What makes sandy bridge so good anyways?

[Zendoren]

As the subject says, What makes the new sandy bridge worth waiting for?

From what I have read so far, it will be a bear to overclock unless you have an unlocked processor (source) Which I'm sure Intel will take us to the bank us for. Also, the new south bridge, ICH11, seems a ways out from an official release date which makes a single chip solution for SATA 6Gb and USB 3.0 out of reach. To top all this off, it seems that we are going back to the DMI architecture instead of QPI. (Correct me if i am wrong with this one)

So besides a smaller dye for more cores, what are the perks of this processor?

 

[drescherjm]

So besides a smaller dye for more cores, what are the perks of this processor?

The initial revision will not have more cores.

1. IPC improvement extending Intel's lead to 30 to 50% faster clock for clock than any processor AMD currently sells.
2. Also lower power. Instead of 130W on the high end 95W max.
3. And at that slightly higher frequency.
4. On die GPU will be good enough for over 90% of users.

 

[Furystrike]

Are your issues with 1155 AND 2011, or just 1155? 2011 will be the high end counterpart to 1155 and will have high end features not available on 1155.

 

[Zendoren]

Are your issues with 1155 AND 2011, or just 1155? 2011 will be the high end counterpart to 1155 and will have high end features not available on 1155.

Will sandy be available for LGA 2011 on release?
I guess a better question would be, will there be MBs that till support this on release? (Opps, found my answer, Q4 2011)

Also, I'm sure the Sandy Bridge EX will be un-godly priced on release as well!?

Lastly, will there be support for the new PCIe 3.0 with the LGA 2011 if/when the specs are released!?

@drescherjm, I see you your point with the IPC improvements, should the new AVX instruction set also make me drool a little as well?

 

[drescherjm]

should the new AVX instruction set also make me drool a little as well?

If you have software that can use them, yes..

 

[Zendoren]

If you have software that can use them, yes..

We barely have software that uses 64 bit multi-core processors. So I'll take this feature with a grain of salt!

 

[drescherjm]

At least in my profession as a programmer in medical imaging research I can write my own 64 bit multithreaded code and I can tell you for my applications this is a tremendous gain and a lot less frustration when dealing with images that are larger than 2GB. No longer do I have to worry about not being able to allocate a contiguous array of larger than ~1.2 GB (under windows).

 

[plugwash]

it seems that we are going back to the DMI architecture instead of QPI. (Correct me if i am wrong with this one)

QPI and DMI are different interfaces for different functions.

DMI serves to connect what would conventionally be referred to as the southbridge (with LGA1156 intel calls it a PCH but it's essentially the same thing) into the system. It does this function in all current intel systems.

With LGA1366 QPI serves to interconnect the processor(s) and the IO Hub(s) (an IO HUB is essentially a northbridge without a memory controller, most boards have only one but supermicro do some boards with two processors and two IO hubs). Fast PCIe lanes come off the IO Hub(s). The IO hub(s) provide fast PCIe connections and also provice the DMI link for the southbridge. This means you get plenty of fast PCIe lanes (36 per IO hub). But it also makes the platform relatively expensive.

With LGA1156 and LGA1155 the fast PCIe comes directly off the CPU rather than via QPI to an IO Hub. This makes things cheaper but does mean fast PCIe is a bit limited (though bridge chips will help with this to some degree).

With sandy bridge intel seems to be introducing the midrange first which kind of makes sense since the midrange stuff is currently still on 45nm while the top and bottom end stuff is on 32nm.

From what I can gather the high end sandy bridge will be using a different system again with on-chip PCIe but also QPI availiable for multiprocessor interconnect and add-on chipsets. Information on high end sandy bridge stuff is somewhat sketchy at the moment though.

As the subject says, What makes the new sandy bridge worth waiting for?

The reasons people are excited about sandy bridge are simple, the combination of higher clockspeeds and higher per core performance at a given clock will make them quite a bit faster than the fastest current quad cores and rumours are the unlocked ones will be very overclockable too (though noone really knows on that until people get their hands on the production silicon).

