Crunching comparison 2600K and 3770K

[OldChap]

Thought it wouldn't do any harm to have this info so anyone who has these feel free to add anything.

I have my 3770K in an Asus 5 gene and would be happy to recommend that MB. It is my first with UEFI so I have been learning. Overclockers dream, very forgiving and will clock ram for fun.

So, the 3770k in the gene currently running 4.7 so that comparisons are easier. It has 8gb 1600 samsung green ram running 2200 10,10,10,21 at the moment. I may get that faster or tighter...we'll see

for comparison My 2600k is in a ud4 running @ 4.7 with 4gb corsair 1600 ram running 7,8,7,24 and has one Samsung 103j 1TB HDD the same type fitted to the other rig

First observation is that hooked up to the same cooling and using the same water blocks and Tim, the IB is running an average of 10 degrees hotter which makes me wonder if reduced clocks may be in order in hotter climes.

It will take a week to get some results from the 2600K as just now it is on sustainable water only so bear with me on that but I think that completion times should give us a clue here

From memory I would say that Hcc completes in around 0.1 hours quicker ...The few results so far show a range of 0,89 to 0.99hrs to complete

__________________________________________________ __________________________________________________ _______________________

EDIT:Now that things are stable and before I start presenting results some final info;

I consider the slight improvement here could be both mem speed and cpu imc related and as such I should not try to match as this is just better performance from newer tech/availability of cheaper higher speed ram.

2 Memory and Cache screenshots of..... Hell read them yourself



Ambients at time of the next two = 18.5° Air in to rad, 20.2° water from rad remember both rigs cooled by this system and both use EK HF and ICD



Voltages: The SB 2600K has been at this voltage for a year. The IB 3770K was LinX and Prime stable on 1.26 but gave me errors so Bumped higher for safety which has pushed the temps up again. This needs more time to find the true safe crunching voltage.

Max ambient today was around 25°-26° From the above the SB max average is 56° IB max average is almost 68°

__________________________________________________ __________________________________________________ __________

First up will be the results from both machines crunching Fight Aids @home

Incidentally although you will see differing completion times, WCG shows the following TARGET completion times;

2600K = 3:34:09 and 3770K = 3:51:27

Make of that what you will

These results taken from 75 work units done by each machine

2600K / 3770K

totals

Runtime 227.18 222.91

Claimed pts 8321.7 8585.3

Awarded pts 8528.2 8399.3

Averages

Hrs/wu 3.029 2.972

Pts claimed/wu 110.96 114.47

Pts awarded/wu 113.71 111.99

Points/hr

Claimed 36.63 38.51

Awarded 37.54 37.68

(man, I wish they would enable the table feature on this site.)

[WFO]

Great post OC! I'm, afraid you may be the only one to provide an apples to apples comparison. All my 2600Ks are on air in the 4524mhz range. 1.28-1.335 v-core. 1 3770K is on water, the other 2 on air. 2 @ 4600+ 1 @ 4500. I can post pix of 1 week production for 3 of each if you think it will add to your thread.

Thought it wouldn't do any harm to have this info so anyone who has these feel free to add anything.

I have my 3770K in an Asus 5 gene and would be happy to recommend that MB. It is my first with UEFI so I have been learning. Overclockers dream, very forgiving and will clock ram for fun.

So, the 3770k in the gene currently running 4.7 so that comparisons are easier. It has 8gb 1600 samsung green ram running 2200 10,10,10,21 at the moment. I may get that faster or tighter...we'll see

for comparison My 2600k is in a ud4 running @ 4.7 with 4gb corsair 1600 ram running 7,8,7,24 and has one Samsung 103j 1TB HDD the same type fitted to the other rig

First observation is that hooked up to the same cooling and using the same water blocks and Tim, the IB is running an average of 13 degrees hotter which makes me wonder if reduced clocks may be in order in hotter climes.

