Sandy Bridge Reviews

[tech960]

What? Noone noticed this?

I thought the road map stated 8 core in Q3. I was waiting for that, but I think I'll just pull the trigger on a 2600k an OC to 4.5ghz on air. I have a IP35 Pro now, am I going to loose my raid storage due to the new controller ?

 

[frostedflakes]

I hate how Intel does stuff like this. They need to take a page from AMD's book on this and keep it simple, much less confusing for consumers.

Other than the usual Intel shenanigans like this, Sandy Bridge looks really, really good. Poor AMD is going to have trouble keeping up.

 

[VirtualLarry]

I hate how Intel does stuff like this. They need to take a page from AMD's book on this and keep it simple, much less confusing for consumers.

Other than the usual Intel shenanigans like this, Sandy Bridge looks really, really good. Poor AMD is going to have trouble keeping up.

WTF!!! No VT-d on the 'K' models, when it is present on the non-K versions? What kind of BS is that? K editions should be supersets.

 

[sihv]

WTF!!! No VT-d on the 'K' models, when it is present on the non-K versions? What kind of BS is that? K editions should be supersets.

Agreed. I was pretty annoyed when I read about this. I was hoping to grab the 2600K and give Xen's VGAPassthrough a try so I could run games in a virtualized Windows. If that were to work, I could keep Linux as my primary OS and get rid of having to dual boot.

If I settle with the non-K 2600, does the choice of MB or chipset (P67 or H67) still matter? I haven't found any references to VT-d in motherboard specifications from several manufacturers. I even downloaded manuals to see if it's mentioned in mobos' BIOS settings and all I saw was a vague "virtualization technology" toggle.

Getting all the needed features is already ridiculously complicated even without having to wade through a minefield of crippled motherboards or buggy bioses. According to the Xen wiki, finding one with a working VT-d implementations isn't that rosy even with chipsets that supposedly should work: http://wiki.xensource.com/xenwiki/VTdHowTo

 

[adamantinepiggy]

Just saw this: http://www.theinquirer.net/inquirer/news/1934536/intels-sandy-bridge-sucks-hollywood-drm. I don't take much stock with the Inquirer, buthave not seen any mention of DRM in Sandy Bridge CPU's anywhere else.

Intel's Sandy Bridge sucks up to Hollywood with DRM

Locked-down PCs to watch flicks
By Nick Farrell
Mon Jan 03 2011, 11:01



CHIPMAKER Intel has cut a deal with Hollywood in its Sandy Bridge chips to put digital restrictions management (DRM) in them.
Intel's sellout to the big media companies makes it 'safer' for Hollywood studios to offer premium movies to consumers on locked-down personal computers. No doubt the media content cartels and Intel are dreaming monopolistic visions of extracting big rents to stream everything online.
According to Reuters, Time Warner's digital distribution unit and other studios plan to offer high-definition movies to consumers who have PCs with Intel Sandy Bridge chips inside, simultaneous with releasing them on DVD.
Apparently the media cartels are dead keen to get their content on PCs that are locked-down even to the level of the chip.
Mooly Eden, Intel's VP and general manager of the PC client group said that Intel's embedded DRM allows content to be streamed to computers with Sandy Bridge chips and that Chipzilla is making deals with all the studios and content distributors to make it available.
So while punters will get saddled with DRM on their computers, Intel thinks that it can suck them in and buy them off with improved multimedia processing.
Intel's Sandy Bridge chip designs combine central processing and graphics functions on the same piece of silicon, making them faster, more energy efficient, and likely more profitable.
But Chipzilla is taking a big gamble. As with Intel's earlier, aborted scheme to put a serial number in every CPU chip, it is possible that consumers will refuse to buy the chips because of the DRM lock-down.
It is questionable whether Hollywood making available paid-for content through its toll-booth outlets will be seen as an advantage or a drawback, and will on balance attract the PC punters. µ

 

[Makaveli]

This is interesting would love to read more about this.