From what I have read so far, it will be a bear to overclock unless you have an unlocked processor (source) Which im sure Intel with bleed us for.

Well the pricing "leak" seems to indicate they won't be bleeding overclockers that heavilly. We won't know for sure until they actually release it of course

edit: got the number of PCIe lanes wrong, mustn't post while half asleep

 

[Zendoren]

So the story, for Gaming Only rig builders, is that 1366 socket is the way to go until the 2011 socket comes out?

Even then,as a gamer only user, anything over quad core would be a waist of $$$ until the games code for the extra cores.

With this said, will there be a unlocked i7 quad core flavor available?

 

[plugwash]

So the story, for Gaming Only rig builders, is that 1366 socket is the way to go until the 2011 socket comes out?

1366 will give you more fast PCIe lanes but 1155 will give you more CPU. I think for gamers the latter will probablly end up more important but we will have to wait for the benchmarks to know for sure.

 

[ilkhan]

1366 will give you more fast PCIe lanes but 1155 will give you more CPU. I think for gamers the latter will probablly end up more important but we will have to wait for the benchmarks to know for sure.

So the story, for Gaming Only rig builders, is that 1366 socket is the way to go until the 2011 socket comes out?

Even then,as a gamer only user, anything over quad core would be a waist of $$$ until the games code for the extra cores.

With this said, will there be a unlocked i7 quad core flavor available?

If you're running top end SLI/X-fire and actually need the PCI-E bandwidth *dual or triple 5970/6990/480/580 might need 16x16, dual 6870 doesn't), s1366/s2011 is the way to go. If you need hex cores (and are willing to pay for them), s1366/s2011 is the way to go. If you need 24-48GB of RAM, s1366-2011 is the way to go. If not, s1155 is faster per dollar by a good margin. For the *VAST*majority of people, none of those apply.

s1155 is the replacement for s1156, which replaced s775 (s1366 is really a replacement for s771 with lots of PCI-E). The market got confused by s1366's launch pricing and the long delay before s1156 launched. But its a specialized subset of the market; s115x is the mass market socket for 95% of users.

 

[havermeyer]

5GHz on air with better scaling clock-for-clock at a nice price, even unlocked? Seems pretty good. Obviously, we'll need production parts after launch to confirm this, but it's looking bright so far.

With this said, will there be a unlocked i7 quad core flavor available?

The "K" versions have an unlocked multiplier with the i7-2600K supposed to be available at launch.

The leaked prices are showing a ~$10-30 premium for unlocked parts.

 

[Zendoren]

Will sandy be available for LGA 2011 on release?
I guess a better question would be, will there be MBs that till support this on release? (Opps, found my answer, Q4 2011)

Also, I'm sure the Sandy Bridge EX will be un-godly priced on release as well!?

Lastly, will there be support for the new PCIe 3.0 with the LGA 2011 if/when the specs are released!?

@drescherjm, I see you your point with the IPC improvements, should the new AVX instruction set also make me drool a little as well?

Ye shall ask and Wikipedia shall provide!

s2011 will have PCIe 3.0 support! (source) (However gaming wise, unnecessary until the video card producers max out the current PCIe 2.0 bandwidth. With OpenCL up and coming this should prove useful for biotech industry in need of repetitive processing power)

I'm guessing this socket will be supported by the new X68 NB and the ICH11 SB?

I'f so, I'm liking what i'm seeing.

Update: looks like 4 channel memory will be used with this socket and chip set. Makes since seeing how this is aimed at server builds more than gaming builds.

 

[ilkhan]

s2011 is quad memory, 40 PCI-E v3 and X68 in Q3'2011
s1155 is dual memory 20 PCI-E v2 (16 on-die + 4 via southbridge) and P67/H67/etc on 1-8-10 (IIRC that supposed to be the announcement date)

 

[djsb]

I remember hearing a rumor that the 4 southbridge PCIe will not be artificially slowed to v1 speeds on Sandy Bridge, but I can't remember where. Does anyone else remember hearing something similar?

