Slashdot Mirror
Intel Announced 8-Core CPUs And Iris Pro Graphics for Desktop Chips
MojoKid (1002251) writes "Intel used the backdrop of the Game Developers Conference in San Francisco to make a handful of interesting announcements that run the gamut from low-power technologies to ultra-high-end desktop chips. In addition to outing a number of upcoming processors—from an Anniversary Edition Pentium to a monster 8-core Haswell-E — Intel also announced a new technology dubbed Ready Mode. Intel's Ready Mode essentially allows a 4th Gen Core processor to enter a low C7 power state, while the OS and other system components remain connected and ready for action. Intel demoed the technology, and along with compatible third party applications and utilities, showed how Ready Mode can allow a mobile device to automatically sync to a PC to download and store photos. The PC could also remain in a low power state and stream media, server up files remotely, or receive VOIP calls. Also, in a move that's sure to get enthusiasts excited, Intel revealed details regarding Haswell-E. Similar to Ivy Bridge-E and Sandy Bridge-E, Haswell-E is the 'extreme' variant of the company's Haswell microarchitecture. Haswell-E Core i7-based processors will be outfitted with up to eight processor cores, which will remain largely unchanged from current Haswell-based chips. However, the new CPU will connect to high-speed DDR4 memory and will be paired to the upcoming Intel X99 chipset. Other details were scarce, but you can bet that Haswell-E will be Intel's fastest desktop processor to date when it arrives sometime in the second half of 2014. Intel also gave a quick nod to their upcoming 14nm Broadwell CPU architecture, a follow-on to Haswell. Broadwell will be the first Intel desktop processor to feature integrated Iris Pro Graphics and will also be compatible with Intel Series 9 chipsets."
So they finally caught up to AMD.
Does anyone else find it kind of weird that Intel seems to have gotten into a pattern where their supposed top of the line CPUs are perpetually a generation behind their supposed commodity CPUs in terms of technology?
Finally! I have been waiting for next gen Iris graphics since like forever!
There's no reason for most programs to be 64bits. Most programs don't need to address that much RAM nor do they need the additional registers that you get with 64bit processors.
Now for programs that use massive amounts of RAM or need the additional registers, going 64bit makes sense, but it's silly to suggest that there's something wrong with 32bit programs in general that would be fixed by moving to 64bit.
You asked for it, you got it! Though the downside is these two fast cores don't include AVX, AVX2, or a few other instruction sets.
(T>t && O(n)--) == sqrt(666)
The eDRAM simply makes the chip way too expensive.
If you look at the price of i7 core 4770R: $358. It's an i7 but has only has 6 MB of cache (compared to the 8mb of the regular i7 4770). So basically, it's about the same value as a i5-4670K which cost $243. With the price difference you could buy a Radeon R7 260X, which will trash Iris Pro in performance.
AMD's Bulldozer cores have Clustered Integer Core which has two true ALU "cores" and one shared FPU. For integer instructions this is two true cores and not "hyper-threading". For FP instructions this is "hyper-threading" and why Intel has been regularly handing AMD it's arse in all benchmarks that aren't strictly ALU dependent (gaming, rendering, etc). AMD's FPU implementation, clock for clock, is a bit weaker on most instructions as well. And yes, the FPU _is_ shared on AMD processors.
EMT64 is not "32 bits on each 1/2 of the clock cycle". That doesn't even make any sense. EMT64 is true 64 bit. x86-64 does have 32 bit addressing modes when running on non-64bit operating systems. This is part of the x86-64 standard and hits AMD, Intel and VIA.
Hardware Queuing Support is part of the Heterogeneous System Architecture open standard and won't even be supported in hardware until the Carizzo APU in 2015. Since this is an open standard, Intel can chose to use it.
Both architectures have shared caches.
WTF does nVidia's IEE-754 compliance have to do with Intel vs AMD?
I'm not an Intel or AMD fanboy, I try to use the right one for the job. I prefer AMD for certain work loads like web servers, file servers, etc because they have the most integer-bang for the buck. If I'm doing anything that involves FP, I'm going to use an Intel Chip. Best graphics solution?... yeah, I'm not even going to go down that hole.
I just did a ps -e | wc -l and got 245. Maybe most of my processes are only single threaded but since there's 245 of them I'm glad my processor has 8 hardware threads to handle them.
You'll have 16GB unbuffered DIMMs so you aren't losing anything. With Haswell-EP using LR-DIMMs allows 3 per channel for 768GB per CPU.
The few times I'm ever waiting on CPU, it's multi-threaded. Video transcoding, occasionally compiling. I can't remember the last time I heard of a game being CPU bound - that's always GPU-bound these days.
Socialism: a lie told by totalitarians and believed by fools.
The pro has a +10 core if you max it out.
http://saveie6.com/
It has a pci express which is several multitudes faster as it is directly on the PCI bus. It is rated for over 700 megs a second.
http://saveie6.com/
Silicon tops out at ~ 5 GHz.
Germanium X tops out at ~500 GHz.
The average consumer doesn't give a rats ass about GHz, which means that you will never see cheap 10 GHz CPUs anytime soon.
Hell, we're STILL waiting for Knights Corner / Landing 48+ core CPU to ship to the general public.
Really.
Ok, but aside from the n% increase over the n% increase over the n% increase over the n% increase, what has Intel done for us?
Intel makes the 2 fast core processor right now, today, and it'll cost you a staggering $120 to $150. It's called the Haswell Core i3 and each of its cores is faster than any of the cores in your $5000 machine from 2007. It will run Dwarf Fortress faster than anyone would have imagined back then.
Of course there's no limit to what you'd like, but if you have a problem with the amazing shit you can buy today, the problem is with you, not the amazing shit.
As copyright owner of this comment, I authorize everyone to defeat any technological measure which limits access to it.
"You are full of outdated rubbish."
Please, GMABooster is already preparing to unlock the underclocked Iris GPU.
Because this is what Intel does. Use it at max power, show performance, then underclock the shit out of it and ship it off.
The 950 would have been competitive if it had shipped at its stock 400 MHz (minus not having hardware T&L) instead of the fucked 166MHz the drivers forced at low level.
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
Make that 798Ghz
"When we tested the IHP 800 GHz transistor at room temperature during our evaluation, it operated at 417 GHz,"
http://en.wikipedia.org/wiki/S...
Waterfox - a Firefox fork with legacy extension support, security updates and better privacy by default.
Memory is fairly cheap (though if you really want lots of memory in addition to the cost of the memory itself you have to consider the cost of the platform to accomodate that memory) but cache, particulally the lower levels of cache that are closest to the CPU isn't so cheap, if you have a pointer-heavy workload (e.g. data structures that are mostly cross-references implemented using pointers) then you can fit a lot less of your workload in cache with .64-bit pointers.
For java (which is very pointer heavy) this was so bad that they implemented tricks in the VM so that if the java heap size is below a certain level then "ordinary object pointers" can be stored in 32-bits.
On linux there is something out there called x32 which aims to combine the advantages of the 64-bit processor mode with the advantages of 32-bit pointers but it's unclear if it will ever catch on.
you can fit 32Gb+ in a consumer grade desktop system.
If you had left off the + then you would have been more on the mark. The largest reasonablly available modules of desktop memory are 8GB and the most slots you will find in a mainstream desktop board is 4 (and many boards only have two). If you want to go over 32GB you have to move to the high end desktop platform (and if you want to go over 64GB you have to move to workstation/server platforms).
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register