What is Digg?Sponsored by Travelzoo
$52 and Up-Airlines Slash Fares On Peak Holiday Flights. view!
travelzoo.com - This year, waiting until the last minute is NOT the best strategy. See why.
53 Comments
- AdmiralAcbar, on 05/19/2009, -0/+27You accidentally put the words "intel" and "high end GPU" in the same sentence.
- inactive, on 05/19/2009, -0/+24Octo-core!
- bratterscain, on 05/19/2009, -0/+24Because it's not everything. It's like using a wheelbarrow to haul dirt. Instead of tweeking the guy up on speed, they just increase the wheelbarrow size, not how fast the guy hauls which can be more efficient and get more work done if done right. I've always likened it to that.
Also, to answer more direclty, with the current implementations, you can only go so small until you hit a speedbump. In this case, it's to do with what I've heard is like crosstalk. Electrons behave more erratically at faster speeds so unless we use some other implementation such as fiber optics, there will be limits. - bratterscain, on 05/19/2009, -0/+17EX stands for extreme, which means it kicks more ass. Kind of like the stickers on ricers, it adds another 5mph.
- anexanhume, on 05/19/2009, -0/+12Because heat output and energy usage increase with frequency. Current processors do more in each clock than previous processors. That's why they are faster without running at a higher clock rate. Cranking the MHZ up is a cheap way of getting more speed out.
- krisrm, on 05/19/2009, -0/+12Adding "EX" to the end of product names doesn't automatically make them cool...
...adding 4 processor cores, however... :D
(pre-emptive response: yes, it can run Crysis) - robbob, on 05/19/2009, -0/+12"The chip itself will not actually ship in systems until late 2009 or early 2010"
- WDAndrew, on 05/19/2009, -0/+9Hell yea! I need my extra speed boost from the aerodynamic stickers! :D
- WDAndrew, on 05/19/2009, -1/+10I think he means we are only reaching 8 cores by now. Quad-cores launched in 2003; 6 years ago. :)
- ultrafez, on 05/19/2009, -0/+9But can it run Doom?
- RogerMcDodger, on 05/19/2009, -0/+7Still? These are the first x86 8 core processors.
- WDAndrew, on 05/19/2009, -0/+6Great analogy, I'll have to keep this one in mind :)
- twitchster, on 05/19/2009, -1/+6anexanhume is right.
Higher clock frequency means higher energy use. This causes a by-product of heat. Because the processors pack transistors so closely heat becomes a major issue.
Collution is probably remembering the P4 which ran at speeds approaching 4Ghz.
The P4 was HOT, and used alot of power for the processor of that size. Intel essential made a very fast processor that did as little as possible during each cycle to achive the fast speeds.
Intel was primarily focusing on the consumer market and using the ignorance of the average Joe. Equating clock speed to performance.
This was not so.
This allowed AMD to take the performance crown from Intel. By producing a processor that had a shorter pipe line. AMD had processors that essentially did more calculations per clock cycle.
Combine the amount of work done with clock speed and you get your performance.
The industry has found itself in a bit of a pickle. The current chip fabs cannot produce chips reliably at very high clock speeds, without being cost pro-hibitive.
Instead of going the IBM direction with vast cell processors (EX: PS3)
AMD and Intel are compromising. Delivering multiple core proccessors, that do more per clock cycle. At slower speeds.
This gives us increased perfomeance, in no way does it impact moore' law, transistors on a chip are still doubling every 2 years.
What it does do is impact how programmers design thier software. And it does add more work. A game programmer would have to write his program to be aware that there are multiple processors or cores, and break up different tasks to take advantage of the processors.
This lengthens development time.
But for day-to day operations, Multiple cores deliver vastly improved performance over a single core machine.
All of the programs you run, your OS, AIM client, Web browser, etc spread out among the processors. Rather than having to wait in line for each other.
So yeah, long answer but that's the slightly non technical answer. - WDAndrew, on 05/19/2009, -0/+5Yeah... six years to go from quad to octo seems a tad long, but maybe both AMD and Intel are taking precautions in developing these chips as they are going to hit servers first.
- benologist, on 05/19/2009, -0/+4Intel's GPUs are very high end.
For digital picture frames. - collution, on 05/19/2009, -3/+7Question for the tech heads: how come clock speeds aren't increasing?
- cky3, on 05/20/2009, -0/+4They better nickname it "The Ocho".
- WDAndrew, on 05/19/2009, -0/+4Can't wait to see benchmarks between this Intel chip and the AMD Hydra chip. Also can't wait to utilize an 8 core chip on my desktop PC :)
- RogerMcDodger, on 05/19/2009, -0/+4EX stands for expandable.
- tuka, on 05/20/2009, -0/+4Finally, a CPU for Octo-mom...
Sorry for bringing her up :( - alarchy, on 05/19/2009, -0/+4This will be awesome for vSphere. 32 logical cores in a dual CPU server box? Yes please.
- nullcodes, on 05/19/2009, -5/+8Intel, make me one with 4 CPUs and 4 high end GPUs on the chip.
Let me know when it's ready .. thanks. - anexanhume, on 05/19/2009, -0/+3There isn't any theoretical problem with electrons moving faster in and of itself. First of all, in the case of higher clock rates, the electrons aren't moving that much faster, if it all (it depends on lattice collisions, which can increase with more potential behind them). In most of these cases, higher frequency comes from higher voltage, which increases electron leakage (electrons punching through the gate of the device rather than going through the channel), and you'll also get more lattice collisions, as stated. Getting the correct logic is more about current (number of electrons in a given time), than electrons moving faster. The technologies we are developing, such as graphene, have higher electron and hole (positive electron "receptor") mobility, which means you can move more of them at once. This will allow faster clock rates without having to crank up the voltage, which produces the heat and leakage we don't like.
