IDF Sandybridge overclocking - 4.9Ghz

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Dark_Archonis

Member
Sep 19, 2010
88
1
0
The original Core 2's performed 90% faster per clock than the Prescott based Pentium 4. With Sandy Bridge even the LV chips should threaten the 5GHz Pentium 4. With multi-threading its going to reach that with the ULV chips. Of course this is all per core comparisons. :)

I don't think a 10Ghz Pentium 4 processor could top a Sandy Bridge :p
Look at it this way:
Took a 5.2Ghz Pentium 4 to beat a K8 Athlon at 2.6Ghz
Which would be equal to about a ~2.3Ghz 65nm Core 2 Duo
Which would probably be equal to a 1.66Ghz 32nm Clarkdale
So if the Sandy Bridge processors will ship at 3Ghz and above--

Well I guess I was being far too generous with the P4 comparison :p.

s2011 has an extra memory channel (quad channel), AFAIK no GPU. It also moves the PCI-E controller on die (since it was on the NB before, and AFAIK thats gone) and thats 40 lanes of PCI-E 3.0). Those account for the vast majority of new pins.

Doesn't 1155 have on die PCI-E 2.0, according to AT's article?

Nehalem using Bloomfield and Lynnfield core is ~10% faster in single thread than Penryn and with Clarkdale its less than 5%.

Core 2: 2x clock P4
Core 2: 45nm-2.2x clock P4
Core i7: 2.5x clock P4
2nd gen Core i7: 3-3.3x clock P4

By "2nd gen i7" are you referring to SB, or Bloomfield/Lynnfield?
 

ilkhan

Golden Member
Jul 21, 2006
1,117
1
0
Doesn't 1155 have on die PCI-E 2.0, according to AT's article?
1156 and 1155 have 20 PCI-E 2.0 lanes (including 4 for DMI->southbridge). I was comparing to s1366, which has 40 PCI-E 2.0 lanes on the northbridge.

"2nd gen core" is now the marketing name for sandy, so thats what he was referring to.
 
Last edited:

ydnas7

Member
Jun 13, 2010
160
0
0
I don't think a 10Ghz Pentium 4 processor could top a Sandy Bridge :p
Look at it this way:
Took a 5.2Ghz Pentium 4 to beat a K8 Athlon at 2.6Ghz
Which would be equal to about a ~2.3Ghz 65nm Core 2 Duo
Which would probably be equal to a 1.66Ghz 32nm Clarkdale
So if the Sandy Bridge processors will ship at 3Ghz and above--

but a P4 is probably about the 1/4 the size of each SB core...
 

Triskain

Member
Sep 7, 2009
62
30
91
but a P4 is probably about the 1/4 the size of each SB core...

Not really if you count the transistors. A single Pentium 4 core takes up ~ 50 million transitors, but a Westmere core needs only about 55 million transistors while being about 2,5 times the speed of the Pentium 4 core at the same clock.
For reference a Phenom II core takes up ~ 32 million transistors.
 

ydnas7

Member
Jun 13, 2010
160
0
0
prescott chip was 112mm2 at 90nm
shrink that by about 8 to get 32nm equivalent, thats a 14mm2 die including 1mb cache, 64bit and HT. and whatever glue they used.

put 10 of those cores on a chip today and you could make the most intelligent oil heater on the market.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,786
136
Prescott isn't that compact. For comparison

Pentium D, 2x cores, 4MB L2, 65nm: 162mm2, 376 million transistors
Core 2 Duo, 2x cores, 4MB L2, 65nm: 143mm2, 291 million transistors

Die shots of single core Cedarmill also shows the core to cache ratio is greater than on the Core 2. Nehalem is about 10-15% larger than the core in Core 2, and Sandy Bridge is another 10% larger than that.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
Prescott isn't that compact. For comparison

Pentium D, 2x cores, 4MB L2, 65nm: 162mm2, 376 million transistors
Core 2 Duo, 2x cores, 4MB L2, 65nm: 143mm2, 291 million transistors

Die shots of single core Cedarmill also shows the core to cache ratio is greater than on the Core 2. Nehalem is about 10-15% larger than the core in Core 2, and Sandy Bridge is another 10% larger than that.

Not surprising either, if you want clockspeeds then you need lots of Idrive and Ion which means wider transistors and less "xtr/mm^2" density.

(I don't have the data at the tip of my memory but I also wouldn't be surprised if the general Vcc/VID of 65nm P4-based SKUs were higher than Core 2 as well)

It seems people either don't know or are quick to forget that xtors are 2-dimensional components in the circuit and they vary in both directions for a reason.

Anecdotally, I see a similar logical-fallacy crop up a lot in the VC&G forum when people try to make xtr/mm^2 arguments/comparisons between NV and AMD gpus.
 

Ben90

Platinum Member
Jun 14, 2009
2,866
3
0
I just got back from my trip, and it seems the "Omg Sandy Bridge is going to be the worst processor ever because it won't overclock" hysteria is over. Consequentially/Coincidentally a certain moderator with a SB ES spreading the hysteria seems to be absent from this thread.
 

Martimus

Diamond Member
Apr 24, 2007
4,488
153
106
I just got back from my trip, and it seems the "Omg Sandy Bridge is going to be the worst processor ever because it won't overclock" hysteria is over. Consequentially/Coincidentally a certain moderator with a SB ES spreading the hysteria seems to be absent from this thread.

