Large Ivy Bridge 3570K vs Sandy Bridge comparison.

[Eggy]

Just got a 3570K chip a couple of days ago and have been doing some comparison vs Sandy Bridge at the same clock. HT is disabled.

ATM i'm using chilled water, but will be doing LN2 in a couple of days.




3DMark 11
Ivy Bridge 3570K @ 3.4Ghz (stock)


CPU Score
Run 1: 6423
Run 2: 6481
Run 3: 6429
Average: 6444.3

Sandy Bridge 2600K HT OFF @ 3.4Ghz


CPU Score
Run 1: 6411
Run 2: 6392
Run 3: 6386
Average: 6396.3
Improvement from SB to IB: 48P = 0.75%


Ivy Bridge 3770K HT OFF @ 4.5Ghz

8d215851_45_2.jpeg
CPU Score
Run 1: 8059
Run 2: 8041
Run 3: 8095
Average: 8065

Sandy Bridge 2600K HT OFF @ 4.5Ghz


CPU Score
Run 1: 7728
Run 2: 7700
Run 3: 7696
Average: 7708.0
Improvement from SB to IB: 357P = 4.63%


3DMark Vantage
Ivy Bridge 3570K @ 3.4Ghz (stock)


CPU Score
Run 1: 18903
Run 2: 18579
Run 3: 18905
Average: 18795.3

Sandy Bridge 2600K HT OFF @ 3.4Ghz


CPU Score
Run 1: 18233
Run 2: 18192
Run 3: 18068
Average: 18164.3
Improvement from SB to IB: 631P = 3.47%


Ivy Bridge 3770K HT OFF @ 4.5Ghz

2cd8a2e9_IB_45_1.jpeg
CPU Score
Run 1: 24752
Run 2: 24709
Run 3: 24810
Average: 24757.0


Sandy Bridge 2600K HT OFF @ 4.5Ghz


CPU Score
Run 1: 23721
Run 2: 23781
Run 3: 23532
Average: 23711.3
Improvement from SB to IB: 1045,7P = 4.41%

CineBench.
Ivy Bridge 3570K @ 3.4Ghz (stock)


Score: 5.67P

Sandy Bridge 2600K HT OFF @ 3.4Ghz


Score: 5.43P
Improvement from SB to IB: 0.24P = 4.41%


Ivy Bridge 3770K HT OFF @ 4.5Ghz
6ea6079b_Untitdasdasdled.jpeg
Score: 7.47P


Sandy Bridge 2600K HT OFF @ 4.5Ghz

Score: 7.09P
Improvement from SB to IB: 0.38P = 5.35%


Super Pi mod1.5 XS
Ivy Bridge 3570K @ 3.4Ghz (stock)


32m - 10m 00.469sec (600.46 sec)

Sandy Bridge 2600K HT OFF @ 3.4Ghz


32m - 10m 21.443sec (621.44 sec)
Improvement from SB to IB: 20.98sec = 3.37%


Ivy Bridge 3770K HT OFF@ 4.5Ghz

574f7146_Udasdasdntitled.jpeg
32m - 08m 05.002sec (485.00 sec)

Sandy Bridge 2600K HT OFF @ 4.5Ghz


32m - 08m 14.989sec (494.98 sec)
Improvement from SB to IB: 13.72sec = 2.77%


Wprime 1.55
Ivy Bridge 3570K @ 3.4Ghz (stock)


Run 1: 8.766 sec
Run 2: 8.766 sec
Run 3: 8.766 sec
Average: 8.766 sec

Sandy Bridge 2600K HT OFF @ 3.4Ghz


Run 1: 9.205 sec
Run 2: 9.205 sec
Run 3: 9.205 sec
Average: 9.205 sec
Improvement from SB to IB: 0.439 sec = 4.76%

Ivy Bridge 3570K @ 4.5Ghz

4dad8505_Untidasdatled.jpeg
Run 1: 6.645 sec
Run 2: 6.645 sec
Run 3: 6.645 sec
Average: 6.645 sec

