Intel Core I7 Launched, Nehalem and X58 Tested 194
MojoKid writes "Today marks the official launch of
Intel's new Core i7 processor, the most major overhaul of Intel's core processor architecture since the release of their Core 2
design. As has been reported, the Core i7 is a major departure from
Intel's aging Front Side Bus architecture of old, now replaced by
Intel's QPI (Quick Path Interconnect) serial links. This 20 lane
bi-directional (40 lanes total) point-to-point connection provides 6.4 GT/s of
bandwidth and scalability for future multi-socket designs as well. In
addition, the Core i7 now has an integrated triple channel memory controller
offering over 3X the bandwidth of the previous Core 2 architecture with
DDR3 system memory. Though the product is set to ship in volume later this
month,
the early benchmark numbers show Intel's new chip is markedly faster
clock-for-clock versus their previous generation CPU and much faster than
anything AMD has out currently."
Not out... (Score:5, Insightful)
Re: (Score:2, Funny)
You didn't see the bundle offer for one of these with a copy of Duke Nukem Forever?
Re: (Score:2)
The timing would be just about right....
http://www.bestbuy.com/gunsnroses [bestbuy.com]
Re:Not out... (Score:5, Insightful)
The AMD 3-core processor bug was one year ago (Score:4, Informative)
Re:The Pentium 1 floating point bug was 15yrs ago. (Score:5, Insightful)
Every time a brand new processor architecture comes out there are either errata, unforseen shortcomings, or more often both. It's always a good idea not to adopt a new architecture immediately. Let them work the kinks out over the first few steppings.
You obviously know little about processor design nor how many times over the past two decades new architectures have shipped with bugs or design flaws.
Re: (Score:2)
I got my Yorkfield from a kit-builder long before Newegg actually had them in stock. Call around. Someone out there ordered 5 or 6 more than he had plans for.
Sweet! (Score:4, Interesting)
A little hot, but on time, in time for Christmas and slamming the benchmarks. Hey, there is a system that can run Crysis with all the features turned on!
Maybe a price break on the LGA775 quad lineup now please?
Re:Sweet! (Score:5, Informative)
Re: (Score:2)
This isn't too shocking. With the move of the mem control into the i7, there's an extra latency that has to be programmed for/around in video drivers/cards. I suspect that once updated drivers/cards start flowing, the performance of these games on i& will roll back to the top of the pile.
Re: (Score:2)
Yes, but some of the nifty new feature of the core i7 is the new interconnect and triple channel ram. It should have *a lot* of memory bandwidth.
Re:Sweet! (Score:5, Interesting)
But it doesn't magically increases RAM bandwidth.
i7 memory interconnect would help applications which are not hand-crafted to maximize performance. And I expect that games like Crysis already optimized through the nose to utilize all bandwidths to max.
Or to put it in other words: unoptimized code would gain from i7 more than highly optimized code, since in former case CPU would have more opportunities to optimize memory accesses on its own and better fill up the data bus.
But I also can be wrong and hand crafted code of Crysis/etc is simply cannot take advantage of i7 features.
Re:Sweet! (Score:5, Funny)
Or to put it in other words: unoptimized code would gain from i7 more than highly optimized code, since in former case CPU would have more opportunities to optimize memory accesses on its own and better fill up the data bus.
I see!! You mean Vista might actually run well on this processor??!!
Re:Sweet! (Score:5, Funny)
Re:Sweet! (Score:4, Informative)
3 if you count FarCry which was developed by Crytek but published by Ubisoft rather then EA. It's also worth pointing out that IMHO FarCry was the better game.
Please stop using the GT/s performance indicator. (Score:5, Insightful)
Re:Please stop using the GT/s performance indicato (Score:5, Insightful)
Re:Please stop using the GT/s performance indicato (Score:5, Informative)
What is a GT/s? (Honest question, looking for an honest answer.)
Giga-Transfers per second [tmworld.com] (or at least that's what google found).
