How high will ivy bridge overclock

[boczek]

All test so far show that tri-gate = overlockers' nigtmare
Im really dissapointed.

 

[davel]

so for someone like myself who is not huge into overclocking, I just need a stable fast system, should I still get IB? or just get sandy then?

or still too early to know for sure

 

[Absolute0]

If you're not overclocking, get Ivy Bridge. With OCing out of the equation, IB is better in every way.

 

[davel]

If you're not overclocking, get Ivy Bridge. With OCing out of the equation, IB is better in every way.

Cool thanks!

I will wait it out.

Out of curiosity when people are overclocking their chips to say 4.5 - 5Ghz, how much more FPS are they really getting in say BF3 compared to stock speeds?

Are we talking huge differences here?

 

[pelov]

Depends on the GPUs. If you're GPU limited then OCing your chip really won't provide much of an increase at all. If you're going SLi or X-fire with 7970s/GTX680s then yea, OCing provides a clear benefit.

 

[davel]

Depends on the GPUs. If you're GPU limited then OCing your chip really won't provide much of an increase at all. If you're going SLi or X-fire with 7970s/GTX680s then yea, OCing provides a clear benefit.

Well I just bought one GTX 680, and going to pair with the i5-3750 i think..

I am only going to be running on 1080p on one 25 inch monitor.

 

[pelov]

You'll be fine unless you're considering adding another GTX680.

 

[Concillian]

Gaming, in general is usually not very limited by the CPU if you have something decent (a 2500k / 3570k is definitely decent).

You generally turn up the vidual setting until the GPU is the limiting factor in most cases. However there may be some isolated incidents where the CPU is a limiting facor and will result in a short time low FPS dip.

Those dips are really what get eliminated by OCing. How frequent they are and when they occur is VERY dependent on the game. BF3 is not very CPU dependent if you have a SB / IVB 4 core and I don't suspect there is much noticeable difference between stock and OC.

When you get your chip, you can try it out and see for yourself. It's not a huge difference unless the game is very CPU dependent. Very CPU dependent games tend to fall into 2 categories:
-- console ports that didn't get enough PC development (GTA IV, for example. Also Skyrim before patch 1.4)
-- Games that have LOTS of computer controlled units (Starcraft II, Total War series games, etc...)

 

[Phynaz]

All test so far show that tri-gate = overlockers' nigtmare

Which tests would those be?

 

[Imouto]

When you get your chip, you can try it out and see for yourself. It's not a huge difference unless the game is very CPU dependent. Very CPU dependent games tend to fall into 2 categories:
-- console ports that didn't get enough PC development (GTA IV, for example. Also Skyrim before patch 1.4)
-- Games that have LOTS of computer controlled units (Starcraft II, Total War series games, etc...)

Add:

-- Poorly optimized MMORPGs that only perform like intended with the newest hardware even running a 5 years old engine. (fe TERA Online)

 

[gplnpsb]

Think about it, you have more transistors in a smaller space which means the cores will need to have more cooling.

Would they not also be smaller, and have less surface area to generate heat on. So remain cooler?

Well, no...they have less surface area to DISSIPATE heat off of, which is the main issue.

I've seen this argument repeated a lot, but I'm not sure how valid I think it is. Suppose you clocked a i7 2600K and a i7 3820 at the same speed and same voltage, would the i7 2600K run hotter, when its die size is 216 mm2 vs 294 for the 3820? I haven't seen any tests but I'd be surprised if it did. When Intel shrank Nehalem, they added 2 cores and 4MB L3. The resulting Westmere die was 240 mm2 vs 263 for Bloomfield, and yet Westmere ran cooler and used less power than the i7 975.

I have a hard time believing that the decreased surface area for Ivy Bridge would serve to increase temperatures more than a shrink to 22nm would be expected to decrease them. I tend to think that FinFET 22nm works very well for lower power applications, but that there is some inherent property of the process that causes it to heat up more quickly at higher power usage than the previous 32nm planar process.

 

[Endymion FRS]

If you're not overclocking, get Ivy Bridge. With OCing out of the equation, IB is better in every way.

So what if I was planning on doing a small OC, like 4.0Ghz? Would I still be ahead moving forward to IVB?

 

[Absolute0]

So what if I was planning on doing a small OC, like 4.0Ghz? Would I still be ahead moving forward to IVB?

Yeah because Ivy will do 4 Ghz no problem and 4 Ghz Ivy will beat 4 Ghz SB.

We dont have all the info yet but its looking like Ivy Bridge won't quite reach as high. Hopefully if you have a good heatsink you should still be able to do 4.5 Ghz though.

 

[Tempered81]

Ivy at 4.5 would probably be a go. I think it gets hairy from 4.6-5.2 range. Thats where ivy and sandy will compete on air/water. Ivy already owns Sandy on LN2 at 6.5+. I wouldn't be surprised if most Ivy's topped around 4.6-4.8 temp limited, while folks with a 4.9-5.1 Sandy would see no benefit by switching to said Ivy.