Sandy Bridge 2600K HT OFF @ 4.5Ghz


Run 1: 6.941 sec
Run 2: 6.941 sec
Run 3: 6.941 sec
Average: 6.941 sec
Improvement from SB to IB: 0.296 sec = 4.26%


Wprime 2.09
Ivy Bridge 3570K @ 3.4Ghz (stock)


Run 1: 10.639 sec
Run 2: 10.638 sec
Run 3: 10.640 sec
Average: 10.639 sec

Sandy Bridge 2600K HT OFF @ 3.4Ghz


Run 1: 11.138 sec
Run 2: 11.171 sec
Run 3: 11.186 sec
Average: 11.165 sec
Improvement from SB to IB: 0.526 sec = 4.71%


Ivy Bridge 3770K HT OFF @ 4.5Ghz

Run 1: 8.142 sec
Run 2: 8.096 sec
Run 3: 8.095 sec
Average: 8.111 sec

Sandy Bridge 2600K HT OFF @ 4.5Ghz

d6276b30_Untisadasdasdtled.jpeg
Run 1: 8.439 sec
Run 2: 8.455 sec
Run 3: 8.439 sec
Average: 8.444 sec
Improvement from SB to IB: 0.333 sec = 3.94%


Handbrake
Ivy Bridge 3570K @ 3.4Ghz (stock) = 13m 09sec (789 sec)
Sandy Bridge 2600K HT OFF @ 3.4Ghz = 14m 12sec (852 sec)
Improvement from SB to IB: 1m 03sec (63 sec) = 7.39%

Ivy Bridge 3770K HT OFF @ 4.5Ghz = 10m 14sec (614 sec)
Sandy Bridge 2600K HT OFF @ 4.5Ghz = 11m 08sec (668 sec)
Improvement from SB to IB: 54sec = 8.08%


Summarize - Improvements from Sandy bridge to Ivy Bridge

3DMark 11.
3.4Ghz - 0.75% Improvement
4.5Ghz - 4.63% Improvement

3DMark Vantage
3.4Ghz - 3.47% Improvement
4.5Ghz - 4.41% Improvement

CineBench
3.4Ghz - 4.41% Improvement
4.5Ghz - 5.35% Improvement

SuperPi 32M
3.4Ghz - 3.37% Improvement
4.5Ghz - 2.77% Improvement

Wprime 1.55
3.4Ghz - 4.76% Improvement
4.5Ghz - 4.26% Improvement

Wprime 2.09
3.4Ghz - 4.71% Improvement
4.5Ghz - 3.94% Improvement

Handbrake
3.4Ghz - 7.39% Improvement
4.5Ghz - 8.08% Improvement

Note: Later tests have shown that 3770K HT OFF is ~1% faster then a 3570K at the same clocks in all tests but CineBench (Due to 2Mb extra L3 Cache), so the 3.4Ghz comparison is not 100% correct ATM. Will be updated later.


Testing is done on this hardware:

Asus Maximus V Gene
Gskill 1600 6-9-6-24 (running @ 9-9-9-24)
OCZ 1250W

Random 3770K @ 4.9Ghz / 5Ghz

Ivy Bridge 3770K @ 4.9Ghz

Cpu Score: 34523



Score: 9.86


Ivy Bridge 3770K @ 5Ghz

Score: 4.772 sec

Thermal testings

Cooling
Custom loop, dual HwLabs GTX360 GT-15 Push-Pull, Dual Laing D5 @ Setting 5 = 285 L/H (1.255 GPM) 10-15c ambient.