 

[3far4shot]

There were some benchmarks posted showing a significant increase in performance over the current Nehalem chips. Though im personally not upgrading to it anytime soon due to lack of funds and the fact that games these days can run on a toaster (no point in buying the Lamborghini if you're just going to drive it around the parking). The on-board GPU is a benefit for a ton of people (the mainstream), though i personally don't have a huge interest in that aspect of Sandy Bridge.

 

[haxxorboi]

Keep in mind that 2011 is Q3/4 2011 whereas 1155 is Q1 2011. 2011 is likely to be expensive as hell (it is enthusiast after all) whereas the i5-2500k and i7-2600k will be about $200-220 and $275-300 respectively.

5+GHz on air for the K series chip have been reported which should make SB a very good investment for most 775 or current i5/7 users that want to stay bleeding edge.

 

[Vega]

There were some benchmarks posted showing a significant increase in performance over the current Nehalem chips. Though im personally not upgrading to it anytime soon due to lack of funds and the fact that games these days can run on a toaster (no point in buying the Lamborghini if you're just going to drive it around the parking). The on-board GPU is a benefit for a ton of people (the mainstream), though i personally don't have a huge interest in that aspect of Sandy Bridge.

A hand full of console ports does not include all "games these days". I play some games that bring my 4.55Ghz i7 to it's knees. Faster CPU's are still very much needed for a lot of us.

 

[3far4shot]

A hand full of console ports does not include all "games these days". I play some games that bring my 4.55Ghz i7 to it's knees. Faster CPU's are still very much needed for a lot of us.

I would say the large majority of games coming to PC today are crappy coded/unoptimized console ports, i was not trying to overgeneralize with that statement, but it is the truth. , i also never said that faster CPU's aren't needed for others, it largely depends on what software if being ran by the user, and if gaming isn't your main prerogative then you buy the chip that matches your needs/price, i never stated otherwise.

 

[PGHammer]

The initial revision will not have more cores.

1. IPC improvement extending Intel's lead to 30 to 50% faster clock for clock than any processor AMD currently sells.
2. Also lower power. Instead of 130W on the high end 95W max.
3. And at that slightly higher frequency.
4. On die GPU will be good enough for over 90% of users.

Issue #4 you pointed out above is why I call it a glorified G4x (G41 and friends) successor.

Granted - G4x is easily the best IGP that Intel's ever done. However, that's largely because of what it permits that had previously been atypical for an IGP.

1. G4x supports DX10 (atypical for Intel) and supports a dedicated GPU option (either instead of or alongside - the last decidedly atypical for Intel).
2. The IGP has multiple output options (entirely dependent on what the board maker wants to do) - VGA/DB-15, DVI (single or dual-link) and HDMI (singly or in combination).
3. Except for the oldest LGA775s, if it fits, G4x can swallow it - from single-core to quad, and anywhere in-between. The ability to swallow the Extreme quads is unique.

Sandy Bridge duplicates all except #3, as there are no Extreme-series CPUs planned for 1155 - I have no idea yet if it ups the ante on IGP support to DX11 (matching AMD and nVidia; if it does, it would be unique in the 1155 desktop space).

 

[PGHammer]

I remember hearing a rumor that the 4 southbridge PCIe will not be artificially slowed to v1 speeds on Sandy Bridge, but I can't remember where. Does anyone else remember hearing something similar?

And what cards use PCIe x4 today?

As long as Sandy Bridge has a single PCIe x16 slot (as G4x does today) that would tend to be plenty for the target audience.

 

[SHaFT7]

PCIe x4 is being used by a select few USB3+SATA 6Gbps combo cards, and more notably RAID cards (assuming compatibility with the motherboard in question). I'm sure there are other things as well but these are the first that come to mind that I've seen.

 

[Gaiden133]

So, are there any speculations as to which would be faster between LGA1366 and LGA1155?

So, a 2600K OC'd versus i7-950 OC'd ?

 

[R-Type]

Issue #4 you pointed out above is why I call it a glorified G4x (G41 and friends) successor.

Granted - G4x is easily the best IGP that Intel's ever done. However, that's largely because of what it permits that had previously been atypical for an IGP.