Re:Please stop using the GT/s performance indicato (Score:4, Funny)
What is a GT/s? (Honest question, looking for an honest answer.)
Damn, if you had been looking for a biased answer I'd have linked you to Wikipedia...
Re: (Score:2)
That's only if he wanted a liberal bias. If he wanted the God's-honest-bias you need to send him to Conservapedia.
Re:Please stop using the GT/s performance indicato (Score:5, Funny)
Re: (Score:2)
> What is a GT/s? (Honest question, looking for an honest answer.)
On Nehalem, bus is 20 bits wide. 1 GT/s = 20 * 10^9 bits/s = 2.5 * 10^9 bytes/s in one direction (link is bi-directional)
Re:Please stop using the GT/s performance indicato (Score:3, Interesting)
No bits or bytes anywhere to be seen.
Re: (Score:2)
WTF?
Quantities of what then? Digital data breaks down to bits, regardless of how it's transported.
If what you say is literally true, then they have to be converting the digital to an analog signal or there's no data either.
Better explanation required.
Re: (Score:2)
I think the goal is to make the transfer mechanism not care what physical data is sent over the line, much like the physical layer of the OSI model, and to allow the CPU or other handling
Re: (Score:2)
so if you use an encoding that transfers 1 TB of data per packet versus an encoding that transfers 1 bit of data per packet the Quick Path Interconnect will transfer the same number of packets per second for both? wouldn't that mean the total bandwidth would change depending on the encoding?
the HotHardware article states that:
which makes more sense than being able to magicall
Re: (Score:2)
But the bus doesn't transmit bits or bytes always.
You really don't know what you are talking about. A bus ALWAYS transfers bits. ALWAYS.
GT/s is only half of a useful metric - it's analogous to talking about CPU frequency without mentioning what specific CPU. The other half of the metric is the width of the bus. Put them together and you can get the actual bandwidth of the bus - in bits and bytes - of the bus. For example, 6GT/s on a 64-bit bus is 48Gigabytes per second.
Re:Please stop using the GT/s performance indicato (Score:2)
But giga-transfers is a more accurate way to term it, especially when talking about something like the QPI (Quick-Path Interconnect), which is a point-to-point packet-based interconnect similar to AMD's Hypertransport. Each packet is going to involve some number of transfers worth of header information, and some number of transfers worth of data, and these are going to change based on the nature of the transaction. So the actual amount of GB/s of real data you get is going to depend on the nature of traff
Re:Please stop using the GT/s performance indicato (Score:2)
Only the highest-end processor, Core i7 965 Extreme Edition, supports QPI at 6.4 GT/s. The 920 and 940 support only 4.8 GT/s. So the amount of bandwidth on a QPI link is:
This is comparable to the HyperTransport bandwith of a Phenom processor (4.0 GT/s), which AMD is supposed to scale to
i7? (Score:5, Funny)
Of course, "Core 3" was what everyone expected them to do, so Intel couldn't possibly use that. Using imaginary numbers is much more logical.
Re: (Score:2)
Wouldn't that make it "Core 7i" instead of "Core i7"?
Re: (Score:2, Funny)
Re: (Score:2, Funny)
But it defies the common notation. I think "Core 0+7i" would be unambiguous. What the hell is wrong with those marketing types?
Re: (Score:2)
But 7i makes it sound like the processor is fuel-injected, which is also not the case.
Re: (Score:2)
Re: (Score:2)
i is used by mathematicians, physicists, and other riff raff.
The EE crowd would call it Core 7j.
Re: (Score:2)
So a Beowulf cluster would have negative 49 cores?
Re: (Score:2)
I was assuming each core was i7.
Re:i7? (Score:5, Funny)
Well... I was trying to *IMAGINE* a Beowulf cluster....
Why would you expect Core 3? (Score:5, Funny)
Core
Core Duo
Core2 Duo
The correct answer should be the 2Core2 Duo, or the Core2 Duo Dos, or the BiCore2Duo. Maybe the DuoCore2 Duo? Anyway, follow the pattern- keep adding things that mean "2." In several years, we should have had BiDuo2Core2DoubleDuo Dos MarkII.
