• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Core i5 in September @ near price parity with i7?

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
it is quite clear they are not electrically compatible hence my confusion on why anyone, yourself included, would bother stating something that simply need not be stated?
I did not make the assumption that it is completely clear to everyone reading it. To you and me it is, but to some it might not be clear.

My point was more of "X58 was designed with QPI only input in mind, now they are making DMI chips, if they do not change the socket someone will put a DMI chip into an X58 mobo and it will mess things up, hence a different socket"

To be pragmatic about it, x58 not accepting DMI inputs is simply a matter of engineering/management decisions during the design phase, i.e. the decision was made to have the x58 not support DMI.
Pretty much.

Anand's article spoke about platform/chipset pricing, nothing about interface topology costs. I just read it again to confirm my initial impressions weren't misguided. He's quite clear about it with repetitious use of the wording.

Further he said Intel prices the x58 at a premium, not that it cost a premium to implement/produce/manufacture. He further stipulated that Intel's decision was to not lower the price on x58 but rather to introduce a lower-price DMI-based product (again, emphasis is on price to the consumer, i.e. market segmentation, not cost or gross margins impact on Intel's side of it).
Then i misread. in which case you seem to be right, I do not know weather it is more costly or if it is just an unexplained and pointless complication...

Maybe it is marketing, maybe there is some cost we are not aware of (board production cost maybe? power consumption?)
wait... power consumption... isn't QPI more power hungry?
 
Originally posted by: taltamir
wait... power consumption... isn't QPI more power hungry?

You thinking notebook? Like they wanted something other than QPI to be used for notebooks (with an eye towards IGP stuff) for power reasons and then figured why not dual-purpose the SKU for mainstream desktops? Certainly plausible.

To be honest I don't really care what the technical requirement is, I just so deplore the idea of old Intel coming back that early signs of it like DMI and LGA1156 gives me shivers so I am just hopeful someone knows for fact the technical reasoning.

Arguments about cost are exactly the kinds of arguments marketing guys will make as the justification, which aren't 100% misdirection so I can't wholly discount them either, but if someone could just make a good solid technical case for why DMI/LGA1156 must exist instead of a slowed down dumbed down QPI interface for Lynnfield segment (with fewer memory channels, etc, to create the segment as needed) then I can rule out the marketing angle with confidence and be done with it (personally).
 
I didn't follow the discussion but DMI has been PCIe lanes (x4 bandwidth to date, Gen 1) So it's likely that on-die PCIe controller will have more than x16 bandwidth. Perhaps x20 or even x24. At least it's fairly certain that DMI will be 2nd gen PCIe this time. And I figure Intel can simply reserve x4 out of x20/x24 for DMI? As for performance impact of PCIe bandwidth - it's fairly limited. There is of course ways to show the difference between x8 and x16, but that gets really beyond 'synthetic'. The biggest factor is board design from what I've seen, and possibly latency due to physical distance from slots to NB/CPU. AMD/NV both get around this (i.e. crappy board screwing multi-GPU performance) by bridge cables, and that's where most inter-GPU communication occurs, I heard. (and that bridge is PCIe x1)

This might change with next gen GPUs (GT300 is rumored to be a monster), and it gets interesting if that actually happens because now CPUs will govern PCIe lanes. Say if NV's next gen cards really need more than x8 bandwidth, then NV will be at Intel's mercy for necessary PCIe lanes. Users can resort to Socket 1366 but that's going to be a much smaller market so that leaves SLI as a very niche market (more so than now). If I were to be a little more sadistic - maybe Intel can spec out the on-die PCIe controller according to.. Larrabee's developement? 😀

After all is said and done, I think the rumour is correct that Intel will move on to 32nm quickly and that'll give more breathing room to on-die PCIe.. 45nm Nehalem (be it i7 or i5) might be fairly short-lived.
 
Originally posted by: lopri
After all is said and done, I think the rumour is correct that Intel will move on to 32nm quickly and that'll give more breathing room to on-die PCIe.. 45nm Nehalem (be it i7 or i5) might be fairly short-lived.