1. G4x supports DX10 (atypical for Intel) and supports a dedicated GPU option (either instead of or alongside - the last decidedly atypical for Intel).
2. The IGP has multiple output options (entirely dependent on what the board maker wants to do) - VGA/DB-15, DVI (single or dual-link) and HDMI (singly or in combination).
3. Except for the oldest LGA775s, if it fits, G4x can swallow it - from single-core to quad, and anywhere in-between. The ability to swallow the Extreme quads is unique.

Sandy Bridge duplicates all except #3, as there are no Extreme-series CPUs planned for 1155 - I have no idea yet if it ups the ante on IGP support to DX11 (matching AMD and nVidia; if it does, it would be unique in the 1155 desktop space).

That is just flat out wrong. The SB IGP is more than 4x faster than G41 which is itself a couple generations behind the curve for Intel chipsets. Why are you even bring up such an obscure old igp?

 

[E4g1e]

s2011 is quad memory, 40 PCI-E v3 and X68 in Q3'2011
s1155 is dual memory 20 PCI-E v2 (16 on-die + 4 via southbridge) and P67/H67/etc on 1-8-10 (IIRC that supposed to be the announcement date)

Actually, most s1155 platforms will have up to 24 PCIe 2.0 full-bandwidth lanes. The P67 PCH and most of the other s1155 PCHs will have up to 8 PCIe 2.0 full-bandwidth lanes. And based on what's on Wikipedia, most of the chipsets will offer native SATA 6 Gbps support (though there will be no more than two SATA 6 Gbps ports out of a total of six, with the remainder of the six SATA ports limited to SATA 3 Gbps).

 

[PGHammer]

That is just flat out wrong. The SB IGP is more than 4x faster than G41 which is itself a couple generations behind the curve for Intel chipsets. Why are you even bring up such an obscure old igp?

Is or is not Sandy Bridge an IGP chipset?

What did the current LGA1156 version (G5x) bring to the table over and above G4x? The reason why I said G4x (rather than G5x) is because G5x has largely been a fail (and especially compared to G4x). G4x is also the direct successor to the first LGA775 IGP: G3x (Bear Lake and progeny).

G3x was a success, and G4x (and surprisingly, including the bottom-end G41) has been an even bigger one. G5x, on the other hand, has been a flop.
G5x has gotten practically zero traction - not in BYO, and not even in notebooks and laptops (the only traction it has gotten is in prebuilts, and that is strictly for reasons of price; even there, it is finding itself severely undercut by AMD and the A-II/P-II pairing with G78x). In BYO, it's losing to discrete-only P55 (and even to the LGA775-based G41).
Age (and speed) are not everything, even at the bottom end/IGP marketplace. For a new chipset to successfully replace another (even from the same manufacturer, such as Intel) it must offer a compelling reason for it to be chosen, which G5x has miserably failed to do. That is why I said Intel's real target with Sandy Bridge is not the horrible fail that was G5x, but the rousing success that has been G4x (and G41 in particular).

 

[R-Type]

Is or is not Sandy Bridge an IGP chipset?

What did the current LGA1156 version (G5x) bring to the table over and above G4x? The reason why I said G4x (rather than G5x) is because G5x has largely been a fail (and especially compared to G4x). G4x is also the direct successor to the first LGA775 IGP: G3x (Bear Lake and progeny).

G3x was a success, and G4x (and surprisingly, including the bottom-end G41) has been an even bigger one. G5x, on the other hand, has been a flop.
G5x has gotten practically zero traction - not in BYO, and not even in notebooks and laptops (the only traction it has gotten is in prebuilts, and that is strictly for reasons of price; even there, it is finding itself severely undercut by AMD and the A-II/P-II pairing with G78x). In BYO, it's losing to discrete-only P55 (and even to the LGA775-based G41).
Age (and speed) are not everything, even at the bottom end/IGP marketplace. For a new chipset to successfully replace another (even from the same manufacturer, such as Intel) it must offer a compelling reason for it to be chosen, which G5x has miserably failed to do. That is why I said Intel's real target with Sandy Bridge is not the horrible fail that was G5x, but the rousing success that has been G4x (and G41 in particular).