 

[shabby]

Just saw this: http://www.theinquirer.net/inquirer/news/1934536/intels-sandy-bridge-sucks-hollywood-drm. I don't take much stock with the Inquirer, buthave not seen any mention of DRM in Sandy Bridge CPU's anywhere else.

Thats because its disabled on p67, its only enabled on the lower end chipsets.

 

[soccerballtux]

6mb of cache and lack of hyperthreading makes this one unappealing imo.



1.45V for 24/7 at load operation on air cooling is not what I would consider safe for 32nm process.

My 920 runs at 1.41v, so it has to be fine riiiiiiight?

Your 920 is not 32nm

QFT.

 

[cusideabelincoln]

What? Noone noticed this?
It’s codenamed Sandy Bridge-E and it’ll debut in Q4 2011. The chips will be available in both 4 and 6 core versions with a large L3 cache (Intel isn’t being specific at this point).

I noticed that. Perhaps Intel is doing more to differentiate the two platforms again? 1156 doesn't get six cores while 1366 does. 1155 looks like it will get six cores. So I guess 2011 should have some perks, perhaps even more perks? Honestly it really should given the price difference it will probably be.

Most early rumors point to Bulldozer being 15-20% faster per clock than Phenom II. Therefore, AMD is no position at all to make the performance leap that Core 2 Duo did over Pentium D / Athlon 64.

LOL, you fail at reading comprehension. It's nice you ignored the previous sentence, which is IMPORTANT to the statement I made: "Yep, Athlon 64 was just as much of a jump over Pentium D as Core 2 was over Athlon 64."

AMD is in the position to make such a leap simply because they are behind Intel much the same Intel was behind AMD in the Pentium D era. Oh... and this says it best:

It wasn't Pentium D-->Core 2, it was Athlon 64 X2-->Core 2. Intel Pentium D wasn't the top chip of the day, AMD's chips were.

Do I need to explain this further? Actually I've already explained it and provided data in my other posts in this thread, including the one you quoted. I was never suggesting they are actually on track to do it.

Interesting. So you think that going from Pentium D to E6600 2.4ghz dual core in 2006 to Q6600 2.4ghz in 2007 to Core i7 920 in 2008 is comparable to going from a 2008 Core i7 920 to a 2600k in 2011? Intel is improving, at a snail's pace in the last 2 years though. I am sure most 2008 Core i7 920 owners won't bother at all with the 1155 platform. You couldn't say that to the E6600 S775 owner in 2006 when Core i7 920/Core i7 860 arrived.

...

Q6600 launched in the first half of 2007. The 920 launched at the end of 2008. The 2600K launches at the very beginning of 2011. The time gap is not as big as you are making it.

The performance jump is there. In fact... let me bring back one of your wishes

The bottom line is in 2011 a CPU should be 2-3x faster than a $300 CPU in 2007, which SB will be nowhere near.

And the benchmarks...
http://www.anandtech.com/show/4083/...-i5-2600k-i5-2500k-and-core-i3-2100-tested/15

Photoshop: 11.3s vs 25.6s = 2.25x faster
WME 9: 20s vs 40s = 2x faster
x264 1st: 106.4fps vs 53.8fps = 2x faster
x264 2nd: 36.3fps vs 15.2fps = 2.4x faster
3dsmax 9: 20.1 vs 9.6 = 2x faster
Cinebench single: 5991 vs 2778 = 2.15x faster
Cinebench multi: 22875 vs 9681 = 2.35x faster
POV Ray: 4875 vs 2192 = 2.2x faster
Par2: 17.3s vs 47.6s = 2.75x faster
Winrar: 49.6s vs 140.7s = 2.8x faster
Sorenson: 72.7s vs 179s = 2.45x faster
Excel Monte: 11.1 vs. 22.1 = 2x faster

And before you quote what I've already just said and take things out of context and switch contexts on the fly all willy nilly and invalidate your arguments, here's one of the most important points: THESE ARE JUST THE LOW END AND MIDRANGE CHIPS! The high end stuff doesn't come out until later this year. You will see your six and probably eight cores. And socket 2011 will bring even higher IPC too boot. The performance is here and it will also be here.

