Slashdot Mirror
Intel Launches Sandy Bridge-E Series Processors
MojoKid writes "Today marks the release of Intel's Sandy Bridge-E processor family and its companion X79 Express chipset. The first processor to arrive is the Core i7-3960X Extreme Edition, a six-core chip manufactured using Intel's 32nm process node that features roughly 2.27 billion transistors. The initial batch of Sandy Bridge-E CPUs will feature 6 active execution cores that can each process two threads simultaneously via Intel Hyper-Threading technology. Although, the chip's die actually has eight cores on board (two inactive), due to power and yield constraints, only six are active at this time. These processors will support up to 15MB of shared L3 Intel Smart Cache and feature integrated quad-channel memory controllers with official support for DDR3 memory at speeds up to 1600MHz, as well as 40 integrated PCI Express 3.0 compatible lanes. Performance-wise, Sandy Bridge-E pretty much crushes anything on the desktop currently, including AMD's pseudo 8-core FX-8150 processor."
http://www.aria.co.uk/Products/Components/CPUs+%2F+Processors/Intel+Core+i7+(2011)/i7-3960X?productId=47179
So it's very heavy then?
Maybe I'm reading it wrong, but they are bench-marking a $1000 processor against a $300 processor?
$1000 processor wins!
Er, no.. they didn't. It's LGA 2011, a new socket. Again.
Honestly, with a $500+ entry tag plus cooler which is not included plus expensive, low volume motherboard you might want to compare to a dual processor Xeon machine rather than other desktops for some alleged server/workstation stability too. Performance was as expected, 6 cores to 4 so it's faster in well-threaded workstation applications, not that different otherwise.
What's disappointing is the platform, no USB 3.0, two SATA 6 Gbps ports, no SAS support, it seems like PCI express 3.0 made it in but no cards support it yet so there's nothing besides the processor that really screams high end. Well that and 8 memory slots if you feel 4x4GB isn't enough but there's alternatives like the old high end it replaces with 6 slots or 8 GB sticks that have been showing up lately - pricey but you can get 4x8GB for less than one of these CPUs. Don't get me wrong, it's the undisputed performance king but it's like the same car with a souped up engine and fuel system yet none of the features that say this is a $100k Ferrari.
Live today, because you never know what tomorrow brings
That's nice and everything, but I'll wait for Ivy Bridge, which is due March 2012.
According to Wikipedia:
Ivy Bridge feature improvements from Sandy Bridge were expected to include:
Tri-gate transistor technology (up to 50% less power consumption)
PCI Express 3.0 support
Max CPU multiplier of 63 (57 for Sandy Bridge)
RAM support up to 2800MT/s in 200MHz increments
Next Generation Intel HD Graphics with DirectX 11, OpenGL 3.1, and OpenCL 1.1 support
The built-in GPU is believed to have up to 16 execution units (EUs), compared to Sandy Bridge's maximum of 12.
The new random number generator and the RdRand instruction, which is codenamed Bull Mountain.
Next Generation Intel Quick Sync Video
DDR3 low voltage for mobile processors
Multiple 4k video playback
So yeah, just hang on for the die shrink if you care about performance and power consumption. My next system will definitely be Ivy Bridge based.
Trying to future proof an Intel based motherboard is pointless. Considering this new socket replaces the LGA 1155, which is less than a year old, in an ever decreasing release interval, I would estimate this socket will be obsolete by next spring, summer at the latest.
It doesn't replace LGA1155 it replaces LGA1366, which is 3 years old. These chips are server and workstation level chips.
Intel's next desktop architecture is called Ivy Bridge, will be released in the first 3 months of next year, and will be using LGA1155. Ivy Bridge E will use LGA 2011. Only in about 20 months time (by which point LGA1155 will be 3 years old) will Haswell come out on a newer socket.
http://www.pcper.com/reviews/Processors/Intel-Sandy-Bridge-E-Review-Core-i7-3960X-and-X79-Chipset-Tested
http://www.hardocp.com/article/2011/11/14/intel_core_i73960x_sandy_bridge_e_processor_review
http://www.legitreviews.com/article/1773/1/
http://www.anandtech.com/show/5091/intel-core-i7-3960x-sandy-bridge-e-review-keeping-the-high-end-alive
In this instance, no, this is a new socket, replacing LGA1366. Next year though intel will release new desktop CPUs based on the current LGA1155 for desktops.
