Sandy Bridge write-ups

[Sabre]

Toms hardware: Intel’s Second-Gen Core CPUs: The Sandy Bridge Review
Toms hardware: Efficiency Comparison: Sandy Bridge Vs. Intel And AMD CPUs
Toms hardware: The Man Behind 'Sandy Bridge'

PC Perspective: Intel Core i7-2820QM Mobile Sandy Bridge Performance Review
PC Perspective: Intel Core i7-2600K (and friends) Sandy Bridge Processor Review

Arstechnica: Sandy Bridge arrives from Intel with up to 50% performance boost

AnandTech: The Sandy Bridge Review: Intel Core i7-2600K, i5-2500K and Core i3-2100 Tested
AnandTech: Intel’s Sandy Bridge: Upheaval in the Mobile Landscape





CN's:

Check out:
Turbo boost 2.0: Can raise a single core to high speeds till a thermal load is reached
QuickSync: Vastly improved encoding/decoding for video

Overclocking isn’t handled well at all. Really, the only viable option for power users is a K-series SKU.

The graphics situation, at least on the desktop, is also pretty whacky. Of the 14 models introduced at launch, the two best suited to enthusiast-oriented gaming machines with discrete GPUs are the ones armed with Intel’s HD Graphics 3000 engine. The other 12—conceivably candidates for more mainstream gaming builds, office desktops, and HTPCs—sport the downright average HD Graphics 2000 implementation.

Those two gripes out of the way, how could we not be impressed by Sandy Bridge’s performance? Existing Lynnfield- and Clarkdale-based processors already offer strong performance compared to AMD’s lineup. Significant gains, clock-for-clock, compound in the face of notable frequency increases across the board (thanks to a mature 32 nm process), giving Sandy Bridge an even more commanding position.

I’m also a big fan of Quick Sync. Neither AMD nor Nvidia have an answer to Intel’s decode/encode acceleration, and they’re not expected to any time soon. If you do a lot of video editing or transcoding, an upgrade to Sandy Bridge might be warranted based solely on the time you’ll save by virtue of this feature.

Intel's new Sandy Bridge processor architecture and H57 platform are impressive both in terms of performance and power efficiency. Together, they easily outperform the competition in terms of performance per watt. That include the existing Core i3/i5/i7 lineup employing LGA 1156 and AMD's fastest Socket AM3 product portfolio.

If there was one Sandy Bridge-based SKU that I’d personally recommend to friends and family building new PCs, it’d be the Core i5-2500K. Its performance relative to AMD’s lineup and the rest of Intel’s stack is noteworthy—especially given its price tag just north of $200. The i5-2500K circumvents Sandy Bridge's overclocking challenges with an unlocked multiplier, and I'm counting on gamers to drop it onto a P67-based motherboard, skirting the integrated graphics debate entirely.

Oh dear god do I want to buy a new PC and throw one of these in it... but I'm trying my best to wait to see how Bulldozer performs... whenever it comes out!

[Raven]

I'm using a 2 year old Core 2 quad and still have way more CPU power than I need.

[PGT]

what's the new architecture with security stuff embedded?

[Sabre]

Sandy bridge does have a bunch of logic dedicated to encryption and such (if that's what you're talking about). Look up AES-NI. I know I saw a performance graphic in one of those articles, but I can't find it right now I'm also very interested in the UEFI support. I'm shocked that they did not include USB 3.0 or more than 16 lanes on the PCIe bus though. Hopefully with the Z68 or the X78 we will see this.

[complacent]

i was surprised as well. the rivalry might be back on!

[Sabre]

I've always like AMD... even way back when they were making 486 clones I really do hope Bulldozer comes our ahead in this race! Whatever does happen, one thing good definitely has: Intel has had to step up their game, both in speed and price. Without AMD, we'd probably be paying twice to three times what we do for processors that are not nearly as fast.

[Raven]

I had a Pentium 3, then a Pentium 4 (northwood, the good one) and then I jumped ship to the dual core AMDs and stayed there for many CPUs until a friend of mine bought a Core 2 quad that wouldn't fit in his computer. I bought it off him half to help him out and half to upgrade. I'll probably hang with this system until a even faster memory standard emerges. I'll probably go back to AMD I liked supporting the under dog. Same reason I always buy Radeon cards.

[Sabre]

BTW, that brings up a good point... the memory interface is actually slower on the new chips.

[Raven]

You sure, every new DDR standard has felt faster than the last.

[Sabre]

You sure, every new DDR standard has felt faster than the last.

Yep

Sandy Bridge - Dual channel DDR3-1333
Nehalem - Triple channel DDR3-1066

Interestingly enough, the actual memory connect is faster... but they don't have the memory to keep it full There is one note that gives me hope "Improved memory controller with maximum 25.6 GByte/s bandwidth supports DDR3-1600 dual channel RAM and two load/store operations per cycle". Hopefully the two load/store ops/cycle will double the bandwidth!

[Raven]

Sandy Bridge - Dual channel DDR3-1333
Nehalem - Triple channel DDR3-1066

Oh, triple channel channel ram is a scam. It's really not faster than dual, which in turn isn't that much faster than single channel ram. I have a theory that dual channel memory controllers were created to help shift more memory sticks.

