Ivy bridge overclocking is disappointing

[bret emerald]

Hi guys, not sure if anyone's seen the ivy bridge previews over at Tweaktown.com but the results are a bit disappointing. They tested both the i5 3570k and i7 3770k and have an overclocking article as well. It seems ivy bridge runs hotter than sandy bridge, possibly due to a smaller die size which has less area to dissipate heat. The samples they tested also didn't overclock as well as the sandy bridge cpus. Looks like Mario Galaxy with LLE might be causing slowdowns for quite some time yet

[Duke Nukem]

(03-24-2012, 03:42 AM)Damned Wrote: Well I guess I'm lucky cause the only Wii games I'm interested in are Mega Man 9 and 10.

Well I guess so too, Mega Man 9 and 10 shouldn't be a problem for anyone with a decent computer these days.


As for the Ivy Bridge thing, I think people sees it as a savior or something but it won't change a thing.
We'll all have to wait a little bit longer till games like Super Mario Galaxy runs 100% with DSP LLE.

My guess is that most games will run 100% speed with DSP LLE within 2-3 years given that the Dolphin project is still actively worked on then.
But that's my guess, anything can happen and fast. Heck one day someone magically finds out a way to improve Dolphin making everything runs smoothly.

...I don't know

[Zee530]

We can all dream i guess

From the start ivybridge was thought to be roughly only 10% better than sandybridge but if sandybridge still wins in the overclocking department i might have to start rethinking my choice of processors

[AnyOldName3]

Of course, someone who was talking about the tick tock model of intel processors (architectural overhaul, then die shrink) gave me the thought that because sandy bridge was the overhaul, and ivy bridge was the shrink, the next one will actually be the one with the benefits and increased overclocking. For me, ivy bridge was never going to affect me, as I've only been on sandy for under 4 months, so wasn't going to go out and buy one. By the time I do need a new processor enough to justify buying one the next one after ivy will be done, and it won't matter if the socket changes, as my motherboard doesn't support overclocking, so I'll need a new one anyway.

[NaturalViolence]

.....but that makes no sense. Die shrinks usually raise clock rates while overhauls usually lower clock rates, it's been this way for 6 years now.

And sure enough the stock clocks of ivy bridge are higher than sandy bridge while the TDP and power consumption are much lower. I don't see why it would scale so poorly with temperature though, die size certainly isn't the reason.

[Zee530]

Dolphin is aimed at accuracy rather than performance so the emulator is likely to get slower with every new revision, our hope lies with companies that make computer parts, so far they make parts powerful enough, all our games would run fullspeed

[bret emerald]

(03-25-2012, 06:54 AM)NaturalViolence Wrote: .....but that makes no sense. Die shrinks usually raise clock rates while overhauls usually lower clock rates, it's been this way for 6 years now.

And sure enough the stock clocks of ivy bridge are higher than sandy bridge while the TDP and power consumption are much lower. I don't see why it would scale so poorly with temperature though, die size certainly isn't the reason.

Yeah I was surprised with the results too. The incresed temps are more likely due to higher levels of resistance/impedance caused by using shorter wires (due to smaller die). Perhaps we've reached a point where the thermal benefits of requiring less voltage are outweighed by the higher electrical resistance. It will be very interesting to see what happens with the move to 14nm, but at least we'll have one more architecture revision to boost performance before then.

Anyway, regardless of the reasons, all testing so far points to one thing.....we're going to have to live with higher temps with ivy bridge.

[ExtremeDude2]

(03-25-2012, 06:54 AM)NaturalViolence Wrote: .....but that makes no sense. Die shrinks usually raise clock rates while overhauls usually lower clock rates, it's been this way for 6 years now.

And sure enough the stock clocks of ivy bridge are higher than sandy bridge while the TDP and power consumption are much lower. I don't see why it would scale so poorly with temperature though, die size certainly isn't the reason.

This ^^

[Shonumi]

(03-25-2012, 06:35 PM)Zee530 Wrote: Dolphin is aimed at accuracy rather than performance so the emulator is likely to get slower with every new revision, our hope lies with companies that make computer parts, so far they make parts powerful enough, all our games would run fullspeed

True, as Dolphin advances, it's less likely to cut corners. Most of the speed gains we'll see will come from hardware upgrades. Still, you shouldn't discount cases where existing code is improved to provide better performance. One recent example I remember was 3.0-415. skid's commit not only lowered the CPU usage on my third core with LLE on Thread (100% vs 50~60% now) it also allowed me to play games that required LLE at full-speed without having to overclock at all. Such improvements are rare, but they happen.

[Starscream]

(03-26-2012, 01:10 AM)Shonumi Wrote: One recent example I remember was 3.0-415. skid's commit not only lowered the CPU usage on my third core with LLE on Thread (100% vs 50~60% now) it also allowed me to play games that required LLE at full-speed without having to overclock at all. Such improvements are rare, but they happen.

That actually slowed things down even more for me, approx 5% slower on my AMD chip.