All results are in Celsius, load done with Prime95.

ff1dd25b_Thermal.jpeg

73f2d28a_3570kvs2600ktemperature.jpeg


Note: Core 2,3 and 4 are showing wrong readings on start (Idle) on the 3570K (use core 1 as true reading)


@ 1.1V IB runs 4,25C hotter then SB with HT off. (16.03%)
@ 1.2V IB runs 10,75C hotter then SB with HT off. (36.44%)
@ 1.3V IB runs 15,00C hotter then SB with HT off. (44.11%)
@ 1.4V IB runs 24,75C hotter then SB with HT off. (65.56%)

[rouvio]

wow!

thank you for such a comprehensive benches.
so much material to process here..!
personally I love threads with lots of information

[iluv2fly]

Thank you for the benchmarks, very useful info! It looks like the i5-3570k is a better buy then the i7-2600k at the same price.

MicroCenter.com lowered the price of the i7-2600k to $199 yesterday. No wonder, they need to get rid of it before the 3570k hits the shelves :-)

I thought a scored a good deal when I bought it for $199. I'm returning it! If you want pci-e 3 you need the 3570k anyways...

[Mechromancer]

This test was done with Hyper Threading disabled. How much faster is the 3570K in multithreaded applications when the HT is enabled on the 2600K? I bought a 2600K yesterday for $199 too!

[clubber_lang]

I think for gaming purposes since most don't really take advantage of HT ( I think? )......I definitely wouldn't be selling my 2600K or 2700K any time soon. In the tests above......I've seen way bigger margins between just two different 2600K's. This is bumming me out.

[Callsign_Vega]

I think for gaming purposes since most don't really take advantage of HT ( I think? )......I definitely wouldn't be selling my 2600K or 2700K any time soon. In the tests above......I've seen way bigger margins between just two different 2600K's. This is bumming me out.

Yup, pretty much for gaming there is no reason to go with IB unless you need PCI-E 3.0. But if that't the case you can just go with X79 and get a lot more lanes. For 24/7 gaming computer SB and SB-E may vary well be better than IB due to virtually the same speed in games clock for clock and IB may clock lower due to high temperatures.

[jcniest5]

I think for gaming purposes since most don't really take advantage of HT ( I think? )......I definitely wouldn't be selling my 2600K or 2700K any time soon. In the tests above......I've seen way bigger margins between just two different 2600K's. This is bumming me out.

Did you try its OCability yet? What batch # is yours? I got a 3202B190 and OCing isn't that good...Can't even get it to run stable at 5Ghz using 1.544V (LLC Enabled and vcore set at 1.475v in the BIOS).

[PatRaceTin]

Very extensive !!!

[Eggy]

Did you try its OCability yet? What batch # is yours? I got a 3202B190 and OCing isn't that good...Can't even get it to run stable at 5Ghz using 1.544V (LLC Enabled and vcore set at 1.475v in the BIOS).

Batch# L209C207

Mine is a bad clocker. i've not tried it that much but it is 3d stable @ 1.3v @ 4.5Ghz but once core is failing in Prime95, i've tried 4.8Ghz @ 1.45v, but it wont even get into windows.

As for the OCability, i have not tried that much, i'll do that now but i need some help in form of input regarding mobo settings.

I've just got this Max V Gene Z77 because my old P67-UD7 is not 100% supported (locked multi) on IB, and i've not clocked much on Asus since the old MIIF 775 days, so there is a lot of new stuff to adjust here that is not present on a UD7 board. So to sum it up, all info and tips regarding OC'ing on Max V Gene is really appreciated, this will also help big time on LN2.

[bwanaaa]

do you have a watt meter by any chance? how much less juice did IB need?

btw, thank you very much for good info. obviously this is an IB engineering sample-how did you get it w/o NDA?

[Andrea deluxe]

about temperatures?

[bwanaaa]

about temperatures?

+1
please

[zalbard]

Keeping in mind that 2600K has more cache.

[Ket]

IB has never been intended to provide any real raw CPU performance improvements over SB so I'm not sure why people were expecting anything to write home about in that regard. The main reasons for IB are for a improved on-die GPU and for PCI-E 3.0. IB is more a incremental step forward, or a side-grade. Not a real next gen step forward.

