Complete list of Sandy Bridge processors leaked

[Acanthus]

not excited.

motherboards will be overpriced (1 chipset vendor)

the K processors will also be overpriced

Unless the advantage over AMD is devastating, i will be going AMD.

 

[iCyborg]

No sub 35W is a bummer, but I'm more disappointed that 25W Arrandales are still MIA despite being announced in Jan 2010 with the rest. What happened there?

 

[IntelUser2000]

not excited.

motherboards will be overpriced (1 chipset vendor)

the K processors will also be overpriced

Unless the advantage over AMD is devastating, i will be going AMD.

You'll be probably be going AMD regardless. The motherboard prices have been going down over the years unless you were adamant on obtaining boards with double-digit phase VRs and other features which is questionably expensive.

The K is what? $40 more? The i7 2600 on the earlier rumors had price of $280, which is same price as the i7 860 and isn't the highest clocked chip.

 

[IntelUser2000]

Given how up to now everyone(that is, except Intel) thought that the Sandy Bridge parts would have 6MB L3, I wonder how much of the future roadmaps are true.

For one thing I doubt the existence of a Socket 2011 as a high end desktop platform rather than Socket 1356. Two sites believe the high end desktop will be Socket 1356, not Socket 2011 is PCWatch and 4gamer.net. PCWatch is the first site to leak news that there are dual core Core i5's when every other site believed its a 4 core Hyperthreading disabled version.

The 2P workstation platform that uses Socket 1356, still exists. Why would they not use that for desktop and instead use much more expensive Socket 2011 and 4 channel(useless for desktop) on PC?

 

[Soleron]

Given how up to now everyone(that is, except Intel) thought that the Sandy Bridge parts would have 6MB L3, I wonder how much of the future roadmaps are true.

Some of the L3 is set-aside for the IGP. The 8MB ones set aside 2MB.

For one thing I doubt the existence of a Socket 2011 as a high end desktop platform rather than Socket 1356. Two sites believe the high end desktop will be Socket 1356, not Socket 2011 is PCWatch and 4gamer.net. PCWatch is the first site to leak news that there are dual core Core i5's when every other site believed its a 4 core Hyperthreading disabled version.

The 2P workstation platform that uses Socket 1356, still exists. Why would they not use that for desktop and instead use much more expensive Socket 2011 and 4 channel(useless for desktop) on PC?

Yeah, Socket 2011 is the Nehalem-EX replacement on the very high end.
Socket 1356 is the 2P workstation and high-end desktop platform.
Socket 1155 is the mainstream desktop platform.

It fits a lot better than S-2011 being the high-end.

 

[IntelUser2000]

I can't see Intel investing into a high performance interconnect like the Ring Bus and and THEN doing a fixed allocation of L3 cache for the GPU. It's just not efficient. Dynamic allocation depending on usage is better.

Especially now the specs are said to be "8MB", that means it'll be available to both CPU and GPU.

 

[Acanthus]

You'll be probably be going AMD regardless. The motherboard prices have been going down over the years unless you were adamant on obtaining boards with double-digit phase VRs and other features which is questionably expensive.

The K is what? $40 more? The i7 2600 on the earlier rumors had price of $280, which is same price as the i7 860 and isn't the highest clocked chip.

The cheapest X58 is still $200 for the cheapest "premium" brand (looking at Asus and Gigabyte). How old is x58?

I cannot expect the newer successor to be cheaper than that.

Are the "k" series processors in fact completely unlocked? as in i can set multipliers well beyond what i would need to achieve a reasonable goal?

Most importantly, am i going to have a 25% performance advantage over a comparable overclocked AMD solution? Because the cost difference is going to be that large.

 

[IntelUser2000]

Yea, but the 6 series chipsets coming in early 2011 isn't successor to X58, so it'll be cheaper. If you are talking about the one coming in second half of next year, then I guess you are right.

As for overclocking:

http://www.youtube.com/watch?v=RaHD_ciI4w0

2:06

 

[IntelUser2000]

http://www.expreview.com/11683.html

The same info in page 1 now with GPU clocks.

