Core i7 980X doesn't seem much faster than Core i7 860?

[andy5174]

If you have the money to go 1366/920 D0 (or Xeon W3520), why would you ever get an 860? Just the voltage necessary to hit 4+ ghz on s1156 chips is enough to put me off.

I would feel much more comfortable running a 920 at such a speed with lower volts over an extended period of time. Of course, I don't have the money for computer upgrades right now, so for me, it's a moot point . . . but if I did, I wouldn't want to push the volts that s1156 chips need for higher clocks.

Not much voltage increase being required on LGA1366 for aggressive OC is due to its worse BJTs which has a lot of leakage current even when they are in the idle state. Since you have a lot of extra unnecessary current for LGA1366 (and therefore LGA1366 runs much hotter which is not good), you don't need to increase much voltage to raise the power for high OC according to P=VI.

Read the article from the link below. It is about the new BJT & PCU(power control unit) on LGA1156 platform which acts as a TRUE switch. This means that there's almost no leakage current at idle state.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3634&p=4

Regards
Andy

 

[andy5174]

In most cases, HT is required to be turned off on both i7-920 and i7-860 at 4.2GHz which make your CPU no different from i5-750 at 4.2GHz.

i5-750 FTW!

 

[ChorniyVolk]

I bought and built a 1366 set-up specifically out of spite for andy5174.

 

[Mars999]

I now see there will be a Core i7 930 coming out in Mar along side the 980x... I see this 930 will have smart cache tech and maybe 12MB cache on it? Only thing is the core speed is at 2.8Ghz like he 860.... Wish it was 3.06 at least... It's also going to be a 32nm version.

I am going to wait and see what benchmarks show and how much faster this 930 vs. 980x vs. the 860 ends up being.

If not much faster I will probably keep my current build and wait until Haswell....

Besides IMO an SSD would a bigger performance increase vs. a 930 or 980 CPU upgrade over the 860 but I could be wrong.

 

[Interitus]

While you may or may not have problems, I had 24 sticks of DDR3-1600 Ballstix Tracer and could not get the system stable at ALL in any configuration on an EVGA classified 759. GS Kill and Corsair sticks had no problems, however. The Tracer memory runs extremely hot too.

Thanks for the heads up, I'll keep an eye on the RAM if installation goes well.

 

[Mars999]

Nevermind I just read on some other sites that is most likely a 45nm version with 8MB cache... ugh probably a modest 133mhz speed bump....

 

[alyarb]

nm, yeah the 930 is due in february and is 45. just wait for the 32nm xeons in mid march.

 

[ChorniyVolk]

IIRC, wasn't the 930 basically the exact same thing as the 920, only at a slightly higher speed?

 

[andy5174]

IIRC, wasn't the 930 basically the exact same thing as the 920, only at a slightly higher speed?

No. 930 has one more BIN than 920.

 

[aigomorla]

Pish..

Having 12 working tasks = L33T

A lot of gulftown haters this year.. :\

 

[alyarb]

let me see a coretemp screen

 

[aigomorla]

let me see a coretemp screen

Core temps reads delta TjMAX.

So take 105 - (value posted) = core temp.

Gulftown will wipe the floor off any i7 processor you throw at it. (multi scaled / Folding / BOINC / Benches)
I have enough benchmarks soon to be posted to make you go WTF.

 

[alyarb]

i know, i just wanted to see if it puts up a reasonable VID.

i don't think we have any reason to believe singlethreaded performance has improved except for the larger cache. you would need a well threaded algorithm to provide a good enough floor for wiping with bloomfield.

 

[aigomorla]

i know, i just wanted to see if it puts up a reasonable VID.

u do know vid changes as u change the multi...

Meaning lower multi, lower reported values.

 

[alyarb]

yeah and you're at 23 there.

 

[aigomorla]

yeah and you're at 23 there.

thats core temp.

Vid is the 1.1125 value being shown.

On the lowest multi it drops below 1.0V.

Thats how C1E works.

 

[Makaveli]

If you have the money to go 1366/920 D0 (or Xeon W3520), why would you ever get an 860? Just the voltage necessary to hit 4+ ghz on s1156 chips is enough to put me off.

I would feel much more comfortable running a 920 at such a speed with lower volts over an extended period of time. Of course, I don't have the money for computer upgrades right now, so for me, it's a moot point . . . but if I did, I wouldn't want to push the volts that s1156 chips need for higher clocks.

^^^ This

 

[Makaveli]

I just ordered all my parts for a 1366 920 system.

I swayed back and forth between 1366 and 1156 for a long time.

In the end, the 1366 setup I went with cost exactly $94 more than the i7 860 setup I looked at.

X58 setup:

i7 920
EVGA e758-A1
6GB Crucial Ballistix Tracer

P55:

i7 860
EVGA P55 SLI
(RAM was where most of the difference came from) 4GB Crucial Ballistix Tracer

Rest of the setups were identical.