[Dumo]

all info and tips regarding OC'ing on Max V Gene is really appreciated, this will also help big time on LN2.

Heres http://www.xtremesystems.org/forums/...US-V-GENE-quot

[xdan]

Keeping in mind that 2600K has more cache.

+1
It was a mistake from the begining to compare i5 2500K 6MB L3 cache with i7 2600K 8MB L3 cache.
L3 cache has it's role and in some application make's a big difference.
2600K benefits best from HT but bigger cache helps too quite qood.
Here are some of my results with my 2500K.
More results tomorow.
Vantage 3.4ghz Cpu - 17.722 IB 18903 + 6.7%
Vantage 3.4ghz.jpg
Vantage 4.5ghz Cpu -23141 IB 24187 + 4.5%
Vantage 4.5ghz.jpg

Cinebench 3.4ghz - 5.27 points IB 5.67 points + 7.7%
Cinebench 3.4ghz.jpg
Cinebench 4.5ghz - 7.09 points IB 7.67 point + 5.35%
Cinebench 4.5ghz.jpg

wPrime 32M 3.4ghz -10.43s IB 8.766s + 16%
wprime 32M 3.4ghz.jpg
wPrime 32M 4.5ghz - 7.47s IB 6.691s + 10.5%
wprime 32M 4.5ghz.jpg

Stock results are looking pretty good 7-10% and in some cases will be probably 13-15%.
Average will be probably 10% improvement as it was sad from the begining.
What it's strange in both Eggy results and mine is that after overclock to 4.5ghz the improvement of IB drops 40-50%.
May be some early imature bioses which hinder performance of IB during overclock?
Or higher cores temperatures to IB which affects performance.

[Eggy]

do you have a watt meter by any chance? how much less juice did IB need?

btw, thank you very much for good info. obviously this is an IB engineering sample-how did you get it w/o NDA?

Nope this is not a ES, full retail chip, and i'm under no kind of NDA.

about temperatures?

Will see if i can get some test done there.

Heres http://www.xtremesystems.org/forums/...US-V-GENE-quot

Thanks.

+1
It was a mistake from the begining to compare i5 2500K 6MB L3 cache with i7 2600K 8MB L3 cache.
L3 cache has it's role and in some application make's a big difference.
2600K benefits best from HT but bigger cache helps too quite qood.
Here are some of my results with my 2500K.
More results tomorow.
Vantage 3.4ghz Cpu - 17.722 IB 18903 + 6.7%
Vantage 3.4ghz.jpg
Vantage 4.5ghz Cpu -23141 IB 24187 + 4.5%
Vantage 4.5ghz.jpg

Cinebench 3.4ghz - 5.27 points IB 5.67 points + 7.7%
Cinebench 3.4ghz.jpg
Cinebench 4.5ghz - 7.09 points IB 7.67 point + 5.35%
Cinebench 4.5ghz.jpg

wPrime 32M 3.4ghz -10.43s IB 8.766s + 16%
wprime 32M 3.4ghz.jpg
wPrime 32M 4.5ghz - 7.47s IB 6.691s + 10.5%
wprime 32M 4.5ghz.jpg

Stock results are looking pretty good 7-10% and in some cases will be probably 13-15%.
Average will be probably 10% improvement as it was sad from the begining.
What it's strange in both Eggy results and mine is that after overclock to 4.5ghz the improvement of IB drops 40-50%.
May be some early imature bioses which hinder performance of IB during overclock?
Or higher cores temperatures to IB which affects performance.

Thanks, i know it was a bit of a bad comparison considering the L3 cache, but it was what i had at hand. I'm expecting a couple of 3770K's in a couple of days so will redo some of the test then with HT on. Also i do not think that it's the right way to compare the way you are doing. You are using your 2500K on your hardware and comparing it with my 3570k on my hardware. Different ram sticks, on same timing, clocks and so on perform different, and different motherboards will also make quite a large impact on the score. I will do some quick benchmarks on the 3770K with HT off and compare them against my 3570K scores to see how much impact the L3 does have.