[Sabre]

^^^ Negative ghost rider... The memory controllers are interleaving the RAM, so you are increasing speed. Take a lookie:

Intel's write up on it

High and Low values taken from Wikipedia:

Code: Select all

Device	                                         Rate (bit/s)	Rate (byte/s)
PC-800 RDRAM (single-channel 64-bit)              12.8 Gbit/s	  1.6 GB/s
PC-4000 DDR-SDRAM (single channel 64-bit)         34.3 Gbit/s	  4.29 GB/s
PC2-6400 DDR2-SDRAM (single channel 64-bit)       51.2 Gbit/s	  6.4 GB/s

PC-1200 RDRAM (dual-channel 128-bit)              38.4 Gbit/s	  4.8 GB/s
PC-4000 DDR-SDRAM (dual channel 128-bit)          67.2 Gbit/s     8.4 GB/s
PC2-8500 DDR2-SDRAM (dual channel 128-bit)       136.0 Gbit/s	 17 GB/s

PC3-8500 DDR3-SDRAM (triple channel 192-bit)     204.6 Gbit/s	 25.6 GB/s
PC3-17600 DDR3-SDRAM (triple channel 192-bit)    422.4 Gbit/s	 52.8 GB/s

Actually, easier to see in a graph

[Raven]

First, that graph is hideously flawed due to the way they mixed different memory standards. The triple channel ram is the only one that is DDR3.

Second, those numbers are theoretical peak numbers. Real world numbers are much lower. Back when I tested and played with such things, I discovered that single ram would run at about 90% of its theoretical value, but dual channel could only manage 75% of its theoretical. So where is the bandwidth going? Overhead. Having the query two sticks of ram to get one piece of info takes it's toll on real world throughput. Triple channel RAM will suffers from this even more. Yes, they're faster, but not as fast as you think.

http://www.insidehw.com/Reviews/Memory/ ... age-2.html

This link shows virtually no difference between the dual and triple. This illustrates the other reason triple channel is silly. Basically no matter how fast the ram is, eventually you hit the limit of the bus. I have the same problem on my Core 2 system, the difference between single and dual channel on my computer is slight. 6.5 GB/s is simply all I'm going to get through a 1333MHz FSB.

[Sabre]

lol, I build that graph... if you notice, I included DDR and DDR2 in the first two groups, but couldn't graph them for triple channel because no one does triple channel DDR, DDR2, or RDRAM).

Originally I posted this... here it is with the theoretical values:
Sandy Bridge - Dual channel DDR3-1333 - PC3-10600 DDR3-SDRAM (dual channel 128-bit) - 170.6 Gbit/s
Nehalem - Triple channel DDR3-1066 - PC3-8500 DDR3-SDRAM (triple channel 192-bit) - 204.6 Gbit/s

So they have indeed dropped the theoretical bandwidth by 34 Gbit/s.

That being said, more bandwidth doesn't always mean more speed. There are very few apps that can use all the bandwidth that dual channel can provide let alone triple. Ram speed and latency will make more of a difference in most of todays apps than bandwidth which is why most people believe triple channel is a waste. In a head to head comparison, there was only a 2% difference in todays apps. What I'm trying to get at is that Sandy Bridge was supposed to be "tomorrows" architecture. While todays apps don't take advantage of the bandwidth, I think tomorrows apps might. They should have built a bigger bus to take advantage of the triple channel memory (bigger than 1333 that is).

lol, in the end it doesn't matter... the new chips are damn fast

[Raven]

I bet the triple channel was an experiment that just didn't pan out like they thought it would and thus is getting dropped.

[Sabre]

I bet the triple channel was an experiment that just didn't pan out like they thought it would and thus is getting dropped.

lol, quite a possibility!

[Sabre]

Origin PC over clocks new processor to 5Ghz

[Sabre]

Toms article

Heise found a recent presentation that provides details about the Panther Point chipset series as well as its USB 3.0 integration. The presentation is apparently targeted at mainboard manufactureres as it is positioned as an effective appetizer to advertise the benefits of 5 Gbps USB 3.0 versus the current 480 Kbps USB 2.0.

According to Intel, Ivy Bridge and Panther Point boards should integrate four USB 3.0 ports. The technology will support Windows 7 and Windows 8, whiel Windows XP systems will run USB 3.0 ports as USB 2.0 interfaces via a native Windows EHCI driver.

There is no information when Intel's USB 3.0 integration will be available, but Heise speculates that Panther Point could be announced as series 7 chipsets at CES 2012.

Why this wasn't integrated into the first gen chips for Sandy Bridge, I have no idea...

[Sabre]

Interesting....
Sandy bridge OC'ing

This article doesn’t set out to look at the highest clock rate you can achieve on a Sandy Bridge-based processor. For that, we'd want to use more aggressive cooling, higher voltages, and generally throw the entire efficiency story out the window completely. At the high end, existing BIOSes support 5700 MHz through a 57x multiplier and a bit more through modest BCLK increases. This is the top end for now, though Intel's engineers have told us that they intend to edge the multiplier limits up slightly.

Practically, though, you should be able to reach anywhere between ~4.5 and roughly 5 GHz on air cooling with all Core K-series processors based on the 32 nm Sandy Bridge architecture.

It certainly didn’t take this article to find that Sandy Bridge overclocks very well, at least as long as we’re talking about the K-series processors in Intel's Core i5/i7 portfolio. Going beyond 4 GHz is easily possible, even without a voltage increase, and our processor sample scaled reliably all the way up to 5 GHz on the standard Intel cooler.

Shocking performance is coming from these damn chips! I'm REALLY hoping that Bulldozer is better though...


That being said, it looks like they have run into a problem: Intel hit with chipset design flaw in Sandy Bridge rollout

[complacent]

that recall is expected to cost north of 300 million dollars. oof.

correction. the cost is more like 700 million.

ouch.

[Sabre]

Updated roadmaps




Source

[complacent]

i gotta ask... why the are usb 3.0 and thunderbolt support not already included?!?!?!

seriously. why? i don't get it.

[Sabre]

I think that everyone has been asking themselves that question. It takes quite awhile to come up with these designs though, and they probably were not even sure those technologies would take off, so they just left it out figuring they could add it later At least that's my opinion, I could be way off, haha.