Intel actually don't release things on new sockets as much as people think. Every tick/tock pairing has one desktop socket, and one server socket, this is the server socket to go with LGA1155's desktop socket. The tock to come (Ivy Bridge) will also use LGA1155 for desktops and LGA2011 for servers.
This is much the same as has happened before: Nehalem introduced LGA1156 and LGA1366, westmere reused them; Conroe introduced (properly) LGA775 and LGA771, Arandale reused them.
(I've always wanted to start a conspiracy theory.)
On the one hand you take life too seriously, and on the other, you do not take playful existence seriously enough. Seth
This is all nice and well, but are there any sites that actually benchmark this CPU under Linux, running some stuff not compiled with intel compiler? AFAIK most of the benchmark software is running on windows is compiled with ICC, and ICC cheats- it disables most optimizations on non-intel CPUs.
How about some linux developer workload? Compile times? IDE performance? Java performance? PHP, Apache, PostgreSQL, MySQL performance? KDE/Gnome performance? CAD/CAM? Matlab or Octave? Bzip2/gzip/SSL/zip under Linux? I know some of these workloads depend on IO/graphics more than on CPU, but I'd like to see results anyway. And I'm sick and tired of reviews that run some Intel compiled synthetic benchmarks and then some games that primarily use GPU anyway. Phoronix is guilty of that as well- they should have more WORK workloads and less FPS counts for games. But at least they are trying- and Bulldozer performance under Linux/GCC isn't that bad compared to Intel CPUs as it is under Windows/ICC.
--Coder
When buying hardware, trying to future proof is dumb. You could try to "future proof" now and buy a $1500 system. In 3 years it'll be shit though.
Alternatively, you could buy a $600 mediocre system now, and another $600 system in 2 years that'll be faster than the above $1500 one. The result will be that you've spent $300 less, you've got machines that are reasonably current for 4 years, and the system you get out at the end is faster.
http://venturebeat.com/2011/11/13/amd-introduces-worlds-first-16-core-pc-microprocessor/
no it doesnt. not when it comes only close with amd's 8 cores in multithreaded apps. 16 cores , becomes unmatchable.
Read radical news here
That's the number of frame per second that the MULTITHREADED VIDEO ENCODER was being able to encode. All of the provided benchmark on that page were about video encoding using multithreaded encoders, and the new proc was beating all others easily. That's why I asked if it was the wrong link perhaps?
What I see is the 2600k, the 4 core $300 chip, matching or beating the Bulldozer, and the 3960X beating everything by a decent margin.
You are correct in that the Bulldozer doesn't have much to worry about from the new E series as they are much higher priced and compete in a different market. What it does have to worry about is the regular SB chips, which are killing it. Even when things are stacked in what should be its favour: Heavily threaded tasks, the SB does as good or better. Then if you take many other tasks that are not as multithreaded, the SB pulls way ahead.
THAT is the BD's big problem... Well that and the fact that the Ivy Bridge comes out in a few months. The E series is just for people with too much money. In the consumer market, the regular SB is an amazing performer.
There are only three things that you can bench it against usefully:
1) The 2500/2600k CPUs that are the high end for the consumer boards. The question there is "What do I get moving up to the much more expensive E series?"
2) The top of the line AMD Bulldozer. The question there is "How much faster is Intel's high end than AMD's high end?"
3) The previous Intel high end, the i7-990X. The question there is "How much faster would it be if I upgraded?"
In all cases, you are talking a very high priced, over spec'd part. There are no other chips in its category really. It is for people who demand the max performance and aren't concerned with the stiff price premium to have it.