True, but if the Anandtech review is any indication, there will not be a SB model released that will be faster than the 980X. (at least until socket 2011, and who knows when/if that will be released for the desktop)

That disappoints me :(
 

khon

Golden Member
Jun 8, 2010
1,318
124
106
True, but if the Anandtech review is any indication, there will not be a SB model released that will be faster than the 980X. (at least until socket 2011, and who knows when/if that will be released for the desktop)

That disappoints me :(

That would depend on the application.

The i7-2600 will be faster than the i7-980X for most tasks.
 

Edrick

Golden Member
Feb 18, 2010
1,939
230
106
That would depend on the application.

The i7-2600 will be faster than the i7-980X for most tasks.

True. In most cases, unless you are running more than 8 threads, the 2600 should be faster.
 

exar333

Diamond Member
Feb 7, 2004
8,518
8
91
I just got back from my trip, and it seems the "Omg Sandy Bridge is going to be the worst processor ever because it won't overclock" hysteria is over. Consequentially/Coincidentally a certain moderator with a SB ES spreading the hysteria seems to be absent from this thread.

This was the same discussion everyone had 2 years ago when Nehalem was launched. "How will we OC this?" "Will it be possible" etc....

As tempting as SB sounds, I will probably wait and jump on it when an affordable (<$500) 6 or 8 core is released.
 

Riek

Senior member
Dec 16, 2008
409
14
76
True. In most cases, unless you are running more than 8 threads, the 2600 should be faster.

doubt it. The moment HT makes an impact on SB, the i7-980X will be getting very close or be faster. This starts from 5threads, not 8.
 

Edrick

Golden Member
Feb 18, 2010
1,939
230
106
doubt it. The moment HT makes an impact on SB, the i7-980X will be getting very close or be faster. This starts from 5threads, not 8.

Well, also keep in mind that all the numbers we have seen from SB so far have been from a 2500 part. The 2600 part will be clocked faster and have more cache (6 vs. 8). Plus, again from we have seen, SB (K series) may have more OC headroom than the 980X.

So until we see hard numbers of the 2600, we will not know for sure. But it is my opinion that the 2600 will beat out the 980X in most applications.
 

khon

Golden Member
Jun 8, 2010
1,318
124
106
doubt it. The moment HT makes an impact on SB, the i7-980X will be getting very close or be faster. This starts from 5threads, not 8.

If we look at the AT review there are 14 benchmark comparisons of the i5-2400 and the i7-980X, of which the i7-980X wins 12-2.

However, if the i5-2400 had been replaced with the i7-2600 it would have scored 15-25% higher (higher clock, working turbo, more cache). Looking at the 14 benchmarks again, how many would the i7-2600 then win ? The answer is it would probably beat the i7-980X 10-4, and some of those are benchmarks where HT does make an impact (adobe cs4, 3dsmax 9, WinRAR 3.8).
 

khon

Golden Member
Jun 8, 2010
1,318
124
106
I keep counting over and over and there's only 12. :p

Maybe you can go back to kindergarten and practice your counting ? ;)

Adobe cs4 - i7-2600 would win
DivX - i7-980X would win
x264 first pass - i7-2600 would win
x264 second pass - i7-980X would win
Windows media encoder - i7-2600 would win
3dsmax 9 - i7-2600 would win
Cinebench single threaded - i7-2600 would win
Cinebench multithreaded - i7-980X would win
POV-Ray - i7-980X would win
PAR2 - i7-2600 would win
WinRAR - i7-2600 would win
Batman - i7-2600 would win
Dragon Age Origins - i7-2600 would win
World of Warcraft - i7-2600 would win

That's 14 benchmarks, with the i7-2600 winning 10-4.
 
Last edited:

Haserath

Senior member
Sep 12, 2010
793
1
81
25% IPC boost is crazy amazing...especially when that comes on top of Bloomfields already impressive IPC.

You think this boost is basically coming from the addition of that L0 uop $?

Or are we looking at a special compile of Cinebench that takes advantage of some of the new ISA extensions?



http://en.wikipedia.org/wiki/Strain_engineering

It is actually yet another one of those major "pros" to the lengthy list of performance advantages that are intrinsic to doing a disposable gate (also called replacement gate or "gate last") integration scheme.

Don't get me wrong, my rant above wasn't to say that 4.9GHz on air isn't great, means we'll probably be seeing some 6GHz water-cooled rigs in the forum. Hmmm...6GHZ...
Sorry if this is off-topic;) and sorry about crushing your hopes about 6ghz clocks but:
http://www.anandtech.com/show/3922/intels-sandy-bridge-architecture-exposed/8
These will be fully unlocked, supporting multipliers up to 57x.
100X57=5700mhz, so even with an amazing water setup made by the gods, you won't be able to clock it to 6ghz:(
 

ilkhan

Golden Member
Jul 21, 2006
1,117
1
0
6Ghz is only 57x105. Even as un-OC friendly as the bus is supposed to be, 5Mhz may be possible.
And yeah, 5700x100 making 6Ghz nearly impossible was one of the first things I thought of when I saw that number. I wonder how they picked it; AKA was it a hardware limited thing or arbitrary?
 

khon

Golden Member
Jun 8, 2010
1,318
124
106

Kreed

Junior Member
Nov 29, 2001
5
0
0
If an i7-2600 is going to beat an i7-980x in most benchmarks, what point is there in selling quad core 1366 Bloomfields come Q1 2011? Is there a chance that Intel will just replace all the quad core Bloomfields with similarly clocked Gulftowns?