Instead, it looks we're heading for the e8, or the pi9, or the ln10, or maybe the 11!. Except for that they'll change the pattern again, because now everyone's expecting math terms.
Re: (Score:2)
Because that's not the progression.
The technology is Core -> Core 2.
The "Duo" indicates that there are two cores of the appropriate type (Core or Core 2). (And the alternative "Quad" indicates four cores, and "Extreme", oddly enough, is used for 2 or 4 cores, but indicates better support for overclocking.)
So, in terms of the part of the branding used to indicate the core technology, Core 3 would be a not
Re: (Score:2)
And, there's also a Solo moniker.
But, Core 2 (or, rather, the Core microarchitecture that Core 2 is based on) is as big of a leap over the ancient P6 (from 1995) that the original Core Duo and Core Solo were based on. (Core Duo essentially being two Dothan Pentium Ms sharing a cache, with better SSE support, and a die shrink, and Core Solo being the single-core version.)
Re: (Score:2)
Er, I screwed up. The Core microarchitecture was as big of a leap over P6 (or bigger) as Nehalem is over Core.
Re: (Score:2, Funny)
I think they were on the same path you were thinking. Then they decided to have a naming scheme that didn't include confusion.
Core (1) = 1
Core Duo (1 + 2) = 3
Core 2 Duo (1 + 2 + 2) = 5
2 Core 2 Duo (1 + 2 + 2 + 2) = 7
Thus i7
Re: (Score:2)
Re: (Score:2)
But... (Score:2, Funny)
Re: (Score:2)
Comment removed (Score:5, Interesting)
Re:We're all serialists now? (Score:5, Informative)
Comment removed (Score:5, Informative)
Re: (Score:2)
FB-DIMMs are dead. Intel is dropping them in favor of QuickPath.
FB-DIMMs use 10 parallel paths (14 in the other direction). RDRAM used 4, 8 in later version (for more performance).
I have no idea where people are getting the idea parallel is dead. Most these busses mentioned are parallel.
Re: (Score:2)
Gigabit ethernet uses 4 parallel links. PCIe uses 1,2,4, 8 or 16 parallel links. QPI uses 20 parallel links.
Serial just isn't fast enough for things like RAM, PCIe or QPI.
Re: (Score:2)
Ob car analogy: Picture a 4 lane highway. Every clock you can send 4 regular cars (serial) or one four-lane-wide monster car (parallel). Queue the Hummer jokes.
Re:We're all serialists now? (Score:4, Insightful)
"Remember Rambus? And all the rigamarole that surrounded it? Faster but more expensive didn't work out in that case."
There was nothing wrong with Rambus technology that caused it to ultimately fail. It was the lawsuit happy tactics of Rambus Inc. that caused the problems. The technology was sound, but the owner of the patents went out of their way to repeatedly shoot themselves in the foot.
Re: (Score:2, Insightful)
I don't think the crappy Rambus controller on the Intel i820 chipset helped it's technical reputation too much, but you're right that the legal shenanigans probably damaged them to most.
Being an innovator not always smart? (Score:5, Insightful)
Comment removed (Score:5, Interesting)
Re:Being an innovator not always smart? (Score:5, Insightful)
I thought HyperTransport was developed as open technology, allowing anyone to use it. I thought it was one of AMD's advantages, and I can't believe it took Intel so long to ditch the traditional FSB. What hurts AMD is pushing release dates back over and over again. What hurts AMD is not being able to keep up with Intel's fab processes. What hurts AMD is Intel using illegal tactics to bump AMD out of the market. AMD decides the only way to stay in the market is to sell their procs super-cheap, but then they don't make any money doing so.