Is PCIe for clarkdale integrated into westmere or is it integrated into the MCM'ed IGP just like the memory controller?
 
Dunno but it'd make more sense to have it integrated, instead of MCM'ed. In an MCM CPU, the core and the GPU need to talk to each other via something.. so that means the core should have either QPI or PCIe built-in anyway. It'd make more sense to have PCIe controller built-in, but as usual I have no source or factual basis.
 
Originally posted by: taltamir

Maybe it is marketing, maybe there is some cost we are not aware of (board production cost maybe? power consumption?)
wait... power consumption... isn't QPI more power hungry?

Power consumption is definite advantage for DMI vs. QPI, but that isn't enough of an advantage in mainstream desktop.

DMI should be cheaper as its really a dedicated PCI Express x4 port and PCI Express is already on the chipset, while QPI is a brand-new high bandwidth interface. It might also require more routing lines for QPI than DMI which would reduce board costs.

I also like to quote an article first written when Nehalem came out: http://www.realworldtech.com/p...icleID=RWT040208182719

"Nehalem differs from the previous generation in that it was explicitly designed not only to scale across all the different product lines, but to be optimized for all the different product segments, from mobile to MP server. This implies a level of flexibility above and beyond the Core 2."

See, Bloomfield and Tylersburg(Core i7 and X58 chipset), is really a Gainstown and its partner chipset that had MP capabilities and ECC taken out. As mostly a 2P server platform, flexibility like 2x16 PCI Express 2.0 lanes and triple channel memory are probably more important over things like DMI and 2 chip configuration.

The current Core i7 only has 1% of Intel's sales, which is miserable. The Core 2 Quads have around 10% market share which means Core i7 isn't doing its job as its "successor" and its really up to Lynnfield to do it.

The rest of the users don't care about SLI/Crossfire and debatable advantages that tri channel brings over dual channel in PC setups. 2 chip configuration, lower power, possibly better single GPU performance with integrated PCI Express controller and cheaper board costs make Lynnfield attractive over Bloomfield with vast majority of the users.

I guess you can conclude the reasons are marketing. But I think they aren't just mere marketing based crippling but a solid reason for more buyers to migrate to the platform. It also provides a "trickle-down" for the cheaper and lower power platforms. Like, you don't want to have QPI on a ULV platform.

Dunno but it'd make more sense to have it integrated, instead of MCM'ed. In an MCM CPU, the core and the GPU need to talk to each other via something.. so that means the core should have either QPI or PCIe built-in anyway. It'd make more sense to have PCIe controller built-in, but as usual I have no source or factual basis.

The original "Havendale" had entire Northbridge on the "other" die. I can't see why it would be different as Clarkdale was a product that was pulled forward. Having PCI Express on the IMC/GPU side would have given flexibility to quickly migrate the 45nm core to 32nm core on the CPU, which might have been not true with integrated PCI Express controller.

Since the memory controller and GPU is already on the "other die", a connection like QPI is needed for communication between the two MCM chips.
 
Power consumption is definite advantage for DMI vs. QPI, but that isn't enough of an advantage in mainstream desktop.
It is for the new intel, the old intel had a 1:2 ratio... if something increases performance by X% and heat by 2X% it stays. the new intel is much tougher on power consumption, and is cutting it EVERYWHERE they can. This is why intel is now actually more power efficient than AMD, while historically it has been the other way around... it also helps them claim to to be more "green" and sells more computers to the "eco conscious" or whatever they call themselves today.
 
Confirmed: Core i7 920, 940, 950 to go "soon"

Guess Intel is making quite a room for their soon to be released i5 series. I dont know if going with i7s was such a good idea for those with limited budget seeing as the only upgrade paths will be EE editions, and 6 core westmeres (when they replace the high end bloom fields), and these will probably cost quite abit.
 
Even if it is true, since when in this industry is 6 months considered "soon"?

