: taltamir
from what i hear, intel feels so little pressure that it is making an entirely new anti overclocking capability to integrate into all non "Extreme" version chips...

:shocked:

Noooooooooo!!!....

Please Intel don't do that to us....:brokenheart:

 

[Rebel44]

Originally posted by: taltamir
from what i hear, intel feels so little pressure that it is making an entirely new anti overclocking capability to integrate into all non "Extreme" version chips...

Incorrect - http://www.nehalemnews.com/200...overclock-lock-on.html

 

[Idontcare]

Originally posted by: Rebel44

Originally posted by: taltamir
from what i hear, intel feels so little pressure that it is making an entirely new anti overclocking capability to integrate into all non "Extreme" version chips...

Incorrect - http://www.nehalemnews.com/200...overclock-lock-on.html

Citing speculation posted in a blog does not invalidate taltamir's hearsay.

 

[VirtualLarry]

I thought that the whole reason that Intel had not yet come up with an anti-overclocking mechanism, was that it simply wasn't cost-effective. IE. putting this additional feature in would cost more, overall, than just simply allowing a minute percentage of chips to be overclocked.

Do you think that this will no longer be true? That there are so many people overclocking in the C2D era, that Intel is loosing money from it and can afford to add anti-overclocking features?

 

[dmens]

Originally posted by: VirtualLarry
I thought that the whole reason that Intel had not yet come up with an anti-overclocking mechanism, was that it simply wasn't cost-effective. IE. putting this additional feature in would cost more, overall, than just simply allowing a minute percentage of chips to be overclocked.

Do you think that this will no longer be true? That there are so many people overclocking in the C2D era, that Intel is loosing money from it and can afford to add anti-overclocking features?

i think it costs about 15 minutes of effort from one engineer to permanently disable overclocking.

 

[aigomorla]

Originally posted by: Rebel44

Originally posted by: taltamir
from what i hear, intel feels so little pressure that it is making an entirely new anti overclocking capability to integrate into all non "Extreme" version chips...

Incorrect - http://www.nehalemnews.com/200...overclock-lock-on.html

uhh.. its not saying he is wrong either.

Its saying they will most likely figure out a way to hack it, however if they cant, and intel does indeed make it impossible and "lock" it, then he is correct.

Only time will tell.

 

[PeteRoy]

If AMD doesn't do something fast Intel will probably find excuses why they can't release Nehalem until 2010

 

[Genx87]

Was thinking about holding out for these chips but decided they probably wont be available and affordable for about 12 months. So I did a minor upgrade with an E8400 instead.

I "plan" on getting the 4 core variety and a minimum of 8GB of ram.

 

[Extelleron]

Originally posted by: Idontcare

Originally posted by: myocardia
Actually, it's going to be sometime in 2009 before the mainstream chips are released, possibly even mid or late 2009, if we're to believe the latest rumors.

The lead at this point is so damning that I 100% believe Intel's merely hedging against a dark horse coming out with Shanghai when they roadmap in anything Nehalem mainstream on the desktop coming from 45nm.

As soon as Intel has confirmation (i.e. retail shanghai chips in hand) of whether or not Penryn will be sufficient to fend off Shanghai you can bet everything mainstream relating to Nehalem gets slipped to Westmere and 32nm timeline for cost management and yield reasons.

I think the reason for this is that Intel doesn't care so much about expanding market share at this point - they are more concerned about their gross margins and their profits at the end of the day.

Gross margins have been a problem for Intel in the last few years and are down from where they used to be; they used to be close to 60% on average, and now we see closer to 50%.... Q1 2008 was 53.8%.

Right now Intel is beginning to replace its 65nm lineup with 45nm products that are cheaper to produce and sell in the same price brackets as the previous parts. This is of course good for margins. Penryn (@ 107mm^2) is replacing Conroe (143mm^2) in the <$266 market, and Yorkfield (214mm^2) is replacing Kentsfield (286mm^2) in the >$266 market. Granted there might be a slightly higher cost for an equivalently sized 45nm chip compared to a 65nm chip, but for certain they are saving some money.

Now if in 2009 Intel releases mainstream Nehalem at the same price brackets as the current Penryn/Yorkfield generation, then their margins will be going down. Nehalem is a more expensive part to produce than Penryn without a doubt, and this will be true especially for the quad-core parts; Nehalem quad is a 246mm^2 die, Yorkfield is two 107mm^2 die. Not only is die size going up, but the chip is well over 2x the size as the current ones and yields will definitely go down.

