Sandy Bridge i7 & i5 Reviews

[SpuTnicK]

Yet another Chinese review:
Source
Google translate

[kl0012]

Beating the 980x by that amount!

yeah right... i can make tons of home-made charts like this.

[informal]

yeah right... i can make tons of home-made charts like this.

In that particular benchmark it is actually faster.

[kl0012]

In that particular benchmark it is actually faster.

proof?

[informal]

proof?

It's Phoronix open test suit(runs under linux).Results are summed up here.

[kl0012]

It's Phoronix open test suit(runs under linux).Results are summed up here.

amdzone as an source of information - it is not even funny. just check the results - different versions of test suites, compilers and benchmarks.

[-=DVS=-]

Well it all comes down to encryption used, if you use industry's standard AES encryption , Intel has build in hardware acceleration, nothing even comes close to speed gains, if i am not mistaken all Intel 32nm CPUs will have AES acceleration.

A 2xCore i5-650 with AES beats 12xCore opteron by ~2.48 times, and sandy bridge takes it to a new level.
You could have your OS and all partitions encrypted with AES and you would not even notice performance hit. For a regular Joe it doesn't matter but for those who want it its awesome.

[Movieman]

Where can I grab that benchmark?
I have a 2x12 core(6168) AMD MC system, a 2x6 core Westmere(X5680) system and a almost running SB system to test this on.

[BrowncoatGR]

Where can I grab that benchmark?
I have a 2x12 core(6168) AMD MC system, a 2x6 core Westmere(X5680) system and a almost running SB system to test this on.

http://phoronix-test-suite.com/?k=home Also runs under windows now i think
or you could get OpenSSL and manually try the bench http://www.madboa.com/geek/openssl/#benchmark-speed
http://www.openssl.org/
http://www.slproweb.com/products/Win32OpenSSL.html

[Movieman]

http://phoronix-test-suite.com/?k=home Also runs under windows now i think

Hope so, all three of those machines are on win7 Ent.

[kl0012]

Where can I grab that benchmark?
I have a 2x12 core(6168) AMD MC system, a 2x6 core Westmere(X5680) system and a almost running SB system to test this on.

In fact, testing RSA is pretty useless. Asymmetric-key algorithms (such as RSA) usually aren't used in performance critical encryption. The main purpose of RSA algorithm is to establish secure connection between server and client and then to exchange a cipher keys. After that any of symmetric-key algorithms (DES, AES e.t.c) are used to exchange actual data.

[Movieman]

In fact, testing RSA is pretty useless. Asymmetric-key algorithms (such as RSA) usually aren't used in performance critical encryption. The main purpose of RSA algorithm is to establish secure connection between server and client and then to exchange a cipher keys. After that any of symmetric-key algorithms (DES, AES e.t.c) are used to exchange actual data.

Good, so in a week or two when I decipher what you just wrote I'll be able to make decisions on what to run.

( Yes,just teasing,I understand what you wrote)

[MAS]

proof?

see http://translate.google.com/translat...d%3D7361744320

[informal]

amdzone as an source of information - it is not even funny. just check the results - different versions of test suites, compilers and benchmarks.

No amdzone is just a link where one poster summed the actual results(and you know this,you are just using amdzone as an excuse when you are confronted with actual data). You can search for i7 results,they are pretty consistent as long as 64 bit version of linux is used(no matter what distribution).Also the test suite version has to be the same or similar subversion(as 2.00 vs 2.00.b2).Same goes for AMD X4/X6.

Example
(phoronix test suite 2.0.0) :
http://global.phoronix-test-suite.co...807-9811-25231
2 i7s @ 2.67Ghz compared on 2 diff. gnome kernels(x64),GCC 4.4.1 - result :199

(phoronix test suite 2.0.0b2)
http://global.phoronix-test-suite.co...45-12535-29612
1 i7 @ 2.67 on ubuntu 9.04 (x64),GCC 4.3.3 - result : 191.8

Pretty consistent result for i7 at the same clock,the difference is ~3% even though the OS,compiler and the test suite version are not the same(test suit version is the closest match i could get).

[Aberration]

Well it all comes down to encryption used, if you use industry's standard AES encryption , Intel has build in hardware acceleration, nothing even comes close to speed gains, if i am not mistaken all Intel 32nm CPUs will have AES acceleration.

A 2xCore i5-650 with AES beats 12xCore opteron by ~2.48 times, and sandy bridge takes it to a new level.
You could have your OS and all partitions encrypted with AES and you would not even notice performance hit. For a regular Joe it doesn't matter but for those who want it its awesome.

