Report: Intel Haswell to be Released on June 2, 2013
It appears that Intel's Haswell processor will launch in early June. According to slides that were leaked to the Chinese website VR-Zone, the product introduction time frame will be between May 27 and June 7, 2013.
Consumers apparently will be able to get their Haswell processors beginning on June 2. The slides indicate that Intel is preparing 14 different desktop SKUs as well as seven mobile SKUs for launch.
Of course, Sunday, June 2 as the introduction date for Haswell is still speculation at this point, but Intel traditionally launches new architectures on Sundays and there is no reason to believe that the company will break this tradition with Haswell. The extended introduction is likely due to the Computex tradeshow, which opens its doors on June 4 to 8.
Haswell therefore arrives substantially later than originally anticipated and it appears that Intel's product cadence has become much more elastic recently. The original tick-tock model that was introduced in 2006 predicted die shrinks in uneven years and new architectures in even years. With the a June launch date, Intel is about six months behind its own plan.
The information posted by VR-Zone also reveals a slight change in the sequence numbering system. The models Core i7-4930MX, Core i7-4900MQ, Core i7-4800MQ, Core i7-4702HQ, Core i7-4700HQ, Core i7-4702MQ and Core i7-4700MQ switch their name extensions from "QM" to "MQ". HQ models are new in the lineup.
Well there are currently 26 Ivy Bridge desktop SKUs, so...
(keep in mind many of the chips come in -S and -T varieties too, plus you have chips that only vary on the IGP side)
But let's be very clear on one thing, it couldn't be that Intel is having problems with Haswell, they're delaying it on purpose because AMD sucks. ROFL, yeah right.
But sadly I have this Gen CPU
The 3770K only has VT-x not VT-d though.
Not since the early-mid 1990's has Intel enjoyed this type of dominance. Back then, they deliberately did not release their fastest architecture, the 486 was released years after it was ready, same with Pentium (x586). We might see a repeat of those days now that AMD is not competitive and has no chance of catching up within the next 4 years or so. It could take that long for AMD just to catch up with Ivy Bridge, and I'm talking about matching performance on all levels, not on just select heavily threaded applications.
You're right. I spoke with a Intel engineer and they said they were no longer focusing on speed, and more focused on energy efficiency/reduction. They are only growing CPU speeds at 10% per year right now.
I'm not sure what you were anticipating, but since Sandy launched in Jan and ivy launched in April, i was expecting a July launch. Otellini's surprise announcement then led me to speculate a May-June launch.
Ahaha. No, they're not competing with themselves, their competing with ARM. ARM's behind them on performance, but they're already rivaling ARM in efficiency, and till they dominate ARM here, they're not going to look at raw performance. More evidence for that lies here:
and
http://www.tomshardware.com/reviews/atom-z2760-power-consumption-arm,3387.html
http://www.anandtech.com/show/6529/busting-the-x86-power-myth-indepth-clover-trail-power-analysis
Then i think they're going to look more into graphics, where i suppose they'd compete with AMD in the low-end PC space and Nvidia/PowerVR in the ultra-mobile space.
Who knows, if Steamroller impresses, maybe then they'll switch back to performance.
If that. Ivy was only ~5% increase on CPU over Sandy for most workloads, plus they didn't OC quite as well. Intel's main focus right now is on their iGPU and getting that up to snuff to compete with AMD, and lowering their power usage to compete with ARM. Their next focus is going to be moving to more of an SoC model where pretty much everything except for the feature chips will be moved directly onto the CPU. After that then there may be a chance for us to see them focus on overall performance again with the Sky series of chips.... but that is still some 3-4 years out and by then pretty much all of the focus will be on mobile platforms which can be docked in order to have desktop-like performance, and true custom desktop users (like myself and most of us here) will be an even smaller minority than we already are today.
No reason to rush out the new architecture when there is no real competition for the old one... Intel will become more and more like Intel in the old days, milking the same architecture for years and increase its speed in 100 mhz intervals each half year or soo until there is competition again...
All the low-hanging performance improvement fruits have been picked. Clock rates using low-cost mass-production process and were responsible for the lions' share of the ~60%/year improvement we used to have 10-20 years ago have hit a brick wall at 3.5-4GHz for most of the past 10 years, out-of-order execution, branch prediction / speculative execution, prefetching, etc. have reached a point where making them any more accurate/efficient is becoming extremely expensive so there really isn't much left to improve on in a cost-effective manner.
This is the burden of interactive desktop applications being largely single-threaded and thus heavily reliant on single-threaded throughput.
In the server space, many CPU vendors sacrifice single-threaded ILP in favor of TLP. This allows them to use much simpler/shorter out-of-order queues per thread, forgo speculative execution (branch prediction) and all the extra complexity that comes with those so they can put more resources into increasing SMT throughput by multiplexing 4-8 instruction streams per core, basically hyperthreading on steroids.
In other words, the desktop PC is stuck on a brick wall until desktop applications start making meaningful use of multi-threading. This is where Intel's recent interest and patent in artificial vision as an alternate input method comes in - something to make even the most trivial tasks gobble up a few cores if people want to use it.
intel does tic-toc they used sandy bridge archetecture on the die shrink till they mature the process for good yeilds that was ivybridge. now we have the toc, the new architecture on the 22 mature process, next will be the haswell archetecture on a shrunk node,
The answer is right there in the article:
No indications of performance beyond what we already know though. We won't really know until launch review benchmarks, unless someone breaks NDA.