The G41 was not a 'rousing success' and using language like that to describe an igp that couldn't do basic htpc functions correctly indicates some sort of perverted bias for this particular chipset that I suspect stems from a latent need to defend your sig system.

I also don't see how you could possibly label the 5x series a failure (other than it not being called a G41). It fills the exact same need as the G4x chipsets did, namely to provide graphics on the lowest common denominator systems. To this end it is every bit as prevalent on 1156 machines as the g4x was on 775. If it seems there are less of them out there it is only because of the massive install base for 775 machines and the fact they continue to be sold on the ultra low end even now.


Back on topic, the igp on SB is a huge improvement over previous efforts but still doesn't bring parity to what Nvidia and Ati offer because of its lack of hardware open cl implementation. The big leap forward will come with Ivy Bridge and its completely redesigned graphics core.

Still, if you are buying a laptop anytime soon and aren't in the market for a unit with discreet graphics then you should definitely wait for SB. You get twice the graphics performance and significant power efficiency gains that will make a big impact in the mobile user experience.

 

[ilkhan]

Actually, most s1155 platforms will have up to 24 PCIe 2.0 full-bandwidth lanes. The P67 PCH and most of the other s1155 PCHs will have up to 8 PCIe 2.0 full-bandwidth lanes. And based on what's on Wikipedia, most of the chipsets will offer native SATA 6 Gbps support (though there will be no more than two SATA 6 Gbps ports out of a total of six, with the remainder of the six SATA ports limited to SATA 3 Gbps).

OK, let me correct myself. 4 lanes *TO* the southbridge, X lanes on the southbridge.

On the IGP topic, my arrandale does quite well for what I need it to do, dual 1080p h264 decoding, and very mild gaming (in my case just EVE). It could be faster, but I've never seen a "too fast" GPU. Lots of westmere based laptops out there.

 

[PGHammer]

The G41 was not a 'rousing success' and using language like that to describe an igp that couldn't do basic htpc functions correctly indicates some sort of perverted bias for this particular chipset that I suspect stems from a latent need to defend your sig system.

I also don't see how you could possibly label the 5x series a failure (other than it not being called a G41). It fills the exact same need as the G4x chipsets did, namely to provide graphics on the lowest common denominator systems. To this end it is every bit as prevalent on 1156 machines as the g4x was on 775. If it seems there are less of them out there it is only because of the massive install base for 775 machines and the fact they continue to be sold on the ultra low end even now.


Back on topic, the igp on SB is a huge improvement over previous efforts but still doesn't bring parity to what Nvidia and Ati offer because of its lack of hardware open cl implementation. The big leap forward will come with Ivy Bridge and its completely redesigned graphics core.

Still, if you are buying a laptop anytime soon and aren't in the market for a unit with discreet graphics then you should definitely wait for SB. You get twice the graphics performance and significant power efficiency gains that will make a big impact in the mobile user experience.

The very fact that LGA775 (and G41 in particular) is still alive and kicking despite Intel's attempts to kill it with Arrandale (G5x) and those purpose-built i3/i5 CPUs for it (that can't really be used in anything else) sounds like (except in prebuilts and some laptops and notebooks) a bit of a problem vis-a-vis G41.

Laptop/notebook owners going Intel usually aren't in a market for discrete graphics for one big reason - there's no such thing as a discrete-graphics market in that space. Further, is nV building any laptop/notebook chipsets for the post-LGA775 Intel CPUs? (I already know that AMD is not.)

The reason why I refer to G4x in general (and G41 in particular) as a success is because of the desktop/HTPC market (and the corporate basic desktop market, which has stayed with LGA775 for, as we both agree, largely reasons of price). Arrandale *desktop* motherboards cost more than the discrete-only P55 - you don't consider that a problem?
And while G5x is getting *some* traction in the mobile space, it's still undercut in price - and flat-out smoked in performance - by mobile A-II configurations (mostly with AMD's own chipsets). While some companies (notably Toshiba) refuse to build/sell any AMD-based portables, the same certainly can't be said of HP. While Arrandale may have better performance numbers, it still has that ugly price/performance problem hanging around its neck like a dead albatross, and that is especially true in the desktop/HTPC space. *That* is what Sandy Bridge must address.