I got a lot of important and valuable information on my work laptop. There are certain countries that I do not trust leaving it unlocked, but things happen.

Have not tried it all yet, but its on the agenda. Not noticing any performance hit would be nice.

[Hornet331]

No amdzone is just a link where one poster summed the actual results(and you know this,you are just using amdzone as an excuse when you are confronted with actual data). You can search for i7 results,they are pretty consistent as long as 64 bit version of linux is used(no matter what distribution).Also the test suite version has to be the same or similar subversion(as 2.00 vs 2.00.b2).Same goes for AMD X4/X6.

Example
(phoronix test suite 2.0.0) :
http://global.phoronix-test-suite.co...807-9811-25231
2 i7s @ 2.67Ghz compared on 2 diff. gnome kernels(x64),GCC 4.4.1 - result :199

(phoronix test suite 2.0.0b2)
http://global.phoronix-test-suite.co...45-12535-29612
1 i7 @ 2.67 on ubuntu 9.04 (x64),GCC 4.3.3 - result : 191.8

Pretty consistent result for i7 at the same clock,the difference is ~3% even though the OS,compiler and the test suite version are not the same(test suit version is the closest match i could get).

If you compile openssl with AES-NI enable, it can yield up to 12x the performance improvment...

http://software.intel.com/en-us/arti...ith-intel-ipp/

So if you want to run openssl and intel just compile it right... the same way goes for amd...

Thats the fun part about open source software.. but they arn't exactly benchmark friendly since you can customize them so much and nothing is standardized.

[BrowncoatGR]

If you compile openssl with AES-NI enable, it can yield up to 12x the performance improvment...

http://software.intel.com/en-us/arti...ith-intel-ipp/

So if you want to run openssl and intel just compile it right... the same way goes for amd...

Thats the fun part about open source software.. but they arn't exactly benchmark friendly since you can customize them so much and nothing is standardized.

While that is true, that particular algorithm is not based on AES and it wouldn't have made a difference. It's just a corner case that fits AMDs design more than Intel's.

[GoldenTiger]

Noob question: Is 1.5v DDR3 required for "stock" operation? What is the "safe" voltage for DDR3 on sandy bridge (with i7 920 C0 it was 1.65v or so)? I.E. if I'm shopping for RAM, I know "stock" is 1333. Current RAM I have is 3x2GB...

[xVeinx]

There is newer low voltage ram on the horizon, and I'm wondering if that will be preferred for SB. At the very least, it will run cooler than some of the ram out currently...

[GoldenTiger]

There is newer low voltage ram on the horizon, and I'm wondering if that will be preferred for SB. At the very least, it will run cooler than some of the ram out currently...

Any idea if it's coming in time for SB?

[Johnny87au]

Noob question: Is 1.5v DDR3 required for "stock" operation? What is the "safe" voltage for DDR3 on sandy bridge (with i7 920 C0 it was 1.65v or so)? I.E. if I'm shopping for RAM, I know "stock" is 1333. Current RAM I have is 3x2GB...

Really depends on what dimms you pick up, I always run @ rated speed, mine is 1.65, You should just run them at spec no matter what.. unless you really increase the MHZ where you need a voltage bump

[Silver Bullet]

Any idea if it's coming in time for SB?

I wondered this also; but I noticed that a lot of the Corsair Dominator series is OOS. Which I find weird with SB launch so close. I think Corsair has something up it's sleeve for SB.

[Sgt.McRuff]

Any idea if it's coming in time for SB?

Some 1.35v memory is on newegg already... idk if you want lower.
http://www.newegg.com/Product/Produc...e=&srchInDesc=

btw i have used with out problems gkills 2x4gb 1333 and 1600 kits and 2x 2gb 1333 kits. Both on i5/i7 and my 1055t.

[GoldenTiger]

Really depends on what dimms you pick up, I always run @ rated speed, mine is 1.65, You should just run them at spec no matter what.. unless you really increase the MHZ where you need a voltage bump

What I mean is, with nehalem 1.65v was the max supposedly for the imc. Is there a similar thing for sb?

Thanks for your replies all.

[Movieman]

Some 1.35v memory is on newegg already... idk if you want lower.
http://www.newegg.com/Product/Produc...e=&srchInDesc=

btw i have used with out problems gkills 2x4gb 1333 and 1600 kits and 2x 2gb 1333 kits. Both on i5/i7 and my 1055t.

Same and XMP profile shows up in bios on the Intel Burrage board..No issues with that 2x4 gig kit.
Haven't tried my Corsair GTX2's yet, thats next.