Intel actually don't release things on new sockets as much as people think. Every tick/tock pairing has one desktop socket, (..) Nehalem introduced LGA1156 and LGA1366, westmere reused them; Conroe introduced (properly) LGA775 and LGA771, Arandale reused them.
Which means the motherboard is good for exactly one generation on upgrades, on the same die size. Who'd seriously upgrade their almost-new Nehalem system to Westmere, Conroe to Arandale or Sandy Bridge to Ivy Bridge? Then you'd do better putting all that money into one processor. Socket compatibility is only good if it lasts long enough there good reason to upgrade. With Intel, I assume that any new processor I buy will require a new motherboard, simple as that.
Live today, because you never know what tomorrow brings
I've got a Z80 right here. I can't imagine how fast a Z68 must be!
Did they finally break that 65536 bytes barrier?
The result will be that you've spent $300 less, you've got machines that are reasonably current for 4 years, and the system you get out at the end is faster.
You're also sending twice as much to the garbage pile.
I know this isn't a consideration for most, and it's all but encouraged through the new "disposable electronics" thing that's crept up over the last decade, but at some point we need to consider that some considerations extend beyond the financial, even when talking about buying consumer goods.
For instance, I know people that buy a new printer every time their starter ink runs out because it's still cheaper than buying replacement ink cartridges. Three times a year they're throwing a perfectly good printer into the trash. Yeah, it saves them money, but does that really make it right to throw it in a landfill? I have a hard time saying yes.
Maybe if we required manufacturers to subsidize the disposal of their goods when such goods are non-biodegradable it would help do something to eliminate the whole "designed for the dump" phenomenon?
No –it means that the motherboard is good for exactly one generation on upgrades at a smaller die size.
It's also fairly common for people to buy a low end version of the tick, and then a high end version of the tock.
Socket compatibility is only good if it lasts long enough there good reason to upgrade.
Agreed, unfortunately, that's almost never, no matter who your CPU manufacturer is.
You're also sending twice as much to the garbage pile.
Not that your point in general is wrong, but you're actually sending about 150% to the garbage pile, not 200%. One machine every 2 years, instead of one machine every 3.
No, the i5 is $150 to $200. The around $300 versions are the i7 2600, 2600k, 2700, 2700k and 3820, the 2600k and 2700k have unlocked multipliers, the 3820 has a partially unlocked multiplier.
The result will be that you've spent $300 less, you've got machines that are reasonably current for 4 years, and the system you get out at the end is faster.
You're also sending twice as much to the garbage pile.
I know this isn't a consideration for most, and it's all but encouraged through the new "disposable electronics" thing that's crept up over the last decade, but at some point we need to consider that some considerations extend beyond the financial, even when talking about buying consumer goods.
For instance, I know people that buy a new printer every time their starter ink runs out because it's still cheaper than buying replacement ink cartridges. Three times a year they're throwing a perfectly good printer into the trash. Yeah, it saves them money, but does that really make it right to throw it in a landfill? I have a hard time saying yes.
Maybe if we required manufacturers to subsidize the disposal of their goods when such goods are non-biodegradable it would help do something to eliminate the whole "designed for the dump" phenomenon?
A machine that was bleeding-edge two years ago is still quite powerful today for the majority of people out there. Also, not every server in the rack has to be equal. There are plenty of less-demanding but still important roles that two year old machine can fill when it is kicked down a notch. I'm sure the "weakest link" hardware can be put to good use elsewhere when upgrade time rolls around. I know I consider the mobo-CPU combo as a unit now, rather than thinking "I can upgrade the CPU later". Maybe I can and maybe I can't, but it doesn't matter that much. So long as I have room to boost RAM and storage, I can extend the useful life of the hardware a great deal. It just may not be my fastest, l33t3st system any more. At worst, I can give the machine away -- my 3 year old secondhand hardware is generally as good as most people would buy off the shelf new, and I already have a good idea what it does best.
How is the Riemann zeta function like Trump rallies? Both have an endless number of trivial zeros.
You're also sending twice as much to the garbage pile.
