Dolphin, the GameCube and Wii emulator - Forums

http://en.wikipedia.org/wiki/Wafer_testing#Wafer_prober

Quote:In some very specific cases, a die that passes some but not all test patterns can still be used as a product, typically with limited functionality.
The most common example of this is a microprocessor for which only one part of the on-die cache memory is functional.
In this case, the processor can sometimes still be sold as a lower cost part with a smaller amount of memory and thus lower performance.
Additionally when bad dies have been identified, the die from the bad bin can be used by production personnel for assembly line setup.

http://efreedom.com/Question/E-5621/Faster-Clock-Require-Power
and because of that... increasing voltage maintains stablitiy if the production serie is not perfect

or maybe intel got a problem with
http://en.wikipedia.org/wiki/Leakage_(electronics)

Quote:higher voltage results in higher leakage current.
In high transistor count devices like a modern desktop CPU leakage current can account for the majority of power dissipation.
As process size gets smaller and transistor counts rise, leakage current becomes more and more the critical power usage statistic.

iam not your wikipedia...

This "report" on the ivy bridges wattage is a clone report (with bad grammar) of one from Nordic Hardware. Honestly, I'm not sure about its accuracy. So far I haven't found any kind of review from a real user (or real website) that analyzes the 3770k properly. But I have found a lot on the 3750k. Specifically, this forum post, where several people got them early have been performing tons and tons of tests. It has a lot of interesting results, but first up:

CPU-Z on an i5 3750k - http://cdn.overclock.net/a/a5/a5ca3d66_Untitlssasaed.jpeg

77w as advertised. Now, the 3770k could be different, but I don't know yet.

[color=#DCDCDC]His testing of the chip shows that the i5 3750k tested roughly 5% higher clock for clock with the higher tier i7 2600k, which the 2500k couldn't hope to even match. So, just guesstimating, if compared to the sandy bridge i5 2500k it would appear to line up with the 15-20% clock for clock improvement that intel previously stated. Unfortunately, even at 77w, he was only able to get the clock up to 4.5ghz on air. The 2600k was able to reach 5.2ghz readily, and with the 3750k only reaching 4.5ghz with a 5% clock for clock increase, I doubt the i5 3750k can beat the i7 2600k at overclock. But that doesn't matter, no one expected it to! The testing also showed that even with 77w, the ivy bridges run much hotter. According to the test reports I read, that even with the reduced voltages, the transitors are just so crammed together, the CPU has difficulty channeling heat off the die. Still, even though it ran a lot hotter, it was able to remain stable at higher temperatures, so the temperature shouldn't be that big a deal. We'll see how much temperature plays into overclocking when the hardware sites do their reviews.
[/color]
Update: Eggy's chip (the one the forum thread and the above paragraph are based on) seems bad, as all the other testers are getting far better than him. So ignore that block of text above. (I really wish this forum had a strike-through code right about now)

Another user in the thread was able to hit 4.8ghz on air with his 3750k, with stability. Unfortunately, even at 77w, the temps were MUCH higher than what a sandy bridge would get. Still, even though the temps were alot higher, it was completely stable, and the temperatures were about the same as my core 2 duo's overclock (60C under load). He was able to hit 5ghz, and it booted up fine, but the temperature was high enough that he got worried and backed off before stability testing. A second user was able to confirm this 4.8ghz result, so this appears to be the range of the retail chips, and not the 4.5 max (air or water) result that the original poster showed.

It looks like, while the ivy bridges are WAY hotter, they can reach the same kind of overclocking specs that the sandys could. 4.8Ghz on air is standard overclocking for a Core i5 2500k, with 5ghz on air possible but not stable: pretty much the same as the ivy bridge i5 3750k. Plus, the ivy bridge clock for clock is higher: as shown in the thread, an i5 3750k is roughly 5% faster clock for clock than a higher tier i7 2600k. While I don't have a proper 2500k to 3750k comparison yet, guesstimating from the i7 comparison, it would appear to hold true to the prerelease numbers of 15-20% clock for clock improvement over sandy.