If the initial goal was 4.5 or 4.6ghz, then 3770K>2700K because you get the IPC improvement

 

[exar333]

I've seen this argument repeated a lot, but I'm not sure how valid I think it is. Suppose you clocked a i7 2600K and a i7 3820 at the same speed and same voltage, would the i7 2600K run hotter, when its die size is 216 mm2 vs 294 for the 3820? I haven't seen any tests but I'd be surprised if it did. When Intel shrank Nehalem, they added 2 cores and 4MB L3. The resulting Westmere die was 240 mm2 vs 263 for Bloomfield, and yet Westmere ran cooler and used less power than the i7 975.

I have a hard time believing that the decreased surface area for Ivy Bridge would serve to increase temperatures more than a shrink to 22nm would be expected to decrease them. I tend to think that FinFET 22nm works very well for lower power applications, but that there is some inherent property of the process that causes it to heat up more quickly at higher power usage than the previous 32nm planar process.

150w TDP (just an example for a 4.4ghz+ CPU) is the same regardless if it's 22nm or 32nm. The die size is smaller, thats less surface area for the die to cool from, so the temps rise. Unless the power usage is proportionally smaller compared to the decrease in die size, you likely will see the temps rise.

This could be a big advantage for IB-E this time around with a larger die (8C+).

 

[gplnpsb]

150w TDP (just an example for a 4.4ghz+ CPU) is the same regardless if it's 22nm or 32nm. The die size is smaller, thats less surface area for the die to cool from, so the temps rise. Unless the power usage is proportionally smaller compared to the decrease in die size, you likely will see the temps rise.

This could be a big advantage for IB-E this time around with a larger die (8C+).

Hmm, perhaps you're right, I can agree with that. I guess the question is how does Ivy's TDP/Power consumption compare to Sandy's at 4.5GHz+. If it's essentially the same or only marginally reduced, the reduced die size argument does work.

 

[Edrick]

I've seen this argument repeated a lot, but I'm not sure how valid I think it is. Suppose you clocked a i7 2600K and a i7 3820 at the same speed and same voltage, would the i7 2600K run hotter, when its die size is 216 mm2 vs 294 for the 3820? I haven't seen any tests but I'd be surprised if it did.

I do not have any graphs or charts or anything of the sort, but I can tell you my experience with a 2500K and 3820 using the same case (Antec 902) and same cooler (H60) running at the same clocks 4Ghz.

2500K: 1.19v @ 4.0Ghz (no HT) full LinX load was 56-58C
3820: 1.23v @ 4.0Ghz (with HT) full LinX load is 52-54C

Those voltages are what I set in the BIOS. And the ambient temps were the same (20C). Fan speeds are all stock as I do not mess with the H60 settings.

It is not a huge difference until you consider that HT does in fact generate more heat. I will test with HT off and see how low the temps will go. I will probably require less vCore as well without HT.

Also to note, the 2500K was small compared to the large copper plate on the H60. It only covered about half the surface area. The 3820 is much larger and covers much more surface area on the H60. I think this is the biggest reason for the drop in temps.

 

[Lepton87]

Also to note, the 2500K was small compared to the large copper plate on the H60. It only covered about half the surface area. The 3820 is much larger and covers much more surface area on the H60. I think this is the biggest reason for the drop in temps.

Are you talking about IHS size? If so I don't agree. IHS is only an insulator between heat sink and CPU and doesn't in any way help temps.

 

[Rvenger]

I do not have any graphs or charts or anything of the sort, but I can tell you my experience with a 2500K and 3820 using the same case (Antec 902) and same cooler (H60) running at the same clocks 4Ghz.

2500K: 1.19v @ 4.0Ghz (no HT) full LinX load was 56-58C
3820: 1.23v @ 4.0Ghz (with HT) full LinX load is 52-54C

Those voltages are what I set in the BIOS. And the ambient temps were the same (20C). Fan speeds are all stock as I do not mess with the H60 settings.

It is not a huge difference until you consider that HT does in fact generate more heat. I will test with HT off and see how low the temps will go. I will probably require less vCore as well without HT.

Also to note, the 2500K was small compared to the large copper plate on the H60. It only covered about half the surface area. The 3820 is much larger and covers much more surface area on the H60. I think this is the biggest reason for the drop in temps.

This actually makes sense. Heatspreader is smaller so it actually heats up faster and hotter, especially since fitting everything onto a smaller die is going to make heat dissipation even harder. The 3820 probably has a ton of space on the die since the cores were chopped down to 4.

@ Lepton
If the heatspreader was an insulator we would have a ton of dead chips on our hands. The heatspreader doesn't help cooling, but it certainly isn't an insulator.

 

[Edrick]

Are you talking about IHS size? If so I don't agree. IHS is only an insulator between heat sink and CPU and doesn't in any way help temps.

It is a heat spreader, and as with the case with all heat spreaders, the larger surface area means it can transfer heat more quickly to the CPU cooler (H60 in my case).

 

[Imouto]

A larger IHS does't really help with cooling because it's really thin. All the work is done by the heatsink.

In fact an IHS is usually making your temps worse. I just love that my HD 5850 doesn't have any, the heatsink is attached right to the GPU.