 

[Acanthus]

Yea, but the 6 series chipsets coming in early 2011 isn't successor to X58, so it'll be cheaper. If you are talking about the one coming in second half of next year, then I guess you are right.

As for overclocking:

http://www.youtube.com/watch?v=RaHD_ciI4w0

2:06

the k-series are the "partial overclocking" on that list right?

that is my question... if intel artificially limits overclocking by only giving us 2x over stock, then the k series is shit for overclockers, especially with a price increase.

 

[ydnas7]

from the video
fully and partially unlocked processor available
non XE fully unlocked supports up to 57 bins.
non XE partially unlocked supports up to TDP +TBD bins.

i guess that 1 bin = 100mHz so the range is 0 - 5.7Ghz on the 'K' versions - if you're lucky/skilled

 

[IntelUser2000]

the k-series are the "partial overclocking" on that list right?

that is my question... if intel artificially limits overclocking by only giving us 2x over stock, then the k series is shit for overclockers, especially with a price increase.

The "K" series are just like the "K" series today. Unlimited ratios. The slide says 57 bins but the slide before that says there's no limit, so I'm understanding it as "57 bins above stock". It's not really "unlimited" as they say but the potential to reach 8GHz is good enough, no? Think of it, why would they give you an option to clock it at 100MHz anyway? They had similar wording for the earlier Extreme Editions.

The non "K" series are the ones that's called "Partially Unlocked". Those are the ones that do "few bins above Turbo". It doesn't look like there will be any "Full unlocked versions".

On some unrelated topic, there are rumors that the mobile version of Sandy Bridge will work with the discrete GPU to enable additional monitor support.

 

[PlasmaBomb]

I doubt it, but you can probably look forward to seeing the Ivy Bridge 22nm parts named i5/i7 3520M -> 3920QM.

Naming things isn't exactly what Intel does best.

It was sarcasm... and you should know by know, if you think you can predict future Intel naming conventions, they will probably change them...

 

[PlasmaBomb]

Interesting how the L3 caches aren't 6MB anymore.

erm... Arrandale doesn't have 6MB L3 caches... (that's the 720/740QM)

http://en.wikipedia.org/wiki/Arrandale_(microprocessor)

It said that 20 percent gain was not on a single app but geometric mean, which is slightly worse than adding all numbers and dividing them.

By the way, the Sandy Bridge mobile chip that was demoed running Cinebench is Core i7 2720QM.

I think I should point out that this "slightly worse" (geometric vs arithmetic mean) is better for us (consumers).

Performance increase -

App 1 = 10%
App 2 = 20%
App 3 = 30%

Geometric mean increase ~ 18%
Arithmetic mean increase = 20%

 

[Fjodor2001]

Any idea what performance can be expected from the Sandy Bridge IGP?

I have a discrete nVidia 7600GS now. Is the Sandy Bridge IGP likely to be faster or slower compared to that?

 

[IntelUser2000]

erm... Arrandale doesn't have 6MB L3 caches... (that's the 720/740QM)

You missed later comments I posted.

I was talking about how earlier rumors had the cache at 6MB L3 for Sandy Bridge.

Any idea what performance can be expected from the Sandy Bridge IGP?

I have a discrete nVidia 7600GS now. Is the Sandy Bridge IGP likely to be faster or slower compared to that?

The 7600GS looks 20-30 percent faster than the 890GX/fastest HD Graphics. The GPU in Sandy Bridge will be at least 2x better so in general it should be faster. Intel's drivers suck in OpenGL though, they need to improve on this a lot.

Looks like the GPU will be 2x faster than the previous gen, on the mobile side.

It looks like you need the Core i7 2600 to get the fastest GPU as well. That's going to be conflicting because the HKEPC videos say you need P67, not H67 to get the unlocked multipliers.

 

[Edrick]

For one thing I doubt the existence of a Socket 2011 as a high end desktop platform rather than Socket 1356. Two sites believe the high end desktop will be Socket 1356, not Socket 2011 is PCWatch and 4gamer.net. PCWatch is the first site to leak news that there are dual core Core i5's when every other site believed its a 4 core Hyperthreading disabled version.