So far I like what I'm seeing. The ivy bridge 3750k appears to beat the 2500k on all counts, and could potentially even rival the 2600k. I have been waiting for months for the Ivys to replace my very old Core 2 Duo, so I'm really glad they are performing well so far. Well, minus that one poor guy that got a dud, his results gave me a freakin heart attack >_<. Still, of course, I am cautiously optimistic; all this information is very very early with leaked retail parts. I'll have to wait for the hardware sites to do proper reviews first before making any kind of real decision on buying Ivy or Sandy.

(04-15-2012, 03:13 PM)MaJoR Wrote: [ -> ]Unfortunately, even at 77w, the temps were MUCH higher than what a sandy bridge would get.

Which still doesn't make any sense

Quote:Which still doesn't make ant sense

No one really knows for sure, as Intel isn't talking, but it appears to be because of the 22nm process. Even though it uses less power and generates less heat, the transistors are SO tightly packed it's difficult for the die to shed heat effectively. It may be fixed in the coming months as they refine the chips, but we'll probably have to wait for haswell for them to sort it out. Still, from what I'm seeing, the heat isn't as bad as I get on my core 2 duo, so while sandy users are freaking out, I'm not all that worried. Pentium 4s routinely hit 70-80C and were perfectly fine. As long as it's stable and safe, it doesn't really matter.

(04-15-2012, 05:52 AM)dannzen Wrote: [ -> ]http://en.wikipedia.org/wiki/Wafer_testing#Wafer_prober

Quote:In some very specific cases, a die that passes some but not all test patterns can still be used as a product, typically with limited functionality.
The most common example of this is a microprocessor for which only one part of the on-die cache memory is functional.
In this case, the processor can sometimes still be sold as a lower cost part with a smaller amount of memory and thus lower performance.
Additionally when bad dies have been identified, the die from the bad bin can be used by production personnel for assembly line setup.

http://efreedom.com/Question/E-5621/Faster-Clock-Require-Power
and because of that... increasing voltage maintains stablitiy if the production serie is not perfect

or maybe intel got a problem with
http://en.wikipedia.org/wiki/Leakage_(electronics)

Quote:higher voltage results in higher leakage current.
In high transistor count devices like a modern desktop CPU leakage current can account for the majority of power dissipation.
As process size gets smaller and transistor counts rise, leakage current becomes more and more the critical power usage statistic.

iam not your wikipedia...

I already know both of those facts. Now answer my question please.

Quote:Even though it uses less power and generates less heat, the transistors are SO tightly packed it's difficult for the die to shed heat effectively.

This makes no sense to anyone who has studied thermodynamics. Packing transistors closer together would not cause that, that's not an opinion, that's physics.

You know, I'm going to have to do a redaction.

Ivy Bridge 3750k - 4.8ghz @ 70C (load) on air

Sandy Bridge 2500k - 4.7ghz @ 73C (load) on air
Sandy Bridge 2500k - 4.8ghz @ 70C (load) on air

Why are all these people complaining about about ivy bridges running hot and weak? It overclocks to the same levels as sandy, runs at roughly the same temps, and is better clock for clock. What the hell? I guess I made a mistake by believing them and not looking at just the data.

They are supposed to be much cooler, not the same temperature.

I'm no physicist but IIRC smaller dies are harder to cool since there's less surface area for the HSF to make contact with.

Quote:I'm no physicist but IIRC smaller dies are harder to cool since there's less surface area for the HSF to make contact with.

Yes.

However:
1. Putting transistors closer together does not necessarily make the die size smaller. Ivy bridge however does have a smaller die size than sandy bridge. What I mean to say is "ivy bridge runs hotter because it has a smaller die" is not the same thing as saying "ivy bridge runs hotter because the transistors are closer to each other".
2. The HSF makes contact with the aluminum cover, not the die.
3. The TDP of ivy bridge is supposed to be much lower than sandy bridge.

Some new info. According to the leaked retail users, the boxes are labeled with 95w for the 3570k, as well as the 3770k. So they appear to run at the same power level as the sandy bridges. This lines up pretty well with the other info we have, as they run at the same general temperature at the same overclock rates. So for those that wanted alot less heat and power use, looks like that didn't happen. Still, if it has that 15-20% increase clock for clock with the same power and heat, it's a win.

As for why CPU-Z says 77w, it may just use a database instead of data from the CPU. It's not a sensor reading, so who knows. We'll find out when Intel releases them to the reviewers and starts answering questions.