So expect the most popular Haswell CPUs (i5-4670K and i5-4570) between April and June? I'm assuming 4670K ~= 3570K and 4670 ~= 3450?
Has there been any indication yet as to whether it will be early or late Q2, aside from the "Sources indicate that a number of Haswell CPU's should be released on the 2nd of June" bit?
Invensas puts entire DIMM on a chip
CES 2013: DDR3, DDR4, GDDR5, it doesn't matter to Invensas
http://semiaccurate.com/2013/01/25/invensas-puts-entire-dimm-on-a-chip/
sounds promising. ultrabook cpus as well as amd apus may benefit from tech like this.
Eh, not really. The ACP debate was due Intel rating the Core-based Xeons' TDPs using "typical power" figures rather than using the previous definition of fairly close to maximum power as the Opterons and Netburst-based Xeons used. AMD had also significantly over-rated many of their Opterons' TDPs to make sure that there was plenty of cooling no matter what. Some sleazy vendors had skimped on cooling on some K7 chips and also didn't often implement catastrophic overheat protection, and AMD's chips had acquired a reputation of "burning up" as a result. AMD used ACP to try to make a more apples-to-apples comparison between the Third Generation Opterons and the Core- and Nehalem-based Xeons as Intel was pushing the TDP issue pretty hard. Currently you don't hear anything about ACP as AMD has Core Performance Boost on all Bulldozer-based parts and sets the power limit as equal to the TDP, so ACP can equal TDP if you really whack at the chip. So they dropped ACP.
The Intel "SDP" essentially is that the chip can thermally throttle itself down to an 8 watt power long-term power consumption if you grossly undercool it. The thermal *design* power of the chip is 17 watts or so since that's the amount of power it will draw if you expect it to run at full rated speed. AMD's chips can throttle too but as far as I can tell AMD simply releases grossly underclocked chips that have a design power of whatever watts instead of relying on throttling to keep temps in check in an undercooled system.
de5_Roy :
i am all for gt3 on desktop 😍 . even if it drives up tdp from 84w. amd's top dt apus are 100w so that's not a factor. i was disappointed to see all dt cpus 'rumored' get puny gt2. may be intel got too lazy and decided not to include 40 shaders on the dt silicon.
shaderrage! 🙁
Who knows, it probably has to deal with product line segmentation if history is any guide. You would expect desktop chips to be available with the larger IGP setup as desktops aren't as power- and cooling-constrained and tend to have larger displays. Businesses and HTPC users would do well with a stouter IGP rather than needing to buy a discrete GPU to get adequate desktop performance. I suppose only Intel really knows.
Chad Boga :
Don't think the average punter will get their hands on a CPU made on a 450mm wafer till 2016, it will probably be on the 10nm process.
I would expect the punters to be the FIRST ones to get these chips. Intel's strategy is to introduce new processes on small, midrange and low-end chips. For example, the first chips on 32 nm were the dual-core Clarkdales and the first 22 nm chips were only quad-core at best. The really big chips like Xeon EXes and Itaniums are always the last ones to move to a new process. Shoot, 8-way Xeon EXes are still 32 nm Westmere based, not even Sandy Bridge let alone Ivy Brdge.
I would expect the punters to be the FIRST ones to get these chips. Intel's strategy is to introduce new processes on small, midrange and low-end chips. For example, the first chips on 32 nm were the dual-core Clarkdales and the first 22 nm chips were only quad-core at best. The really big chips like Xeon EXes and Itaniums are always the last ones to move to a new process. Shoot, 8-way Xeon EXes are still 32 nm Westmere based, not even Sandy Bridge let alone Ivy Brdge.
The reason why I specified "average punters", was to distinguish between a commercial release and an engineering sample(that might end up with a non-average punter), I wasn't making an assessment on which segments of the market Intel would target first.
I think the most noteworthy of these is that Intel is actually going to release a Xeon MP (er, EX) when its architecture is still current! I don't believe Intel has ever released a Xeon MP with the same microarchitecture as the current desktop chips. The current EXes are four year old Westmeres for crying out loud. Intel does have the Sandy Bridge-EP E5-4xxx units that are sort of Xeon MPs because they are four-way capable but they aren't the giant die with extra cores/cache that are found with other Xeon MPs, such as the Westmere-EX.
I think the most noteworthy of these is that Intel is actually going to release a Xeon MP (er, EX) when its architecture is still current! I don't believe Intel has ever released a Xeon MP with the same microarchitecture as the current desktop chips. The current EXes are four year old Westmeres for crying out loud. Intel does have the Sandy Bridge-EP E5-4xxx units that are sort of Xeon MPs because they are four-way capable but they aren't the giant die with extra cores/cache that are found with other Xeon MPs, such as the Westmere-EX.
The E5 Sandy Bridges go up to 8C/16T. Not quite the 10C core beasts of past but the architectural changes and the much higher turbo clocks make up for that.
For the E7s it looks like they're skipping to Ivy.
With Xeon Phi and GPGPU taking off the need for 8-way is on the decline.
The E5 Sandy Bridges go up to 8C/16T. Not quite the 10C core beasts of past but the architectural changes and the much higher turbo clocks make up for that.
For the E7s it looks like they're skipping to Ivy.
With Xeon Phi and GPGPU taking off the need for 8-way is on the decline.
I would more say that the increasingly multicore CPUs are killing off the 8+ socket machines, just like they have reduced overall socket count for everything else. Go ask anybody who's sold to the enterprise market for the last 20 years and ask them what the mix of 1P/2P/4P/8P+ was 8-10 years ago before multi-core CPUs came out, compared to what it is today. GPGPU has its niche but it is a niche. For example, they wouldn't do very well in serving VMs or databases.
Or is it your hope that Haswell will (and should) be better than what these numbers show ? 😛
DrWho aka Francois Poedenal has said these numbers arent real, and his numbers are much better.
It could be that these are from an earlier ES, older bios etc etc.
Either way, its early on, they will get better, and yes, I do hope they get better
DrWho aka Francois Poedenal has said these numbers arent real, and his numbers are much better.
It could be that these are from an earlier ES, older bios etc etc.
Either way, its early on, they will get better, and yes, I do hope they get better
This "review" if you want to call it that says nothing except that Haswell is in the ES stage and somebody in Asia leaked one. We know that the first item happened and the last one happens so often that it's barely newsworthy :lol: I would expect more benchmarks that actually say something to occur in the next few months.
If they are launching in June Haswell is passed the ES phase. Companies would be testing w/ QS right now. Actually, maybe that's how this Russian site would have gotten their ES..as the original recipient would be moving onto QS.