Just look at the 2500 and 2600 chips. They are bringing the performance of the fastest chips of "last generation" at a cheaper price point, and lower thermals.

Newegg has Core i7 950 3.06ghz for $290. So if anything 2600k should be compared to Core i7 870 2.93ghz or Core i7 950 3.06ghz at minimum. The performance difference is not commensurate with a 2011 new generation CPU compared to a 2008 architecture. Once Bulldozer ships 6- and 8-core processors, I hope we will quickly see Intel release 6-core SB at more reasonable prices because this idea of selling a $300 4-core processors in 2011 is just Intel nonsense (which I am sure shareholders love).

Ok, so this is where you change your goalposts (i7 870 launched in Q3 '09 and the 950 Q2 of '09; the 950 was also much costlier at 500 bones.) within the same argument. LOL, whatever. The 2600K is still faster than both of those chips, by a good margin.

 

[ilkhan]

New CPU vs old CPU (for me). Then increas the 2500K performance by 40% (4.6Ghz) for how it'll actually be.
Impressive. http://www.anandtech.com/bench/Product/288?vs=52&i=2.4.5.25.27.28.38.39.46.47.48.49.50

 

[CosmicMight]

And the benchmarks...
http://www.anandtech.com/show/4083/t...2100-tested/15

Photoshop: 11.3s vs 25.6s = 2.25x faster
WME 9: 20s vs 40s = 2x faster
x264 1st: 106.4fps vs 53.8fps = 2x faster
x264 2nd: 36.3fps vs 15.2fps = 2.4x faster
3dsmax 9: 20.1 vs 9.6 = 2x faster
Cinebench single: 5991 vs 2778 = 2.15x faster
Cinebench multi: 22875 vs 9681 = 2.35x faster
POV Ray: 4875 vs 2192 = 2.2x faster
Par2: 17.3s vs 47.6s = 2.75x faster
Winrar: 49.6s vs 140.7s = 2.8x faster
Sorenson: 72.7s vs 179s = 2.45x faster
Excel Monte: 11.1 vs. 22.1 = 2x faster

And before you quote what I've already just said and take things out of context and switch contexts on the fly all willy nilly and invalidate your arguments, here's one of the most important points: THESE ARE JUST THE LOW END AND MIDRANGE CHIPS! The high end stuff doesn't come out until later this year. You will see your six and probably eight cores. And socket 2011 will bring even higher IPC too boot. The performance is here and it will also be here.

Just look at the 2500 and 2600 chips. They are bringing the performance of the fastest chips of "last generation" at a cheaper price point, and lower thermals.


Quote:
Originally Posted by RussianSensation
Newegg has Core i7 950 3.06ghz for $290. So if anything 2600k should be compared to Core i7 870 2.93ghz or Core i7 950 3.06ghz at minimum. The performance difference is not commensurate with a 2011 new generation CPU compared to a 2008 architecture. Once Bulldozer ships 6- and 8-core processors, I hope we will quickly see Intel release 6-core SB at more reasonable prices because this idea of selling a $300 4-core processors in 2011 is just Intel nonsense (which I am sure shareholders love).

Ok, so this is where you change your goalposts (i7 870 launched in Q3 '09 and the 950 Q2 of '09; the 950 was also much costlier at 500 bones.) within the same argument. LOL, whatever. The 2600K is still faster than both of those chips, by a good margin.
__________________

Nice points. I've never understood the "but it's not that much better than 1366!!!11!!one!!!" crowd.

Those of you on 1366 who use your comp for gaming/general tasks don't really need SB. Hell, you don't even need a 1366 for that matter, there were plenty of benches where the AMD chips were still competitive with the 2500k. Furthermore, if you are happy with your computer, don't upgrade.

But debate away.

