Early Ivy Bridge Sample Tested

[Zoran]

Well, I do not have a "k" but the 3770 retail sample i have runs 41 (max allowd multi for all cores) x 108 stable @ 1.224V (Load), consumes ~102W (only the CPU under P95 Small FFT) and does so with 2*4GB G.Skill @ 2160. (For now )

Temps with HR-02 Macho

@ 700rpm pull. [dead silent] are ~77C average on cores [realtemp 3.7], with PCProbe showing 51C on CPU
@ 1300rpm push the tems are down to ~62C average and PCProbe showing 41C.

Unfortunately, I can not go any higher till I get a higher BCLK out of the Maximus V GENE.

@4GHz the CPU needs 1.09V and consumes ~70W!

[Hans de Vries]

The GPU on Ivy Bridge is 2.25 times bigger as the GPU on Sandy Bridge.

http://www.chip-architect.com/news/2...es_Sandys.html

The latest Ivy Bridge die as shown by UBM is larger as previous Ivy bridge
dies. (170 mm2) Somehow the area for the GPU has grown is size. Might
be for better yield but maybe for higher performance.


Hans.

[Andi64]

I've never seen such a drastic change between two similar steppings from the same CPU

Some said that E0 steppings were clocking as high as 5Ghz, but E1 is stuck at 4.5 to 4.7Ghz. Maybe this explains it.

[Cold Fussion]

Because

- Stock temps looks a lot better and the temp seems to rise at the higher clocks a lot faster than the increase in voltage required.
- Subzero-OC friendly
- Comparing 45nm to 22nm with barely any performance increases I doubt Intel decision makers would just pat the back of the engineers and say good job making 10% faster chip with 50%+ higher temp.
- 40C temp differences only due to different internal designs when going from a 45nm to 22nm process seems to go out of the border what's realistic "goals" using common sense. Maybe if we're talking about entirely different kind of company/resources/devices whatever but not for such a large scale important player in the IT business.

Leakage makes more sense than Intel accepting a design which would be the cause of such a large boost in temp as if it's leakage it would only be a prob at clock frequencies which are way out of their targeted clock range (no big deal if it does handle the targeted clocks good enough)

I don't see how you can say it has high leakage if you haven't seen power consumption figures ov heavily oc'd chips.

[aussie-revhead]

Can someone 'leak" a 3D11 cpu score? I really want to know if the ipc and ram clocks help much.

[[XC] Oj101]

I'll leak total scores

Core i5-2500K @ 5.00 GHz + GeForce GTX 580 @ 960/1,920/1,100 MHz = 7,991 3DMarks
Core i5-3570K @ 4.55 GHz + GeForce GTX 580 @ 960/1,920/1,100 MHz = 8,014 3DMarks

[thebanik]

I'll leak total scores

Core i5-2500K @ 5.00 GHz + GeForce GTX 580 @ 960/1,920/1,100 MHz = 7,991 3DMarks
Core i5-3570K @ 4.55 GHz + GeForce GTX 580 @ 960/1,920/1,100 MHz = 8,014 3DMarks

hmmm....3DM11 score, would have been more interesting to compare with 2600K........

[aussie-revhead]

Are you fair dinkum? Same rams and clocks in both tests? If so then that's much better than what the 6% IPC would indicate .....

Thebanik- they are both non-HT CPUs so comparison is valid.

[thebanik]

Thebanik- they are both non-HT CPUs so comparison is valid.

Ah!!!!! missed to look at the CPU, .

[MaddMuppet]

It's looking good for me and a Phase unit.

[[XC] Oj101]

Same RAM was used, yes. 3D will be the playground of Ivy Bridge, the gains can be greater than 2D. Those gains are not guaranteed across the board though, for example Vantage on the Core i5-3570K at 4.55 GHz is more or less the same as the Core i5-2500K.

Oh, and Lucid MVP was disabled for both chips.