Much ado about nothing methinks, every CPU reaches EOL eventually.

The bit-tech folks are really trying to create the impression of drama where none exists IMO.
 
I dunno. I typically expect to put at least 2 CPUs in my mobo. My Socket A mobo had a 900 MHz Duron and a 1700+ T-bird. My 939 had a 3000+ Winny and a Opty 165. My 775 had a E5200 and it will get a quad eventually.

If I had just bought an X58 board and a 920, thats great. But when I go to upgrade, it's going to be a lot of money for a new CPU. Or a new mobo, CPU, and 1 wasted DDR3 DIMM(or 1 more). It just doesn't look like it has a good upgrade path unless you have a ton of money and need Tri or Quad SLI/Fire.
 
Originally posted by: Zensal
I dunno. I typically expect to put at least 2 CPUs in my mobo. My Socket A mobo had a 900 MHz Duron and a 1700+ T-bird. My 939 had a 3000+ Winny and a Opty 165. My 775 had a E5200 and it will get a quad eventually.

If I had just bought an X58 board and a 920, thats great. But when I go to upgrade, it's going to be a lot of money for a new CPU. Or a new mobo, CPU, and 1 wasted DDR3 DIMM(or 1 more). It just doesn't look like it has a good upgrade path unless you have a ton of money and need Tri or Quad SLI/Fire.

But when was that not to be expected for a platform created specifically for the crème de la crème of the enthusiast market?

Think many folks who could actually afford skulltrail were overly concerned with the upgrade path for their $600 mobo? Or those folks who found the coinage laying around to pickup a quadfather even earlier?
 
Originally posted by: Zensal
I dunno. I typically expect to put at least 2 CPUs in my mobo. My Socket A mobo had a 900 MHz Duron and a 1700+ T-bird. My 939 had a 3000+ Winny and a Opty 165. My 775 had a E5200 and it will get a quad eventually.

If I had just bought an X58 board and a 920, thats great. But when I go to upgrade, it's going to be a lot of money for a new CPU. Or a new mobo, CPU, and 1 wasted DDR3 DIMM(or 1 more). It just doesn't look like it has a good upgrade path unless you have a ton of money and need Tri or Quad SLI/Fire.

an i5 will never be an UPGRADE for an i7. the i5 is not REPLACING the i7, the LOWER END i7s are cut and new faster i7s are released, and cheaper i5s that are good enough are also released.

When you want to upgrade EITHER ONE, intel will have a 32nm i7 and i5 for you to go to.
 
Originally posted by: MrK6
Wow, that's kind of disappointing. I still don't see any reason to upgrade from my Q6600. If the power consumption numbers were better, maybe, but it'd be a side-step rather than an upgrade anyway.

You are probably good 'till 32nm...
 
Originally posted by: taltamir
Originally posted by: Zensal
I dunno. I typically expect to put at least 2 CPUs in my mobo. My Socket A mobo had a 900 MHz Duron and a 1700+ T-bird. My 939 had a 3000+ Winny and a Opty 165. My 775 had a E5200 and it will get a quad eventually.

If I had just bought an X58 board and a 920, thats great. But when I go to upgrade, it's going to be a lot of money for a new CPU. Or a new mobo, CPU, and 1 wasted DDR3 DIMM(or 1 more). It just doesn't look like it has a good upgrade path unless you have a ton of money and need Tri or Quad SLI/Fire.

an i5 will never be an UPGRADE for an i7. the i5 is not REPLACING the i7, the LOWER END i7s are cut and new faster i7s are released, and cheaper i5s that are good enough are also released.

When you want to upgrade EITHER ONE, intel will have a 32nm i7 and i5 for you to go to.

You are certain that P55 will have a 32nm upgrade? To what? There might be a 32nm Nehalem die shrink that is 10% faster but don't expect more than that. Sandy Bridge will have a different platform. I highly doubt I will buy a $1000 6 core CPU. I should be good with this X58 until Sandy Bridge in 2011. Those that buy a P55 won't be able to run XFire/SLI high-end or high mid-range cards before that without a substantial performance penalty.