 

[edplayer]

The performance jump is there. In fact... let me bring back one of your wishes

The bottom line is in 2011 a CPU should be 2-3x faster than a $300 CPU in 2007, which SB will be nowhere near.

And the benchmarks...
http://www.anandtech.com/show/4083/...-i5-2600k-i5-2500k-and-core-i3-2100-tested/15

Photoshop: 11.3s vs 25.6s = 2.25x faster
WME 9: 20s vs 40s = 2x faster
x264 1st: 106.4fps vs 53.8fps = 2x faster
x264 2nd: 36.3fps vs 15.2fps = 2.4x faster
3dsmax 9: 20.1 vs 9.6 = 2x faster
Cinebench single: 5991 vs 2778 = 2.15x faster
Cinebench multi: 22875 vs 9681 = 2.35x faster
POV Ray: 4875 vs 2192 = 2.2x faster
Par2: 17.3s vs 47.6s = 2.75x faster
Winrar: 49.6s vs 140.7s = 2.8x faster
Sorenson: 72.7s vs 179s = 2.45x faster
Excel Monte: 11.1 vs. 22.1 = 2x faster

Finally! Some closure.


There are a lot of people on this forum that know a good amount about computer performance. Russian is definitely one of them. Prior to this thread I would have said that you could always count on him to be objective and impartial but something happened and he snapped. Starts twisting everything to make SB appear less than it is. Don't know what did it, maybe because he owns an i7?

Even one of the mods is talking all crazy about SB saying its absolute bullshit and even trying to bet people that S1366 will outsell SB! What have you done Intel, What have you done!

 

[notty22]

Finally! Some closure.


There are a lot of people on this forum that know a good amount about computer performance. Russian is definitely one of them. Prior to this thread I would have said that you could always count on him to be objective and impartial but something happened and he snapped. Starts twisting everything to make SB appear less than it is. Don't know what did it, maybe because he owns an i7?

Even one of the mods is talking all crazy about SB saying its absolute bullshit and even trying to bet people that S1366 will outsell SB! What have you done Intel, What have you done!

Same thing happens in every sector, Chevrolet releases a new Corvette with a extra 100 hp from the year before, and proud owners of the prior model will find fault. 'its 12 1bs heavier,' lol
Head over to SA, for the MAD magazine , rant of the year against SB.
Sandy Bridge is the biggest disappointment of the year

A rant

by Charlie Demerjian

 

[AtenRa]

The most impressive SB CPU is the i3 2100 IMHO. For $117 a dual core + HT CPU that can compete with Quad core i5 and Phenom II X4.

Just imagine what this CPU could do with unlock multi, or what damage it could do to AMDs Llano CPUs if it had all 12 EUs enabled.

Does anyone knows if Dual core SBs are Quads with 2 cores disabled or Dual core die ??

 

[Tsavo]

The most impressive SB CPU is the i3 2100 IMHO. For $117 a dual core + HT CPU that can compete with Quad core i5 and Phenom II X4.

Just imagine what this CPU could do with unlock multi, or what damage it could do to AMDs Llano CPUs if it had all 12 EUs enabled.

Does anyone knows if Dual core SBs are Quads with 2 cores disabled or Dual core die ??

I'd wager they are dual core dies.

 

[IntelUser2000]

Does anyone knows if Dual core SBs are Quads with 2 cores disabled or Dual core die ??

Not only there is a seperate dual core die, there seems to be different one for 6EU and 12EUs that further seperate the dual core dies into two.

That is interesting, especially since IBM has its own architectures (Power7 and z196).

Maybe because their own chips are low volume chips and Sandy Bridge is a very high volume chip. When they released the monstrous 140,000 CPU supercomputer few years ago, they were using their own chips, but were few years old at least.

I noticed that. Perhaps Intel is doing more to differentiate the two platforms again? 1156 doesn't get six cores while 1366 does.

The strange thing isn't the differentiation, but that it will only get up to 6 cores, contrary to previous rumors. Possible explanation for that is that only the server platforms get 8 cores and they'll get 8 core for enthusiasts with Ivy Bridge.