 

[R-Type]

The very fact that LGA775 (and G41 in particular) is still alive and kicking despite Intel's attempts to kill it with Arrandale (G5x) and those purpose-built i3/i5 CPUs for it (that can't really be used in anything else) sounds like (except in prebuilts and some laptops and notebooks) a bit of a problem vis-a-vis G41.

Laptop/notebook owners going Intel usually aren't in a market for discrete graphics for one big reason - there's no such thing as a discrete-graphics market in that space. Further, is nV building any laptop/notebook chipsets for the post-LGA775 Intel CPUs? (I already know that AMD is not.)

The reason why I refer to G4x in general (and G41 in particular) as a success is because of the desktop/HTPC market (and the corporate basic desktop market, which has stayed with LGA775 for, as we both agree, largely reasons of price). Arrandale *desktop* motherboards cost more than the discrete-only P55 - you don't consider that a problem?
And while G5x is getting *some* traction in the mobile space, it's still undercut in price - and flat-out smoked in performance - by mobile A-II configurations (mostly with AMD's own chipsets). While some companies (notably Toshiba) refuse to build/sell any AMD-based portables, the same certainly can't be said of HP. While Arrandale may have better performance numbers, it still has that ugly price/performance problem hanging around its neck like a dead albatross, and that is especially true in the desktop/HTPC space. *That* is what Sandy Bridge must address.

You are correct that the arrandale platform hasn't seen nearly the market penetration that 775 with the g4x chipsets has, but you are incorrectly attributing this to the igp. Rather, the reasons rest entirely on the 775 platform itself.

For one, the platform had such a long life that most oem's were able to tweak and reuse designs for several years which let to terrific cost savings for them.

The second, and biggest, reason is that it was on 775 that C2D hit and we collectively reached the point where 95% of users stopped needing something faster. This, combined with the global economic downturn, created a market that was not buying many new PC's.

At this stage in the game however, virtually every new corporate purchase will have at least arrandale. You say the p55 chipset is 'only used in some prebuilt laptops'... wtf? That is 99% of all laptops! If you see a laptop with i3, i5 or i7 in it, it would be the exception if it DIDN'T have the intel igp.

As much as I wish they did, Intel will have no problem pushing any of their IGP's no matter how horrific the performance is. If you buy one of their processors in an oem system, you are far more likely than not to receive the corresponding IGP as well. Trying to skew this fact to infer preferences about the igp itself is missing the big picture completely.

 

[crippledlemming]

I'm in it for more cores

 

[PGHammer]

You are correct that the arrandale platform hasn't seen nearly the market penetration that 775 with the g4x chipsets has, but you are incorrectly attributing this to the igp. Rather, the reasons rest entirely on the 775 platform itself.

For one, the platform had such a long life that most oem's were able to tweak and reuse designs for several years which let to terrific cost savings for them.

The second, and biggest, reason is that it was on 775 that C2D hit and we collectively reached the point where 95% of users stopped needing something faster. This, combined with the global economic downturn, created a market that was not buying many new PC's.

At this stage in the game however, virtually every new corporate purchase will have at least arrandale. You say the p55 chipset is 'only used in some prebuilt laptops'... wtf? That is 99% of all laptops! If you see a laptop with i3, i5 or i7 in it, it would be the exception if it DIDN'T have the intel igp.

As much as I wish they did, Intel will have no problem pushing any of their IGP's no matter how horrific the performance is. If you buy one of their processors in an oem system, you are far more likely than not to receive the corresponding IGP as well. Trying to skew this fact to infer preferences about the igp itself is missing the big picture completely.

And you just pointed out why that's the case, in terms of portables - lack of chipset (not CPU) competition.

What made G4x different was the presence outside of portables - not just the corporate basic desktop space and even the HTPC space, but the motherboard-replacement space. (To the person that accused me of trying to talk up why my rig is the way it is, I made no bones about the motherboard purchase being what it was based on price; also, in case you didn't notice, it was an nForce-chipset motherboard, with IGP, that my LX2 replaced, and I'm normally no fan of IGPs from anybody, and especially not in terms of desktops.)