I payed a whole $40-$50 more for that and I get to use it for 6 months before Lynnfield has a stable BIOS which is a great deal. Not to mention Lynnfield may not give you a 1.4Ghz OC or better. It's a no-brainer IMHO.

GTA 4 shows gains from OC'ing PCIE bus. At 1680x1050 and no AA no less.
http://www.pcgameshardware.com...ticle_id=674715&page=1

Not to mention this review. It is tri-SLI but it will be similar for SLI this fall. Lynnfield should be similar to the 9770.
http://www.guru3d.com/article/...-performance-review/19

The games are even GPU bottlenecked at 1024x768 with a 965 i7 with tri-SLI.
 
There'll be 32nm 6-cores for s1366, and a 32nm dual core for s1156. We haven't gotten any info on a 32nm quad for either platform.
 
Originally posted by: taltamir
an i5 will never be an UPGRADE for an i7. the i5 is not REPLACING the i7, the LOWER END i7s are cut and new faster i7s are released, and cheaper i5s that are good enough are also released.

When you want to upgrade EITHER ONE, intel will have a 32nm i7 and i5 for you to go to.

I know that i5 is not an upgrade from i7.

All I'm saying is that with LGA-775, I could take my E2100 in (almost) any LGA-775 board and drop in a QX9770 and I've just upgraded to a top of the line CPU.

AMD is doing the same thing with there AM2/AM3 boards. You can buy any processor from AMD and drop in into (almost) any motherboard. People like to not have to worry about which board they have, or if they have some money and want to upgrade, have to figure out if they want to change their whole system for a processor upgrade.

Intel should have used the i7 LGA-1366 for Lynnfield and made it somewhat crippled for P55, just like P45 vs X48. I know there are technical differences between the two, such as QPI or triple channel memory. Intel could have released Lynnfield with dual channel memory and lower QPI, or less bandwidth for PCI-E, and left the LGA1366 so that people could upgrade the same way they have been on LGA-775.

The only reason they didn't is to force people who want the highest performance to buy an overpriced motherboard with an overpriced EE processor instead of people like me buying an E5200 and overclocking it. You can't overclock whats not physically there. 920 was to test out the platform and fix the bugs. We already know that it's being discontinued.

The LGA-1156 could and should have been relegated to SFF and laptops.

 
P55, P57, H55 (not sure on this one), H57 all at display.. (Computex) Reviewers already received Lynfield samples.. Intel planning to hit the market in a full swing with Lynfield.. Lynfield not likely be named i5.. (maybe Intel wants to take marketing advantage of number "7" as Windows 7 will arrive at the same time?)

It's all hearsay, but I returned my un-opened 920 today.
 
what about the i7 960? When does it come out/not come out/eol? I saw it on the evga forums, someone had a D0 960. Why would intel have all these different bloomsfield chips if they were just going to eol them in 6 months? Or is the 960 one of the few to be left? I personally think if they are going to axe the i7 for the most part, they should at least keep the 920 (d0) and the 960 or something.
 
read the anadtech article about the subject, it gives exact details on what new chips are made, and which get the axe.
 
Originally posted by: Shmee
what about the i7 960? When does it come out/not come out/eol? I saw it on the evga forums, someone had a D0 960. Why would intel have all these different bloomsfield chips if they were just going to eol them in 6 months? Or is the 960 one of the few to be left? I personally think if they are going to axe the i7 for the most part, they should at least keep the 920 (d0) and the 960 or something.

You should be able to find them through the end of the year at least anyway. 975 will be the only one produced after September. They probably produced them as a stopgap measure until Lynnfield was ready.

But like I said before why moan that they are discontinuing them when you have the opportunity to get the platform as cheap as it is now? You should be grateful for the ability to buy it at all at these prices. They could have easily just produced the 960 for the X58 in the first place. There hasn't quite been an opportunity like this before. The glass seems half full to me.
 
Back
Top