As you said, as long as Penryn is competitive with 45nm Phenom (which it should be; Yorkfield is solidly faster than 65nm Phenom and the 45nm parts are only going to be 10-15% faster. Even if they were equal or faster per clock, the Phenoms will not reach the same clocks as mature Penryn chips) I do not believe they will be making any push for Nehalem to become mainstream with the 45nm process. That is not to say they will not release Nehalem at any price bracket <$1,000, but I believe that only the high-end quads (Q9550 bracket) and high-end duals (E8600 bracket) would be replaced by a Nehalem equivalent in 2009. From comments made by an Intel employee on xtremesystems, the cost of entry for a Nehalem oc'ing platform in 2009 (LGA 1366) will be $400+ for the CPU and likely similar costs to current X38/X48 motherboards.

 

[bryanW1995]

Originally posted by: PeteRoy
If AMD doesn't do something fast Intel will probably find excuses why they can't release Nehalem until 2020

fixed your typo. :|

 

[bryanW1995]

Originally posted by: Extelleron

Originally posted by: Idontcare

Originally posted by: myocardia
Actually, it's going to be sometime in 2009 before the mainstream chips are released, possibly even mid or late 2009, if we're to believe the latest rumors.

The lead at this point is so damning that I 100% believe Intel's merely hedging against a dark horse coming out with Shanghai when they roadmap in anything Nehalem mainstream on the desktop coming from 45nm.

As soon as Intel has confirmation (i.e. retail shanghai chips in hand) of whether or not Penryn will be sufficient to fend off Shanghai you can bet everything mainstream relating to Nehalem gets slipped to Westmere and 32nm timeline for cost management and yield reasons.

I think the reason for this is that Intel doesn't care so much about expanding market share at this point - they are more concerned about their gross margins and their profits at the end of the day.

Gross margins have been a problem for Intel in the last few years and are down from where they used to be; they used to be close to 60% on average, and now we see closer to 50%.... Q1 2008 was 53.8%.

Right now Intel is beginning to replace its 65nm lineup with 45nm products that are cheaper to produce and sell in the same price brackets as the previous parts. This is of course good for margins. Penryn (@ 107mm^2) is replacing Conroe (143mm^2) in the <$266 market, and Yorkfield (214mm^2) is replacing Kentsfield (286mm^2) in the >$266 market. Granted there might be a slightly higher cost for an equivalently sized 45nm chip compared to a 65nm chip, but for certain they are saving some money.

Now if in 2009 Intel releases mainstream Nehalem at the same price brackets as the current Penryn/Yorkfield generation, then their margins will be going down. Nehalem is a more expensive part to produce than Penryn without a doubt, and this will be true especially for the quad-core parts; Nehalem quad is a 246mm^2 die, Yorkfield is two 107mm^2 die. Not only is die size going up, but the chip is well over 2x the size as the current ones and yields will definitely go down.

As you said, as long as Penryn is competitive with 45nm Phenom (which it should be; Yorkfield is solidly faster than 65nm Phenom and the 45nm parts are only going to be 10-15% faster. Even if they were equal or faster per clock, the Phenoms will not reach the same clocks as mature Penryn chips) I do not believe they will be making any push for Nehalem to become mainstream with the 45nm process. That is not to say they will not release Nehalem at any price bracket <$1,000, but I believe that only the high-end quads (Q9550 bracket) and high-end duals (E8600 bracket) would be replaced by a Nehalem equivalent in 2009. From comments made by an Intel employee on xtremesystems, the cost of entry for a Nehalem oc'ing platform in 2009 (LGA 1366) will be $400+ for the CPU and likely similar costs to current X38/X48 motherboards.

Blauhung's conjecture has already been discussed. Just because somebody works for intel doesn't mean that he has any greater insight into bloomfield pricing than many of us do. In fact, we KNOW that blauhung has nothing to do with setting the price of lga1366 cpus, because if he did then he wouldn't be throwing this info out on the xs forums. He basically punches a clock at intel and he THINKS that bloomfields will be available for $400.

I hope that he's right, as do most of us. However, where is the benefit to intel in doing this? If Q9650, which will be available in july for around $500, is faster than anything that amd comes out with in 2009, why in the world would intel kill their margins by selling 1366 cpus for $400???? Intel will be able to maintain market share with a superior product; if that superior product is cheaper to produce (penryn) then they will also make more money.