Only if you actually throw the machines away. If anything, it's the reverse. If you upgrade components of a machine, it's much harder to find a use for the bits you remove than if you replace the whole machine. A two year old machine may be underpowered for you, but there are lots of people who can use it for another few years.
I am TheRaven on Soylent News
> When buying hardware, trying to future proof is dumb. You could try to "future proof" now and buy a $1500 system. In 3 years it'll be shit though.
That's total nonsense.
a) The gaming rig I built in 2001 lasted until 2008 -- upgrading the video card from a GeForce 2 to GeForce 4 to GeForce 6600GT kept it alive much, much longer.
b) I just priced out a complete gaming rig for a friend based on what I have. For $1300 you can build a gaming rig that WILL be perfectly fine for gaming in 3 years. By that time, you can upgrade the video card and it will play all the latest games.
You _do_ know that MOST games are GPU bound at 1920x1080, not CPU bound right?
Here are the benchmarks to back that claim up:
Sabertooth 990 FX
http://www.guru3d.com/article/asus-sabertooth-990fx-review/17
And proof that GPU's are the bottlenecks in the latest games ...
http://www.tomshardware.com/reviews/battlefield-3-graphics-performance,3063.html
I could go on, but there would be little point.
So go ahead, and blow $500 on that latest i7-2600K -- meanwhile us budget AMD guys will be putting money that we saved on buying the cheap 4-core system towards a high end GPU like the 6970, because you are forgetting one tiny, but important fact. Almost ALL the big PC games are designed to run on the 5 year old console hardware -- the PS3 and XBox 360 only have ~6 core and ~2 cores respectively, which means modern CPU's are NOT the bottleneck -- the GPU's are.
LGA2011 really replaces LGA1567 for the Xeon MPs. Intel's on-again, off-again planned replacement for LGA1366 is actually LGA1356. However, it seems like Intel may somewhat simplify their socket lineup and just have LGA1155 for general uniprocessor platforms and LGA2011 for performance UP and all multprocessor applications.
Just "gittin-r-done," day after day.
It ain't (necessarily) an Intel conspiracy to force you to fork out more cash for a new socket when an old one might have worked, since they'd simply be leaving the field open to AMD if that were the only reason (AMD's greatest selling point is the ability to leverage on previous generations). The reasons are technical - when a die undergoes die shrinks, there is also less area for the same number of signals, making it pad limited. Also, die shrinks need more power and ground pins even as the real estate available for such pins are reduced. As a result, the newer processors sometimes may have to undergo an I/O interface overhaul, in order to accommodate the same essential signals into fewer pins, and also drop any pins that can be eliminated.
That is what one is seeing above - when Intel went from 1156 and 1366 to 775 and 771, reason for it was undoubtedly a reduction in the die size, and hence the number of pads that could be connected to the package. It doesn't make sense to retain the same old package, b'cos the extra 400-600 odd pins that are not being used simply add to package cost, which then percolates downstream. As a result of these changes, Intel is able to offer a complete system cost down compared to previous generations, albeit w/ different form factors. Of course, the price for that is that you can't insert a Conroe into a Nehelam socket.
One may then wonder - how come the i7 comes in a larger package w/ 2011 bumps? Chances are that using 6 cores this time has increased the die size, eliminating the pad limitations, while the extra Vdd and Vss lines are needed, or that it is 6 different processor dies on a single multi-chip package. Either way, the larger package is required.
1156 and 1366 to 775 and 771
You got those backwards. 775 and 771 were the old sockets, spanning from late P4 all the way up to Core 2. Socket 1156 was for first-gen i3/i5, ans Socket 1366 was Nehalem (i7-9xx and extreme). They have never reduced the pin count, because each successive platform has introduced wider memory controllers and/or QPI/DMI/PCIe lanes. Sandy Bridge-E is no different, moving up to 4-channel DDR3, up from X58's 3 channels. That right there is an extra couple-hundred pins just for the memory.
-Billco, Fnarg.com
If you posted that using your Z80, THEN I'm impressed. ;)