There *have* been some fantastic IGPs as of late from *all* the major players (by *as of late* I am referring to since Vista's launch). However, where are the nForce IGPs in the desktop space? AMD still has a presence there, but only with their CPUs (naturally, with AMD acquiring ATI, Intel would not allow any part or partner of AMD to be a chipset source, not even in the low-end IGP space). That means Intel must take up that slack itself.

I'm not saying that IGPs aren't important in the portable space, as I know better. I'm saying, in fact, the opposite - that they are still important in the *desktop* space (corporate desktop, HTPC, light/value gaming, etc.) by defining the floor level of acceptable performance for discrete graphics. Discrete graphics that performs *worse* than integrated graphics should not sell, and especially not for any premium over integrated (which is why, until they finally sold out, you could get sub-$10USD 8400GS PCIe graphics cards @ MicroCenter - they performed little better than G41 due to gimped memory buses and loadouts). G41 was, in fact, a major raise in integrated-graphics performance for Intel-based desktop motherboards (especially with both nV and AMD having pretty much left that space). G5x being a non-factor on the desktop side of things is, if anything, a major step backward for Intel (and especially after back-to-back G3x/G4x). That is why I said that Sandy Bridge (in the G6x form) is more critical than ever, and especially on the desktop side of things, as opposed to the portable side.

 

[R-Type]

I gave it two tries, I will not waste any more time trying to get you to see the holes in this delusional argument you have concocted. At this point I fear if I succeed it would be akin to waking a sleepwalker...


Back on topic: SB is 'so great' because it has 20-30% ipc improvement over arrandale on the cpu and 100% improvement over arrandale on igp performance. Its a fast ass chip, thats why its great.

 

[plugwash]

IMO the big reason for the first gen i series IGP not taking off in desktops (in laptops it seems to be doing just fine) is nothing to do with the IGP itself, it's the fact that for whatever reason intel failed to combine it with a quad core processor.

So as an OEM you had three choices
Have seperate model lines for dual core and quad-core (afaict corp buyers like to be able to get a range of specs on the same model line, makes image managment easier)
Stick with LGA775
Supply every quad core buyer with a discrete graphics card (Which will also complicate image management because the quad cores will have discrete graphics and the duals integrated)

To put it another way I don't think the problem is the integrated graphics in i3/ dual core i5, it's the lack of integrated graphics support in quad core i5 and to a lesser extent i7.

This is a problem that sandy bridge WILL be fixing.

 

[PGHammer]

I gave it two tries, I will not waste any more time trying to get you to see the holes in this delusional argument you have concocted. At this point I fear if I succeed it would be akin to waking a sleepwalker...


Back on topic: SB is 'so great' because it has 20-30% ipc improvement over arrandale on the cpu and 100% improvement over arrandale on igp performance. Its a fast ass chip, thats why its great.

Not disputing it; however, unless it appears in the desktop space, at a decent price, it will be a *niche* product, as Arrandale (G5x), unfortunately, has become.

Not saying that portables aren't important - however, they are a niche market. A large niche, but still a niche.

The IGP is important in the desktop space because it defines what the acceptable floor level of performance is. With G3x and G4x, Intel defined what acceptable performance for an IGP in the desktop space was. Right now, there is *no* desktop IGP from Intel in the 1156 space (Arrandale is barely present). If you want a decent desktop IGP and you want something later than G4x, you go AMD, not Intel G5x. That is a step *backward*, in my humble opinion.

Portables are important - however, they aren't everything.

 

[PGHammer]

IMO the big reason for the first gen i series IGP not taking off in desktops (in laptops it seems to be doing just fine) is nothing to do with the IGP itself, it's the fact that for whatever reason intel failed to combine it with a quad core processor.

So as an OEM you had three choices
Have seperate model lines for dual core and quad-core
Stick with LGA775
Supply every quad core buyer with a discrete graphics card.

This is a problem that sandy bridge WILL be fixing.