It didn't help that when AMD was kicking Intel's butt in performance (Athlon 64 vs P4) AMD didn't gain much in market share because guys like Michael Dell said he'd never ship an AMD processor in one of his desktops, regardless of price and performance. Now that Intel is kicking AMD to the curb on high-end performance, all AMD has going for it is the low-cost market.
Re: (Score:2)
Re: (Score:2)
You also forgot why people like me stopped using AMD (after using them exclusively from 1995 till 2005). They kept changing the sockets which meant I couldn't upgrade my system one piece at a time any more.
Re:Being an innovator not always smart? (Score:4, Informative)
Intel didn't change sockets? How many sockets have they launched in the past six years? AMD has launched 3 main sockets in that time (754, 939 and AM2). Anyone remember Intel ditching Socket 423 after less than a year?
And AMD would release one proc on different sockets so you could still upgrade with your old mobo. For instance, when they came out with Socket 939, they were still releasing new procs under Socket 754. Even though they have Socket AM2/AM2+, you can still get Socket 939 procs.
AM2 came out in early 2006, and when I build my next rig in the spring, I'll still likely be building an AM2 rig. That being said, I'll probably go with a new motherboard for a faster bus, and faster memory support.
I could keep my existing mobo which will support quad-core AM2+ processors with a BIOS update, but to get the full potential, I need a new motherboard for the bus speed and memory improvements.
Intel is in the same boat. Chipsets and cores change often enough that you need to replace everything to get the best possible results.
Your logic was that you didn't want to change sockets and replace your entire system (AMD provided you that option to stay on the same socket) so you replaced your whole system and changed sockets to go to Intel.
How does that make sense?
Re: (Score:2)
Why yes I do have socket 754 and 939 cpus sitting in my cupboard with one working machine still running using AM2. I buy all my computer stuff at Fry's and every time AMD introduced new sockets, Fry's pretty much dropped the older CPUs and motherboards. If you went to the wall of motherboards at Fry's all the Intel mobos used the same socket while the AMD ones had a variety. As another data point consider that the Pentium 4 and Celeron D use the same socket whereas the Athlon 64 and Sempron used differen
Re: (Score:3, Informative)
Actually the Gentoo docs tell you to compile for the more generic architectures, and not the real specific CPU for reasons like that. Then again, I always compiled for the specific CPU.
I miss me some Frys. I loved that store. However I live in Nebraska these days, so I use NewEgg. You can still order a 386 motherboard on NewEgg (I kid you not).
Re: (Score:2)
Yes, but because AMD integrates memory controllers and such directly in their processors, AMD motherboards and cheaper than comparable Intel motherboards. This really is a win-win.
Re:Being an innovator not always smart? (Score:5, Insightful)
You didn't read my post. I never said AMD was faster now. I said that AMD *WAS FASTER* at one point, and these days all AMD has is the low price point.
For instance, the last time I built a computer for me (a little over a year ago) AMD offered a dual core processor for $35. The Intel equivalent that it was compared to in benchmarks cost $150. In the price-performance comparison, AMD came out way ahead at the low price point. At the very high end, AMD didn't have anything that could produce Intel's performance.
Not to mention that scientific computing is vastly different from general processing.
For a scientist, you sure don't seem to understand what I wrote. Go back and reread it.
Re: (Score:2)
Intel has been hit with antitrust charges in several countries. I assumed a Slashdot reader would be familiar.
http://arstechnica.com/news.ars/post/20070726-eu-slaps-intel-with-formal-antitrust-charges.html [arstechnica.com]
http://www.amd.com/us-en/Weblets/0,,7832_12670_13242,00.html [amd.com]
http://www.pcworld.com/article/142443/intel_and_antitrust_a_brief_history.html [pcworld.com]
http://en.wikipedia.org/wiki/AMD_v._Intel [wikipedia.org]
"This is because the Korea Fair Trade Commission has issued a fine of US$25.4 million against Intel."
Several vendors have come
Re:Being an innovator not always smart? (Score:5, Informative)
DEC invented that hypertransport for the DEC alpha. AMD liked the idea and adopted it. it was not AMD's idea.