It will take a week to get some results from the 2600K as just now it is on sustainable water only so bear with me on that but I think that completion times should give us a clue here

From memory I would say that Hcc completes in around 0.1 hours quicker ...The few results so far show a range of 0,89 to 0.99hrs to complete

[OldChap]

results (done as a pic because charts is not enabled here and users can read what they need from this one)

Each set of figures is derived from 50 results for each machine






PLEASE consider this a guide only.... You will need to confirm on your own rigs that these results hold true for you too


Decided that this summary would better serve the purpose of this thread. Outstanding results will be added as they become available

Revision as at June 30 2012 reloaded the points for CFSW after the issue with points claimed when the 64bit app was introduced

[OldChap]

It occurs to me that the other information that needs to be gathered is that of efficiency. If I were to run the 3770K so that completion times for projects were approximately the same, is it the opinion of you guys that the points awarded would be approximately the same??? If so then we then start to require the power use info....

I should have a decent result on the overclock by the end of this weekend to show power use as things stand.

[D_A]

It would be helpful to define some constants for such a comparison. Personally I'd go for same core voltage and same (stock or close to it) clock speed to get a good base comparison of both power consumption and arch instruction efficiency, then wind the clocks up and see how they each scale.

[haschioz]

my 2600k @4,2 gets about 37,45points/hour on cfsw. from what i have seen the ivys gets the same bench scores. but i am not sure how the score is factored in.

[OldChap]

I have just switched to all projects for a day or two to get some sample times. cache size is 2 days so before we get to those I need to know what people think should be the constants?

clock for clock? (this is what I have now with the exception of ram speed. I have another set of samsungs if we need to do that)

Volts to Volts? (Difficult maybe unless I slow the 2600K a bunch)

Temp to Temp?

Match times for completion??? (may take a while to set up)

Match Gflops??

EDIT: I know it is the middle of the night there but when you get a chance post up those results DA, to satisfy my curiosity about points awarded at different speeds

[Vinas]

It's going to be difficult to get an absolute power rating from different systems... Maybe you should take a "base rating" with each system, then OC one a little to match the other chip, taking note of the power draw change. IDK seems like a lot of work but I think this is some great information OldChap!

[itznfb]

I'm running a 2600k and 3770k both at 4.2ghz and both at 1.225vcore. They pull the same power from the wall at about 140watts. It'll take about a week or so to see the results for comparison.

I just checked the kill-a-watt and the Ivy has actually used .4 more kwh than the Sandy (39 hours)... however the Ivy machine has a PCIe GPU installed where the Sandy is using the integrated graphics.

[haschioz]

so a 550ti pulls like 15-20 watts in idle. in 39 hours thats ~0,6kwh. no really much of a difference anyways.

[OldChap]

Quick question Off Topic. DA can you link me to best places in Aus to buy good computer gear especially watercooling.

[itznfb]

so a 550ti pulls like 15-20 watts in idle. in 39 hours thats ~0,6kwh. no really much of a difference anyways.

Those numbers would make sense then. Showing that in 39 hours the Ivy was more likely even with the Sandy or .1~.2kwh less on power consumption.

[OldChap]

Is this a simple way to check graphics power use?

PUT any card in pc; measure power.

Add the card in question in second slot; Measure power draw.

Subtract the first figure from the second.

I need to get an adaptor for the Gene from hdmi or display port to vga or dvi to cut down on having a card installed

[D_A]

Quick question Off Topic. DA can you link me to best places in Aus to buy good computer gear especially watercooling.

There's a few places here. I normally use Techbuy but I'm not exactly up on all the latest water cooling so I'd be hard pressed to tell if they had a good range or not.

[OldChap]

With a 5870 graphics card fitted and running at 4.7 with cpu volts of 1.248v I find the Kill-a-Watt showing around 190-200 watts at the wall but this is with an old server psu.

EDIT: Digging around for info. I found that the idle power consumption of the graphics card is 27 Watts.

These figures I will update as soon as I get a new lead (hdmi to vga) which will enable me to remove the gpu and a new seasonic (more efficient).