I totally agree with this, when I did my 920 build in oct of 2009. Everyone was like don't go Bloomfield its so much more expensive blah, blah, blah.

After I priced the two systems together the price difference was just over $100 dollars. Not $200-$300 as some people would have you believe.

 

[aigomorla]

well i'll be getting a retail sample of the 980 soon.

When i get that i think the NDA on my A0 gulftown will expire.
And you'll see me spaming results on this forum.

Unless someone beats me to it.

 

[ghost recon88]

well i'll be getting a retail sample of the 980 soon.

When i get that i think the NDA on my A0 gulftown will expire.
And you'll see me spaming results on this forum.

Unless someone beats me to it.

All's I see is 4.5GHz prime runs up there, that's weak sauce :thumbsdown: If your sponsor is dishing out the bucks to send you a retail 980xe, I sure hope you have something subzero (single stage at the least) to help show off their investment.

The really sad thing is the steppings are getting worse and worse as far as overclocking goes. The B0 chips aren't clocking as well as the A0, and if they come out with 1 more stepping for retail, that's just gonna plain suck

 

[aigomorla]

Well my sponsor honestly doesnt care how far i take the chip OC'd.

The only thing he wants is thermals on it, and how it behaves on such apps.

The OC is purely on my side.

And i do have sub ambient... however im VERY LAZY in setting it up.
More like i dont want to insulate anymore.. i feel im getting too old for that. :X

 

[ghost recon88]

And i do have sub ambient... however im VERY LAZY in setting it up.
More like i dont want to insulate anymore.. i feel im getting too old for that. :X

 

[DrMrLordX]

Double post, blah

 

[DrMrLordX]

The real question is: will the LGA1156 or LGA1366 quad require more power overall to stay stable at the coveted speed of 4 ghz (or higher)? If you read the same article you linked, Anand speculates that LGA1156 quads require more vcore than would be necessary to stabilize the cores at high clockspeeds due to the on-die PCIe controller. In other words, you're going to be pushing a lot of volts through an i7-860 or i7-750 (particularly the 750) to get them stable. Leakage current be damned, if I can get a 920 D0 stable at or near 4 ghz with very little extra vcore, I hardly consider that to be a bad thing when I might have to push 1.45v or more through a Lynnfield to achieve the same clockspeeds. Whatever advantage Lynnfield had in terms of power usage at stock speeds would be effectively negated.

In most cases, HT is required to be turned off on both i7-920 and i7-860 at 4.2GHz which make your CPU no different from i5-750 at 4.2GHz.

i5-750 FTW!

If I honestly thought I could get an i5-750 to 4.2 ghz on air without pushing 1.5v vcore or more through it, I'd be interested. Most 750s out there that I've seen don't even want to do 4 ghz.

Granted, even a 3.7 ghz 750 would still wreck my poor little Brisbane.

 

[andy5174]

The real question is: will the LGA1156 or LGA1366 quad require more power overall to stay stable at the coveted speed of 4 ghz (or higher)? If you read the same article you linked, Anand speculates that LGA1156 quads require more vcore than would be necessary to stabilize the cores at high clockspeeds due to the on-die PCIe controller. In other words, you're going to be pushing a lot of volts through an i7-860 or i7-750 (particularly the 750) to get them stable. Leakage current be damned, if I can get a 920 D0 stable at or near 4 ghz with very little extra vcore, I hardly consider that to be a bad thing when I might have to push 1.45v or more through a Lynnfield to achieve the same clockspeeds. Whatever advantage Lynnfield had in terms of power usage at stock speeds would be effectively negated.



If I honestly thought I could get an i5-750 to 4.2 ghz on air without pushing 1.5v vcore or more through it, I'd be interested. Most 750s out there that I've seen don't even want to do 4 ghz.

Granted, even a 3.7 ghz 750 would still wreck my poor little Brisbane.

You will be definitely within the absolute max just like 920 at 4GHz+. The article do mention higher Vcore on LGA1156, but you will be fine as long as you are within spec. In addition, the absolute max Vcore on LGA1156 is higher! Besides, I have seen a few 750 at 4.2GHz and a lot at 4GHz. You must be unlucky to have a bad chip.

Again:

Not much voltage increase being required on LGA1366 for aggressive OC is due to its worse BJTs which has a lot of leakage current even when they are in the idle state. Since you have a lot of extra unnecessary current for LGA1366 (and therefore LGA1366 runs much hotter which is not good), you don't need to increase much voltage to raise the power for high OC according to P=VI.

Read the article from the link below. It is about the new BJT & PCU(power control unit) on LGA1156 platform which acts as a TRUE switch. This means that there's almost no leakage current at idle state.

BTW, there are not much 920 can do 4.2GHz as well. You can see a lot of 920 D0 owners also have trouble at about 3.8GHz ~ 4GHz if you do some search. Google!