Re: (Score:3, Informative)
You seem to be thinking about the Alpha EV6 front-side bus architecture that AMD used on the original Athlon. It's very different from the HyperTransport bus, and predates it by several years.
Re: (Score:3, Interesting)
yes, I was thinking of that. but how radical is the new amd vs that older ev6 stuff?
the whole idea is that its NOT a front side bus and its pt-pt from every node to every node.
intel still has this FSB notion and amd dropped that years ago (?)
Re: (Score:2)
A bus based around each processor being connected to a central hub, and a bus based around each processor being connected to one or more other processors allowing a variety of system topologies, is a pretty radical change imo. Electrically ev6 is more similar to HT in that they're point-to-point busses instead of a multi-drop FSB, but logically it's more similar to an FSB, with N processors contending for a single north bridge. That 'logic' aspect affects a lot of the design of the protocol since there's
Re:Being an innovator not always smart? (Score:4, Insightful)
Amen. I'm tired of explaining to my colleagues why AMD Opteron servers outperform Intel for use in database servers because of memory bandwidth and ccNUMA architecture. It's nice that Intel has finally realized that they can't keep designing processors for desktop PCs and not care about I/O bandwidth. This does mean I can finally be confident that when I buy a new 8-CPU, 8-core (64 total core) database server from Intel I don't have to worry about my poor MCH (memory controller hub) choking access to that nice 512GB of RAM I have hanging off of it.
Those of us building database servers, VMware clusters, and other high memory bandwidth applications can rejoice because the Nehalem architecture is finally almost here.
Re: (Score:2)
HyperTransport doesn't replace FSB. Only I/O and cache synchronization transfers go over HyperTransport, data accesses do not (as AMD integrated the memory controller).
Data accesses do go over QPI though. And I don't think that QPI really moves the RAM closer to the cores. QPI's main value is that by having fewer pins, it gives more flexibility in locating the bridges and such on the motherboard.
When are the 2 way ones that will be in the mac pr (Score:2)
When are the 2 way ones that will be in the next mac pro coming out?
For the desktop where are the nvidia boards and the lower end MB we need more the just the high cost X58 boards.
Also apple should have a 1 cpu core i7 system as well.
Expen$ive (Score:2, Interesting)
Re: (Score:2, Insightful)
Sure, but I don't buy a new pc whenever I get a haircut.
I got my first PC, an 386, around 1992. Next thing was a Pentium 1. Then it was up to a P4, which died on me some two months ago. Still haven't bought a new one,but when I do, I expect it to last me another five years at least.
2k$ over 5 years makes for 400$ per year. That's a lot less of an investment than what a lot of people spend on their PC.
That being said, I have no burning desire to play the every new game at the top of it's pixel range, either.
Re: (Score:2)
Re: (Score:2)
More than that, likely; $1,600 - $2,000 sounds right for the processor (~$1,000 itself) and RAM for that setup. But you max out a system using the best processor available, and its expensive. The first PC my family got had an MSRP of approximately $4,500 (and the employee purchase price that we actually paid was ~$2,500) -- in 1984.
By comparison, $2,000 in
Another great /. post. (Score:5, Interesting)
Link to the middle of an ad-laden article and to the Cinebench of all pages - because, you know, that is what the average /. reader is running...
Also, add a nice touch: forget to mention that while the i7 is faster clock for clock with the Core 2, it currently tops out at 3.2GHz and has some sort of overclock protection (lowers clock when it goes over 110A or 130w).
My cheap Core 2 is running at 4GHz on just the stock fan, I don't see myself upgrading to the i7 anytime soon.
What did you say? ... What do you mean Cinebench would still run faster?
Re: (Score:2)
The linked article showed a Core i7 running at 4.15ghz with stock air cooling.
Re:Another great /. post. (Score:4, Funny)
You made me RTFA. The same ad-laden FA I was complaining about. Thanks. So, from the article.