 

[Tsavo]

Not only there is a seperate dual core die, there seems to be different one for 6EU and 12EUs that further seperate the dual core dies into two.

Two dual cores or dual dual cores? :|

 

[mosox]

Regular user:
1.Small stuff:
- Internet, winamp, movies, whatnot: any dual core is good for that.
2.Gaming:
- Good CPU (quad is best for some games)
- The best video card you can afford.

You're on a budget:

What do you do? Buy a $400 Sandy Bridge CPU/mobo and a cheapo video card or buy an Athlon II X4 + mobo ($200) and a much better video card? For your other uses the Athlon is overkill already. Do you encode? The 2500K is twice as fast? So what? Can't you leave the thing do its stuff and mind your business (watch TV, sleep or something)? Are you a professional? Buy the 2500K then. Buy two Xeons, that's much better.

Another socket limited to a handful of CPUs? 6 cores on this socket? 8 cores? Nope.
A regular user (say, gamer) needs:

- Cheap quads
- Cheap good mobos
- Very good video card.

What do you do, buy a $400 CPU/mobo and a $150 video card or buy a $250 CPU/mobo and a $300 video card?

The current CPUs are already overkill for a regular user. The video cards? Not so much.

If I pay $400 for a socket, better give me anything on it. Cheap dual cores to fast 6/8 cores. 50 CPUs at least.

 

[IntelUser2000]

Two dual cores or dual dual cores? :|

LOL, does it sound like the latter?

Quad Core = 216mm2
Dual core 12EU 4MB L3 = 149mm2
Dual core 6EU maybe 3MB L3 = 131mm2

The fact that there are different numbers for the dual core variants suggest its a seperate die.

 

[AtenRa]

Not only there is a seperate dual core die, there seems to be different one for 6EU and 12EUs that further seperate the dual core dies into two.

I dont believe they have two dual core dies.

 

[Tsavo]

LOL, does it sound like the latter?

Quad Core = 216mm2
Dual core 12EU 4MB L3 = 149mm2
Dual core 6EU maybe 3MB L3 = 131mm2

The fact that there are different numbers for the dual core variants suggest its a seperate die.

I like walking into a donut shop and asking for 6 of one and half a dozen of another. Some have had to stop to ponder.

 

[IntelUser2000]

I like walking into a donut shop and asking for 6 of one and half a dozen of another. Some have had to stop to ponder.

I must try that when I walk into Tim Hortons next time. Hmm, the only problem is I'd need to want a LOT of donuts.

 

[ydnas7]

LOL, does it sound like the latter?

Quad Core = 216mm2
Dual core 12EU 4MB L3 = 149mm2
Dual core 6EU maybe 3MB L3 = 131mm2

The fact that there are different numbers for the dual core variants suggest its a seperate die.

There will be another Quad core to the one above, as Mooly Eden had earlier explicitly mentioned a 1.12billion (or so) transistor SB in asia halfway through the year.

so if this is 4 separate dies, then it is very aggressive, Pentium (and Celeron) Sandybridge will be serious competitors for the non-overclocking low end. (dell - etc)

 

[ydnas7]

oouch
http://www.anandtech.com/bench/Product/288?vs=203

2 core or 6 core?
http://www.anandtech.com/bench/Product/289?vs=203
(once encode is done by SB encode engine, expect reversal of graphs for encoding)
ok so this was a troll

but these are similar market points
i3 (131mm2 @32nm inc IGP) vs athlon x4 (169mm2 @45nm exc IGP)
http://www.anandtech.com/bench/Product/289?vs=188

 

[Castiel]

oouch
http://www.anandtech.com/bench/Product/288?vs=203

2 core or 6 core?
http://www.anandtech.com/bench/Product/289?vs=203

(once encode is done by SB encode engine, expect reversal of graphs for encoding)

Yup. Now a simple quad core without HT can beat up on Amd's Flagship 6 core