As for the wheelbarrow analogy, what we've effectively done is both increase the wheelbarrow size (better architectures) AND multiply the number of wheelbarrows we can use at once (more cores, threads), rather than just getting a guy that can push the one wheelbarrow faster. - endoffile, on 05/19/2009, -1/+4I can't wait to come by your house and steal your desktop PC, seeing as how the new CPU will probably cost 3000$ or something.
- nextekcarl, on 05/19/2009, -0/+3Neither Linux nor Windows really hand more than 4 cores very well right now, so I think it may have also been a bit of chicken and egg problem. What's the point if most software used doesn't support the hardware very well, and that's just the OSes. The actual programs run on those OSes don't support more than one core very well (in many cases) yet. This is all changing, but sometimes it can be a difficult development issue, as it is another (potentially massive) sort of bugs that can be very hard to track down.
- hkfczrqj, on 05/19/2009, -0/+38 cores are nice and all, but will the memory bandwidth be enough for such a device? I'll wait until I see numbers.
- WDAndrew, on 05/19/2009, -0/+3It can run one million Doom instances at once! :)
- EricDraven, on 05/20/2009, -0/+3Correct.
To be strict, this is news to people who obviously have not heard of older 8-core implementations like Sun's UltraSPARC and IBM's POWER6 MCM from some time back. - EricDraven, on 05/20/2009, -0/+2To be strict, if this is news to you, you obviously have not heard of older 8-core implementations like Sun's UltraSPARC and IBM's POWER6 MCM from some time back.
- bbarker, on 05/20/2009, -0/+3I think AMD is competitive when cost is considered. They just can't compete at the high end.
- Cogboy, on 05/20/2009, -0/+2This is for quad socket. (as the article says, "more than 2")
- linagee, on 05/20/2009, -0/+2EX stands for EXtra cost.
- Vivifyer, on 05/20/2009, -0/+2Am I the only one that loosely holds onto the small threads that are hope, that AMD gets back in the game? :(
- frostbyt, on 05/20/2009, -0/+1I am holding out until the 20 core procs are out.
- Cogboy, on 05/20/2009, -0/+2For most people x86 is the only thing that matters. Then again, most people don't have quad socket servers.
- bratterscain, on 05/19/2009, -0/+2anexanhume, yeah, that's more accurate. My understanding of cpu tech is mostly basic.
- Culyt, on 05/20/2009, -0/+1Good luck with that when 128 cores are around and every button, scroll bar, loop and so on runs in its own processor thread.
- linagee, on 05/20/2009, -0/+1We = intel? Other companies have had more than 8 cores for years. But they are megabucks. (Sun)
- DrDabbles, on 05/21/2009, -0/+1Linagree, You're still thinking of memory controllers in the old sense. The new Nehalem/Corei7 systems, and all future chips in this architecture use NUMA- Non-Uniform Memory Access. Each core has access to the memory bus, which decreases access latency. In addition to that, the memory has been scaled from dual-channel to triple-channel and the speeds increased.
Memory is no longer being measured in MByte/sec or GByte/sec on this new architecture, but rather in transfers per second. These can be translated into data-per-second, but the number is foolish high. In fact, three DDR3-1600's clock in somewhere around 32GByte/sec- which is the speed of the on-die CACHE for many core2 duo CPUs.
In short, yeah, triple channel memory can handle whatever you're tossing at it CONSIDERABLY faster and more efficiently than previous generation architectures. - WDAndrew, on 05/19/2009, -0/+1I completely agree, but six years is a very long time for development on the software and hardware levels; so you would "think" by now multi-core chips would be utilized efficiently by now... 8 Cores will be useful for servers long before they're useful for desktops though.
- linagee, on 05/20/2009, -0/+1I think this is wrong. If you put the memory controller inside of the CPU, it makes it harder to do multiprocessing (I think). They may have moved the memory controller outside of the CPUs on this one. (Unsure?) If they did move it out, this might mean more than 3 channels or less than 3 channels.
- Cogboy, on 05/20/2009, -0/+1Triple-channel DDR3? yeah I wouldn't be too worried.
- pcpimpster, on 05/19/2009, -2/+2Imagine everyone standing in line at the movies, waiting to get a ticket.
There is only one line that is really ***** fast but still can't handle the load as the line just keeps getting longer.
Add more lines (cores) of equal/lesser value to handle the repetitive load & you just fixed the problem.
Then the guy in line one doesn't have to work so hard & he can't act like that dick in that promotion movie with stiffler
- damnshoes, on 05/19/2009, -2/+2I want to steal your SOUL!
- WDAndrew, on 05/19/2009, -1/+1I highly doubt it'll cost $3,000... Maybe about $1,500 at most. Also, be my guest, I wont have an 8 core in my desktop until it's useful to me and what I do. :P
- electrofreak2k7, on 05/20/2009, -1/+1I am holding on... but not to optimistic about it. If I were to build a system soon, I would build it using an Intel... and I've always been an AMD guy. These days... AMD just barely competes. But I still hold out hope that they will come back to the top.
- terry193, on 05/25/2009, -0/+0...hmmmmm perhaps this will finaly be enough to power vista!!!!! :D
- yohannh, on 05/20/2009, -0/+0But can it run Space Invaders?
- andy3109, on 05/20/2009, -1/+1I think 4 CPU's is enough for me for the next 5 years thanks.
-
Show 51 - 54 of 54 discussions
© Digg Inc. 2009
— Content created and posted by Digg users is