Because the Core i7 Extreme 965 has its overspeed protection removed--i.e. its multipliers are unlocked--we overclocked the processor by raising its multiplier to 25 and also experimented with an increased QPI speed.
My 4GHz Core 2 is not a $1000 *Extreme* part. Humanly priced i7s will have overspeed protection.
I have the feeling you knew this was the case anyway, but had me read TFA just for kicks... shame on you!
Re: (Score:3, Insightful)
It would be nice for them to put one or two 'old' processor scores for reference, I am using a 5YO celeron and don't have the slightest idea what these scores mean in to relation to what I am using.
Re: (Score:2)
You seem to know what 110A means (huge current indeed), yet you haven't put 2 and 2 together and realize how you get a 130W TDP (pretty average for a modern CPU) out of 1.16V ;)
Anandtech Review (Score:5, Informative)
Lot of reviews out, but there is one with 64 bit (Score:5, Informative)
http://www.planetx64.com/index.php?option=com_content&task=view&id=1435&Itemid=14 [planetx64.com]
1) 64 bit macro-op fusion is new. See it tested here..
2) Virtualisation is more efficient with nested pagetables.
3) Gaming should benefit, since all x58 mobos support Crossfire
and nVidia SLI.
4) 12 gigs ram supported with 2gb dims - this is rare for desktop boards.
Numerous other minor tweaks, but read it for yourself..
Have fun with your upgrade dollars!
Real world performance??? (Score:4, Interesting)
And I was *just* about to retire my "old" socket 940 dual-core opteron box for a quad core Intel system. I think I'll just wait another month or two and jump to the i7 platform instead. 8-)
Would be nice to see some video and audio encoding benchmarks and some real world application performance numbers instead of teenmarks (gaming performance).
Cheers,
Servers? (Score:3, Interesting)
Re: (Score:2)
check out the gainestown.
Some info here: http://en.wikipedia.org/wiki/Intel_Nehalem_(microarchitecture) [wikipedia.org]
More reviews (Score:4, Informative)
Another review with some more data, including memory channel performance testing, good explanations of overclocking process, etc.
http://www.pcper.com/article.php?aid=634 [pcper.com]
Love the heat sink!!!! (Score:2, Interesting)
..and Intel and AMD, please blast through 3.2Ghz per CPU so all programs work faster all the time.
What good is it? (Score:2, Insightful)
For CPUs and memory, the market needs to focus on power usage
Re: (Score:2)
Absolutely (Score:2)
I second that. Transcoding is a big deal, and not just for folks ripping their commercial videos to servers. It takes several days to complete a full transcode my FLAC audio files to the preferred compressed flavor of my audio player, which changes every year or three. I do it very infrequently because of the time it takes. I would prefer to sync with the device on the fly, but processing speeds are just too slow to do that. I'd rather leave the computer running for a couple days once and then sync from th
Re: (Score:3, Insightful)
Assuming a simple scaling, you're talking about roughly 50% more performance.
Which, in the mid-late 2000s era is huge.
A lot of games that folks play are CPU-constrained. So that's 50% more framerate, or the difference between something that feels pokey vs something that works well.
That's 50% faster encoding / transcoding for videos.
Yeah, it's
And security? (Score:2, Interesting)
I wonder if they have fixed the security problems of the past.
http://hardware.slashdot.org/article.pl?sid=07/06/28/1124256 [slashdot.org]
940 pins? (Score:2)
Fuck it, we're doing 1366!
A quick-take video review on the product, here... (Score:2)
Re: (Score:2, Funny)
It's GT/s, not GTB/s, there's no bytes. Just 6.4 * 10^21 per second. Or maybe 6.4 * 10 ^ 9 Teslas per second. Or 6.4 cars per second.
new unit (Score:5, Informative)
actually, it's gigatransfers per second... thanx for dkf ( http://slashdot.org/comments.pl?sid=1016475&cid=25611995 [slashdot.org] ) for informing that link.
Re: (Score:2)
T=Transfers maybe?