3770k Ivy Bridge benchmarked, power usage.

[badass]

And since the socket has been around for Tock+Tick, I really wouldn't expect Haswell to use the same socket, factor that into any paths you're considering. AMD have kept the same basic AM3+ socket for probably longer than ideal but it implies the trinity socket will last a while.

I fully expect LGA1155 to be discontinued after Ivy bridge, however I don't think any CPU upgrade will be necessary for 5+ years after buying such an incredibly fast CPU. Advancements in CPU speed have slowed down a lot in the past 5 years - a Core i7 2500k at nearly £200 in 2012 is less than twice as fast as a July 2007 Core2 Quad Q6600 that cost circa £150 - that's nearly 5 years for a less than doubling of performance in an industry that was used to doubling performance every 18 months-2 years.

Trinity however will likely be slower, CPU wise - however if it's got the rumored video transcoding engine in it to rival quicksync and it's within 20% on a Core i5 2570K (lightly threaded) then it will also probably last 5+ years without an upgrade.
However to be within 20% it will need to be much faster than The FX-8150 or the A8 CPU's in lightly threaded tests. Or Windows 8 will save it. If Windows 8 does save it, that's enough for me as well as that will be the OS I use.

[CAT-THE-FIFTH]

Haswell uses socket 1150 and there has been some noise its successor will use the same socket.

Trinity lacks L3 cache and AFAIK it is estimated that the top end parts will be around 15% to 20% faster than the Llano A8 CPUs at the same TDP. So I suspect it might be comparable to a Core i3(hopefully),with greater multi-threaded performance and lower lightly threaded performance.

Vishera OTH,has L3 cache and probably will clock higher at stock clockspeeds.

[MadduckUK]

Wow! That is sloppy work indeed.

tell em to give you a job you will turn that ship away from the rocks.

[CAT-THE-FIFTH]

tell em to give you a job you will turn that ship away from the rocks.

I don't think many companies will like me though!! AMD won't as I would probably mark them down for their stupidly long higher end cards,plus they should get negative points for their aluminium cooler of fail they use with many of their CPUs.

Using light load results for load power consumption and not repeatedly updating your benchmarks is sloppy. After all it is their jobs which they are making a living out of!

[watercooled]

Based on past experience, it seems each socket will last tock-tick then get replaced - LGA775 for Conroe/Wolfdale, LGA1156 for Lynnfield/Clarkdale and LGA1155 for Sandy/Ivy. The same probably applies to the 'enthusiast' platforms too, LGA1366 for Bloomfield/Gulftown and LGA2011 for SNB-E/shrink.

[shaithis]

That's recent past experience though. Go back further and you will see there isn't a real pattern.

To be fair, they normally change socket based on need, although 1156>1155 was anything but :S

[CAT-THE-FIFTH]

So how many years will it take until we have one pin??

Please don't answer it.

[CAT-THE-FIFTH]

It looks like IB might run relatively hot when overclocked:

http://www.cpu-world.com/news_2012/2...s_Surface.html

It is due to the actual CPU section having a small surface area when compared to SB.

[EvanJackPenn]

Oh gawd. I presume that'll be mainly for v. high overclocks? I might have to invest in a H100 as opposed to a 212 Evo! :O

[shaithis]

Hurrumph.....

I guess it's a question of whether its stable at those coretemps.....as it's generating less heat than SB I guess some fancy new coolers may be able to target heat dissipation into a smaller area.......other than that, I guess we could have done with intel coming up with a newer IHS.

Just wonder if they will do anything if their flagship K models don't surpass the SB Ks in the upcoming benchies every site will inevitably run...

[watercooled]

Considering IHS size is roughly the same and TDP is lower, it's more to do with conduction from die to IHS.

[EvanJackPenn]

From what I read, they conclude the temps are to do with either (or all!)...

- 'Low leakage' transistors are only so with low voltages, therefore high voltage = more heat
- Engineering sample issue
- Poor cooling provided; apparently the heatsink wasn't hot to the touch.
- Using Z68 gave poor results
- Temperature measuring was incorrect

[CAT-THE-FIFTH]

The CPUs are running hot when overclocked and overvolted from what I gather,and not at stock settings.

IB is supposed to be around 166MM2 whereas SB is around 216MM2. On top of this in IB the GPU part takes up a larger percentage of the surface area of the die than the CPU part when compared to SB:

http://www.tomshardware.com/news/Int...out,14782.html



It is quite possible the watts/mm2 dissipated for the IB CPU section is higher than SB leading to higher temperatures.

OTH,there might be other considerations too. After all the main reason Intel used 22NM FINFETS was for improved performance at lower voltages. Hence at higher voltages we don't know if there are additional problems and TBH since most of the CPUs are not going to be overclocked anyway,Intel is not going to be really that bothered I suspect.

[DanceswithUnix]

I see Anandtech have been doing more work, not sure how sloppy this time

Interesting result though, Llano can be very frugal.

http://www.anandtech.com/show/5696/p...-at-65-watts/4

[CAT-THE-FIFTH]

I see Anandtech have been doing more work, not sure how sloppy this time

Interesting result though, Llano can be very frugal.

http://www.anandtech.com/show/5696/p...-at-65-watts/4

I am more suprised they can fit those CPUs into an Antec ISK110!

[watercooled]

I'm not sure what metric they use for 'absolutely lousy' and 'dog slow', I've been using that CPU on my everyday rig for months and it's plenty fast enough for any normal tasks, heck I often have 3 VMs running at the same time doing their own thing and it copes fine, leaving the system responsive. Not to mention they're comparing a £56 CPU against a £255 one, nearly 5 times the price, I don't see any benchmarks where it has 5x the performance...