[Eggy]

Got a little package today.

[Eldonko]

You try some 3d comparos with Virtu MVP enabled?

[Eggy]

Does anyone know anything regarding Multiplier Wall on Ivy Bridge? Does it have a MP wall like SB did? If so is it the same procedure as SB to find max multi or how does that work? Not really keen on pushing 1.55v into it when only running chilled water (this thing is even hitting 75c after 20min of Prime @ 1.35v with 10c water)

[iboomalot]

If you were upgrading a I7-920 watercooled setup

2011-SBE or 1155-IB ???

Not really keen on pushing 1.55v into it when only running chilled water (this thing is even hitting 75c after 20min of Prime @ 1.35v with 10c water)

Those kinds of temps long term (since I usually go 2-3yrs before upgrading) doesn't sound appealing.

[Eggy]

If you were upgrading a I7-920 watercooled setup

2011-SBE or 1155-IB ???

Not really keen on pushing 1.55v into it when only running chilled water (this thing is even hitting 75c after 20min of Prime @ 1.35v with 10c water)

Those kinds of temps long term (since I usually go 2-3yrs before upgrading) doesn't sound appealing.

I know the delta T on the Block and the IHS is totally insane here. @ 1.35v we are talking 65C and i'm not even sure that is max because i stopped it there. To be honest i would not be surprised if the chip is 0c under LN2, i know we are talking -160c or so, but we are also talking 1.8-1.9v and seeing how the temp scaled here i would assume that under LN2 the chip is not much below 0.

Back to the max multiplier subject.

Does anyone know anything regarding Multiplier Wall on Ivy Bridge? Does it have a MP wall like SB did? If so is it the same procedure as SB to find max multi or how does that work? Still safe to test with 1.55v to check the max multiplier?

[Eggy]

Updated all of the 4.5Ghz 3570K tests to 4.5Ghz 3770K HT OFF tests, i already had the 4.5Ghz 2600K HT OFF tests done so that was the easiest.

As pointed out by some users the extra 2Mb Cache on the 2600K was playing a bit in it's favor, but not much. The 3570K vs the 2600K HT OFF average performance difference was 2.68%, after redoing the 4.5Ghz benchmarks over again with the 3770K with HT off the difference is 4.77%. So 3770K HT OFF vs 2600K HT off is on average 4.77% in favor of the 3770K, both running @ 4.5Ghz

The extra L3 cache was also the cause of the gap between the 2600K and the 3570K @ 4.5Ghz on 3Dmark 2011, now the favor is on IB side again.

I've also added a couple of 4.9Ghz and 5Ghz benchmarks with the 3770K with HT on to see what it is capable off, load temps is plain sick, and i am only able to load for 1-2min at the time @ 1.42v before heat starts building up and the cores jump up in the 70's, that is even with a cool 10c water temp.

Will have to do the 3.4Ghz 3770K tests to complete the comparison on Monday or Tuesday. Tomorrow i'm going over to a friend to trow the chips under the pot.

[El Mano]

Oh boy, waiting for IB for so long and now it seems it's not worth it against socket 2011.
AMD where are you?



I'm seriously considering buying a 3820. It's cheaper than a 2600K here and the platform will last a bit longer, as long as IB-E is worth it.

[Eggy]

LN2 runs done. Best of the 3770K chips maxed out @ 6.360Mhz 2c/2t and 6.250 with 4c/4t. The other was ~100Mhz below on both.
The 3570K was a lousy 5700Mhz or something like that so that was not an option.

Two of the benchmarks that is decent:

http://www.hwbot.org/submission/2276...70k_5sec_756ms

http://www.hwbot.org/submission/2276...770k_11.61_sec

Sry for the damaged screenshot on PiFast, don't know what happened and 50% of the screens taken was like that or worse.