Which is the point I was trying to bring up.

When Arrandale was first discussed, it was pointed out prior to launch that every CPU that would mate up with Arrandale would be dual-core - and I predicted that it would be nothing short of a disaster because it was a step backward, not just from G4x, but even G3x (G4x supports every 45nm LGA775 CPU that Intel has produced, and all the older quad-cores in the LGA775 formfactor as well). Only X48 among LGA775 chipsets matches the length and breadth of CPU support. And on the desktop, the lack of quad-core support, combined with falling prices for the quad-core CPUs, in not just LGA1156 and LGA1366, but even LGA775, have made things worse for the planned successor to G4x (AMD adding insult to injury by coming out with a kick-butt IGP itself for its own multicore CPUs basically kicked desert sand in Intel's face). And I had not even purchased my nForce IGP-based motherboard at that time, let alone the board that would replace it.

One factor that weighed heavily in the purchase of *both* motherboards (the P5N-EM I originally bought, and the P5G41C-LX2/GB I have now) was quad-core support, in addition to price. I did not expect quad-core pricing to stay at the $150 range it was when I bought the first (however, even I would not have dared predict sub-$100 Intel quad-core CPU pricing, even during what amounts to the LGA775 Farewell Tour).

While quad-core CPU support is not critical in the portable marketplace (yet), it certainly is in the desktop space, even at the low end.

 

[ilkhan]

You realize arrandale is the codename for the mobile dual cores, right?
clarkdale is the desktop dual cores, and the IGP die itself is ironlake (IIRC).

I haven't recommended an AMD system, of any kind, since the K8 days. They may have a better IGP, but the westmere GPU is plenty for 99% of systems out there. The OEM market is still dominated by dual cores, so quads not having the IGP die isn't much of an issue there either. G4x was OK, but the IGP for westmere was a huge step up and sandy's is a leap beyond that.

My server is G31 based, my XPS M1330, which would have been GM4x IIRC, had the 8400GS in it, my E6410 is H55M, and none of my gaming PCs use integrated. So personally G41 didn't do anything for me.

 

[Activate: AMD]

I still don't see SB being some kind of god chip. It seems like whenever I hear anyone talking about Sandy Bridge these days its just Intel fanboys foaming at the mouth for the latest and greatest thing like it was instantly going to turn every chip that came before it into a laughingstock. it'll be a solid mid-range chip that takes the shine off of the 1366 chips. 30-50% IPC benefits? maybe running specialized code, but do you honestly think we'll see 30% performance gains realized in normal tasks? the IGP is probably irrelevant for at least 50% of all the people at [H], so why are people getting so excited about it? unless I can use it for GPU accelerated programs while having my discrete GPU do visuals, it would just be disabled on my system anyway. I know its exciting to think that your mom's solitaire machine could run crysis at 1024x768 and all...

the best thing SB will do is drive down the price on 1366 boards

 

[SHaFT7]

Not to jump into this argument, but we work in the "motherboard-replacement" space, and we have sold mountains of G4x based 775 boards in both new systems and for mobo replacement. All of the customers, except ones needing dual monitors (dual VGA that is) and hard-core gamers have been satisfied.

In fact, our baseline new machine is STILL a 775 box with a dual core, and we sell more of that than anything else new-machine-wise. Most of the people we are upgrading are still coming from P4's or early dual-cores, and are most pleased with the performance that our baseline machines gives. 775 is indeed still kickin'

 

[R-Type]

Not to jump into this argument, but we work in the "motherboard-replacement" space, and we have sold mountains of G4x based 775 boards in both new systems and for mobo replacement. All of the customers, except ones needing dual monitors (dual VGA that is) and hard-core gamers have been satisfied.

In fact, our baseline new machine is STILL a 775 box with a dual core, and we sell more of that than anything else new-machine-wise. Most of the people we are upgrading are still coming from P4's or early dual-cores, and are most pleased with the performance that our baseline machines gives. 775 is indeed still kickin'

....so you sell clueless buyers an obsolete platform in a new PC? Just because you are still selling it to them as new doesn't mean it hasn't been on the EOL train for over a year.