 

[Extelleron]

Originally posted by: bryanW1995

Originally posted by: Extelleron

Originally posted by: Idontcare

Originally posted by: myocardia
Actually, it's going to be sometime in 2009 before the mainstream chips are released, possibly even mid or late 2009, if we're to believe the latest rumors.

The lead at this point is so damning that I 100% believe Intel's merely hedging against a dark horse coming out with Shanghai when they roadmap in anything Nehalem mainstream on the desktop coming from 45nm.

As soon as Intel has confirmation (i.e. retail shanghai chips in hand) of whether or not Penryn will be sufficient to fend off Shanghai you can bet everything mainstream relating to Nehalem gets slipped to Westmere and 32nm timeline for cost management and yield reasons.

I think the reason for this is that Intel doesn't care so much about expanding market share at this point - they are more concerned about their gross margins and their profits at the end of the day.

Gross margins have been a problem for Intel in the last few years and are down from where they used to be; they used to be close to 60% on average, and now we see closer to 50%.... Q1 2008 was 53.8%.

Right now Intel is beginning to replace its 65nm lineup with 45nm products that are cheaper to produce and sell in the same price brackets as the previous parts. This is of course good for margins. Penryn (@ 107mm^2) is replacing Conroe (143mm^2) in the <$266 market, and Yorkfield (214mm^2) is replacing Kentsfield (286mm^2) in the >$266 market. Granted there might be a slightly higher cost for an equivalently sized 45nm chip compared to a 65nm chip, but for certain they are saving some money.

Now if in 2009 Intel releases mainstream Nehalem at the same price brackets as the current Penryn/Yorkfield generation, then their margins will be going down. Nehalem is a more expensive part to produce than Penryn without a doubt, and this will be true especially for the quad-core parts; Nehalem quad is a 246mm^2 die, Yorkfield is two 107mm^2 die. Not only is die size going up, but the chip is well over 2x the size as the current ones and yields will definitely go down.

As you said, as long as Penryn is competitive with 45nm Phenom (which it should be; Yorkfield is solidly faster than 65nm Phenom and the 45nm parts are only going to be 10-15% faster. Even if they were equal or faster per clock, the Phenoms will not reach the same clocks as mature Penryn chips) I do not believe they will be making any push for Nehalem to become mainstream with the 45nm process. That is not to say they will not release Nehalem at any price bracket <$1,000, but I believe that only the high-end quads (Q9550 bracket) and high-end duals (E8600 bracket) would be replaced by a Nehalem equivalent in 2009. From comments made by an Intel employee on xtremesystems, the cost of entry for a Nehalem oc'ing platform in 2009 (LGA 1366) will be $400+ for the CPU and likely similar costs to current X38/X48 motherboards.

Blauhung's conjecture has already been discussed. Just because somebody works for intel doesn't mean that he has any greater insight into bloomfield pricing than many of us do. In fact, we KNOW that blauhung has nothing to do with setting the price of lga1366 cpus, because if he did then he wouldn't be throwing this info out on the xs forums. He basically punches a clock at intel and he THINKS that bloomfields will be available for $400.

I hope that he's right, as do most of us. However, where is the benefit to intel in doing this? If Q9650, which will be available in july for around $500, is faster than anything that amd comes out with in 2009, why in the world would intel kill their margins by selling 1366 cpus for $400???? Intel will be able to maintain market share with a superior product; if that superior product is cheaper to produce (penryn) then they will also make more money.

Well I don't know, what blauhung said contradicts what I think Intel will do and what makes sense but I hope he is right and there are Nehalem options in 2009 other than the ultra high end.

It would seem to be that we could see Intel continue doing what they will do in Q3 of this year into '09; release a product one speed grade faster, and push the existing product into the lower bracket.

We see this w/ Q9650 replacing Q9550 @ $530 bracket, just like Q9550 replaced Q6700 at the $530 bracket previously. At $316 bracket, we see the Q9550 replacing the Q9450. At the $266 bracket, we see the Q9400 replace the Q9300, like the Q9300 replaced the Q6600 previously. Same thing w/ dual cores and E8600 replacing E8500 @ $266 and so on.

If Intel continues to improve yields and clocks with new revisions, it would seem likely to me they would follow up with an E8700 / Q9750 sometime in 2009.