[dctokyo]

Has anyone notice this? I thought is was going to be 77 TDP instead of 95

[SubZero.it]

[SIZE=4]Has anyone notice this? I thought is was going to be 77 TDP instead of 95

Yes, more or less 10 pages ago...
Going back to the 3dMark scores I don't call that a difference but I must grab one for PCI-Ex 3

[STaRGaZeR]

I don't see how you can say it has high leakage if you haven't seen power consumption figures ov heavily oc'd chips.

Besides, I don't really get the high temp stuff. All those reported high temps are because the chip is really generating a lot of heat or it's just that it can't dissipate a the "little" amount of heat generated fast enough?

[Leeghoofd]

If you stop the fan on the air cooler and the heatsink get's barely hot : seems conductivity is not so good between IHS and core.... SOF has popped one of his dead CPU's and it seemed there was a lot of thermal paste between the core and the IHS...

[Mafio]

IMHO the CPUs aren't really hotter than Sandy Bridge stuff, it's just that with SB DIE and IHS were soldered togheter while Ivy have just some kind of crappy thermal paste between IHS and DIE.
take off the IHS like we did back in the A64 days should do the trick and solve those "heat issues".

[RPGWiZaRD]

I really wonder why E1 stepping was made such large change to if it went from 77W to 95W, would be quite upsetting as an overclocker if the early rumors of E0 users that 5GHz was reasonably easy to reach on those. Would be very interesting to hear the reason behind the internal change (this is not usually seen on a stepping change) but too bad we're probably never going to.

[aussie-revhead]

If you stop the fan on the air cooler and the heatsink get's barely hot : seems conductivity is not so good between IHS and core.... SOF has popped one of his dead CPU's and it seemed there was a lot of thermal paste between the core and the IHS...

He has killed Ivy's already? Does he have a thread/pics somewhere?

I agree with Mafio, pop the lid.

[ziddey]

I really wonder why E1 stepping was made such large change to if it went from 77W to 95W, would be quite upsetting as an overclocker if the early rumors of E0 users that 5GHz was reasonably easy to reach on those. Would be very interesting to hear the reason behind the internal change (this is not usually seen on a stepping change) but too bad we're probably never going to.

most likely because of the larger graphics core

[Leeghoofd]

TDP on the CPU's boxes is platform TDP. We reached around 99.88 watts with everything stressed with Sandra stress tool.
sorry it was Thomas that killed one :

3770K die shot

[aussie-revhead]

Wow is that all it is? It's going to be tricky balancing a water block on that without chipping edges etc....

Lol

[Leeghoofd]

That's why mankind invented die shims, recalling Athlon XP ?

[SoF]

TDP on the CPU's boxes is platform TDP. We reached around 99.88 watts with everything stressed with Sandra stress tool.
sorry it was Thomas that killed one :

3770K die shot

just was about to point you on that I did not kill a chip yet even I really pushed hard on cascade - IB looks strong to me

[Kain665]

Well, I do not have a "k" but the 3770 retail sample i have runs 41 (max allowd multi for all cores) x 108 stable @ 1.224V (Load), consumes ~102W (only the CPU under P95 Small FFT) and does so with 2*4GB G.Skill @ 2160. (For now )

Temps with HR-02 Macho

@ 700rpm pull. [dead silent] are ~77C average on cores [realtemp 3.7], with PCProbe showing 51C on CPU
@ 1300rpm push the tems are down to ~62C average and PCProbe showing 41C.

Unfortunately, I can not go any higher till I get a higher BCLK out of the Maximus V GENE.

@4GHz the CPU needs 1.09V and consumes ~70W!

If your socket temp is 51c and core temps are reported at 77c, I'd be willing to bed that the TJMax for these chips has changed. That's almost a 30c difference, that's not normal.

[kl0012]

I really wonder why E1 stepping was made such large change to if it went from 77W to 95W, would be quite upsetting as an overclocker if the early rumors of E0 users that 5GHz was reasonably easy to reach on those. Would be very interesting to hear the reason behind the internal change (this is not usually seen on a stepping change) but too bad we're probably never going to.

Here are some reasons:
http://www.nordichardware.com/news/6...um-of-77w.html