Having found a better cpu voltage I have updated the earlier post regards temperatures (now around 10° higher than SB)

Had a few invalids and errors so I will be also updating my second post re: times and points now that I have new ram timings. (9,10,10, @ 2133)

[Conquistador SW]

Hopefully I can also provide some interesting numbers. I had a 2600K at 4.6Ghz 1.34v (rig that is still in signature) and I switched it out to a 3770K at 4.7Ghz 1.28v.
I could not get 4.8Ghz stable at all. I went up to 1.35v but temps would get to 99C with IBT so I gave up. Except for the CPU everything stayed exactly the same, even memory settings etc.

The 3770K uses between 35-40W less than the 2600K which is better than I hoped for as the 3770K is equivalent to a 2600K at ~4.9Ghz (cinebench r11.5 gives me 9.55 point is you want to compare).
The temperatures however are ~15C higher. Using a EK supreme HF and PA120.3 I get 88C max when running Intel Burn Test but realistically never saw higher temps than 79C when running BOINC and gaming etc (Realtemp TJmax 105C / ~25C ambient)

The 2600K did ~46000 points/day running all projects and I will see what the 3770K will do but it will probably take a few days for the old WU's to clear.

I will also switch out my motherboard for a P8Z77-V Pro soon, and I will then do a comparison at 4.6Ghz if you guys prefer this. I know 4.6Ghz is stable at 1.25v or lower. If you guys don't have a preference I will do the comparison at 4.7Ghz as that is the more interesting comparison for me personally because it will be running at that speed.

[itznfb]

So far my Sandy is producing about 6,000 more points per day than the Ivy while pulling the same power from the wall.

[Vinas]

So it sounds like, at best Ivy does about the same PPD with a negligent difference in power consumption. If so, I think I'll skip my planned upgrade to Ivy and just stick with the sandies this time around.

[itznfb]

So it sounds like, at best Ivy does about the same PPD with a negligent difference in power consumption. If so, I think I'll skip my planned upgrade to Ivy and just stick with the sandies this time around.

Yeah I wouldn't be surprised after more testing to see that Ivy/Sandy produce the same at equal clock/voltage settings.

[OldChap]

Not so sure that is true. For the same power I would expect to get 10-15% more output, at least looking at things so far, however, some projects gain and some loose a little so maybe...... Whether that actually converts to points too remains to be seen. It will probably take until later next month to be sure. At that point it should be possible to do things the other way around and match output in order to see what temps and power is being used by a 3770K and how much is being saved that way.

I am noticing a broad variance in completion times on both rigs, Does anyone think it might be advantageous to select each project in turn for a couple of days each?

[Conquistador SW]

In my case there is a very large difference between the two. 40W less consumption and an expecteded 5-10% more output is a lot to me.

[itznfb]

So, for anyone running a comparison it's really apples to oranges unless we are using the same PSU, Motherboard, dGPU (if present), peripherals etc... as well as working on the same targeted set of work units or... just accepting any work units and utilizing a longer testing average such as a 2-3 weeks. I think I might run the 2600k machine for a week or so to get an average and then put the 3770k in that same machine and do the same. Currently thought my Sandy kicking the crap out of my Ivy.

[haschioz]

i guess you can do 4 days for each project except cep 2 (variance in runtime for my 2600k is 3,5h-10,5h) on a 2770k and a 2600k, everything else stays the same. try and write the completion times down and average them as usual. i would imagine that ivy comes out as more efficient and more results, everything else is just icing on the cake (eg ppd).

[OldChap]

Got one more Stability test to do tonight then maybe it will stop the couple of compute errors in the last 24 hours even after passing LinX x 10 and 4 hours Prime. So at the end of the week latest I will start on 1 project at a time, 2 or 3 days each.

[OldChap]

Sorely tempted to try changing the Intel tim but somewhat nervous of removing the ihs. Seems worth it though for a potential gain of 20°

http://www.xtremesystems.org/forums/...05#post5104005