The 2P workstation platform that uses Socket 1356, still exists. Why would they not use that for desktop and instead use much more expensive Socket 2011 and 4 channel(useless for desktop) on PC?

But wouldn't it be cheaper for Intel to support only 2 sockets instead of 3?

 

[IntelUser2000]

But wouldn't it be cheaper for Intel to support only 2 sockets instead of 3?

That's true, but they all exist anyway so why use the most expensive one? It's not like the profit margin will be better unless they want to have $2k CPUs and $500 motherboards.

 

[bamacre]

But if they sell the 3.1-3.4GHz quad at $200 then maybe it won't matter.

LOL, I was looking at that, too. And if it's priced at $200, it looks like a winner in my eyes. When are these supposed to be released?

 

[Edrick]

That's true, but they all exist anyway so why use the most expensive one? It's not like the profit margin will be better unless they want to have $2k CPUs and $500 motherboards.

Well, not only is Intel currently supporting 3 sockets, but also 3 CPU lines (1 for each socket). Maybe they realized that supporting 3 different playforms is not the cost efficient way, and are looking to change that with SB. Why have different CPUs with 2-channel, 3-channel, and 4-channel MCs for example? (Nehalem already proved that 3-channel is not worth the cost in most cases.)

Cutting down to 2 lines instread of 3 saves developent money as well QA money. I am really hoping that socket 2011 does replace the X58 platform and the costs will be on par with X58.

 

[IntelUser2000]

Maybe, there's two 2011 sockets instead. The cheaper version with 3 channel DDR and server version with 4 channel DDR. Supporting more memory channels increase routing required which necessitate more expensive 6 and 8 layer motherboards.

Somewhere along here is the truth. Plans might change but its hard for an outsider to know.

 

[khon]

Looking it this list I know I should buy the i5-2400, its perfectly adequate for my needs, and very reasonably priced at $180. Yet I can't help but look at that i7-2600K and wonder if I could take it up to 4.5GHz with an H70 cooler ......

 

[PlasmaBomb]

You missed later comments I posted.

I was talking about how earlier rumors had the cache at 6MB L3 for Sandy Bridge.

Yeah I was a bit behind in the thread...

Do you have a link to the 2720QM that was demoed with cinebench?

 

[RussianSensation]

Complete list for LGA1155 Sandy Bridge is what's meant indeed.

I used to be very enthousiastic about these cpu's, planning to upgrade as soon as they came out. But as it turns out, the chipset doesn't really offer anything compelling (no full sata III, no usb3, still 16 pci-e 2.0 lanes) and if the no-overclocking news is true, I don't really see the point anymore.
My i5 750 already runs faster than the i5 2500 i was interested in (3.33-3.98GHz).

Unfortunately we knew for a while that 1155 was neutered from the start (much like 1156 was). If SB is at least 10% faster per clock than the Bloomfield, then the i5 2400 @ 3.1ghz will be = i5 750 @ 3.4ghz (although there is the 6mb of cache vs. 8mb of cache). Still that's quite a massive performance boost from a stock 750 for new buyers (we don't know the IPC for SB though). We'll have to see on the performance front -- most games are still GPU limited and I don't really see many upgrading from 3.8-4.0ghz i7s until 2011 arrives.

 

[khon]

Two things I'm wondering about:

#1. If you overclock the CPU will you effectively be downclocking the GPU ?
This might seem a bit odd, but as far as I can tell it might be the case. As we know all parts of the processor share the same baseclock, so overclocking the CPU will have to be done by increasing the multiplier, which won't affect the GPU. However, if you overclock the CPU you will be operating above the TDP most of the time, which means the GPU turbo won't be able to activate, thus effectively slowing down your GPU.

#2. Can you overclock in steps smaller than 100MHz ?
The GPU shares the 100MHz baseclock with the CPU, yet it uses speeds that aren't divisible by 100MHz, which means it appears to have a multiplier that isn't a whole number. If the same holds true for the CPU it might be possible to overclock in steps smaller than 100MHz, since you might be able to increase the multiplier in steps of 0.5 or even 0.1