backtop


Print 33 comment(s) - last by phorensic.. on Aug 25 at 5:18 PM


Nehalem's new Turbo Mode is similar to the 386's turbo button, in that it provides more power to the CPU. On a whole, it saves power and delivers superior performance in single and dual core applications.  (Source: Techgage)

Intel's new on-die memory controller features three channels, and appears to be working superbly as it more than triples Penryn's bandwidth. The final picture should become clear with latency information arrives.  (Source: Techgage)
There's a lot of new stuff going under the hood of Intel's upcoming "Nehalem" processor

Intel's plans for the next four years were leaked a bit prematurely before its Intel Developers Forum (IDF) this month.  Now with the IDF in full swing, Intel is releasing many more interesting details on the upcoming processor that is on top of its list -- Nehalem.

Formally known as the Intel Core i7, Nehalem features eight logical cores, the return of HyperThreading, on die memory support, QuickPath, and more.  It will launch later in 2008.

One interesting revelation which Intel left off with was that it would probably release its notebook processors at almost the same time as its desktop variants, unlike with its Penryn processors.  It expects many of the power-saving technologies found in Nehalem to be very useful in preserving laptop battery life.

The first intriguing technical detail of Nehalem according to Techgage is the new "Turbo Mode".  Similar to the Turbo button of 386 days of yore, the turbo functionality will up the power to the CPU.  Overall performance will be reduced in Nehalem to save power.  For example, three cores of a four-core Nehalem processor could be turned off while a system is idle. More power can be transferred to that single core as well when performing simple operations that don’t require four cores blazing away.

However, if more CPU power is needed, power can be distributed to the remaining cores.  Turbo Mode can be turned off in the BIOS and does not replace traditional overclocking.

Also discussed was Intel's new on-chip memory controller.  It uses three channels in an unusual approach.  The approach is apparently paying off as memory bandwidth with its QuickPath Interface (QPI) is up to 25.6 GB/s, almost 3.5x the performance of current gen memory.  Unfortunately, Intel refused to divulge latency information, so the memory picture is still incomplete, though it’s looking bright on the bandwidth side of things.

As Nehalem samples have been show capable of working with even the highest end DDR3, DDR3-2000, many are now beginning to wonder whether such high speeds will even be necessary given the massive bandwidth.  Some are speculating that Nehalem's impressive memory performance will spell death for the high end memory market, already hurt by sinking prices.

It’s hard to say whether this will happen as increasingly there will be scenarios that call for massive memory use.  Intel is making big claims about its bandwidth, however.  According to Intel one stick of DDR3 on and Intel X58 motherboard with a Nehalem nearly equals the bandwidth of a dual-channel DDR3 configuration with a Penryn processor.  Nonetheless, memory manufacturers are planning to release 3 GB DDR3 kits to coincide with Nehalem for 32-bit OS's and 6 GB and 12 GB kits for 64-bit OS's.

While IDF attendees were sworn to secrecy on performance specifics, they did say that the overall performance shows good gains, similar to the Penryn.  They say that HyperThreading features few surprises and aside from some minor improvements is virtually identical to that used in the Pentium 4.

In conclusion, there are a few new juicy details on Nehalem, but in the end even more questions.  Fortunately answers will be coming in a few short months.  With AMD's Shanghai set to compete with Nehalem, perhaps the most interesting question of all is when will Nehalem's competitor materialize.



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

Turbo
By Adonlude on 8/20/08, Rating: 0
RE: Turbo
By MrBlastman on 8/20/08, Rating: 0
RE: Turbo
By IsDanReally on 8/20/2008 2:26:18 PM , Rating: 2
Unfortunately, the article does not claim a turbo button is coming, just a turbo mode, controlled automatically by the BIOS (and I imagine in some cases software from the motherboard manufacturer). Still, interesting we come back to such old ideas.


RE: Turbo
By barjebus on 8/20/2008 2:30:10 PM , Rating: 5
I should hope so. If there was seriously a BUTTON for increasing the power you may as well create some kind of boat throttle instead and lean back in your chair as you proceed to warp speed, pulling the throttle back, and encoding video's by the dozen.


RE: Turbo
By Ctsephion on 8/20/2008 2:36:18 PM , Rating: 2
If you arn't aware, back in the 3/486 era, there were cases that made it possible to jump up from 33Mhz to 66Mhz. But since there were programs incompatible with such a high processor speed, the ability to "switch back" was needed. Thus, the 'turbo' button was etched into the pages of history...


RE: Turbo
By MrBlastman on 8/20/2008 2:55:42 PM , Rating: 2
My 286 had a Turbo button...

It took it from a blazing 4 Mhz to 8 Mhz! We were really flying fast!

What is funny is, for some "really old games" at the time (remember, this was 1987), say games made in 1983/84, I needed to turn the turbo mode off in order for these older applications to be controllable.

Imagine... 8 Mhz being too fast...


RE: Turbo
By Fnoob on 8/20/2008 3:05:22 PM , Rating: 2
Had that one too - ah, those were the golden days.

The interesting thing to note in both the 286 and 386 cases - the Turbo button actually DOUBLED the speed. For some reason I doubt this new tech is going to take us from ~3Ghz to 6Ghz magically.


RE: Turbo
By AFMatt on 8/20/2008 4:22:48 PM , Rating: 4
Actually, the Turbo button's purpose was to reduce the speed. With the button on, you were running the actual specs for the processor.
When you took it off "Turbo" mode, the processor clock was downclocked to a certain level to allow for compatibility with software written with older processors in mind. It was generally a 40% or so difference.


RE: Turbo
By masher2 (blog) on 8/20/2008 6:10:29 PM , Rating: 2
> "My 286 had a Turbo button...It took it from a blazing 4 Mhz to 8 Mhz! We were really flying fast!"

I think you're misremembering slightly. My 286 went from 6-8-10 mhz depending on the turbo setting, but I don't believe any 286 scaled below 6mhz. Even the old 8086 itself debuted at 4.77mhz.


RE: Turbo
By littlebitstrouds on 8/20/08, Rating: -1
RE: Turbo
By littlebitstrouds on 8/22/2008 11:46:09 AM , Rating: 1
I get rated down by a masher fan boy even though he's proven wrong 1 line later... brilliant.


RE: Turbo
By 306maxi on 8/24/2008 6:37:55 AM , Rating: 1
You got rated down for acting like an @ss.......


RE: Turbo
By MrBlastman on 8/21/2008 10:38:50 AM , Rating: 2
Tandy 1000 TX:

http://www.old-computers.com/MUSEUM/computer.asp?c...

Here is the manual:

ftp://ftp.oldskool.org/pub/tvdog/tandy1000/documen...

Page A-15

I quote:

"Slow Speed. Changes the CPU speed to 4 megahertz. The normal CPU speed is 8 megahertz."

Anything else? ;) I loved my Tandy 1000 TX. The 16 color TGA graphics gave me PC Jr equivalent colors with a much faster processor, I had 640k ram (later upgraded to 768k) and 3 VOICE SOUND! The 3 voice sound alone gave me a huge edge against everyone else I knew at the time as it was melodious and beautiful to the ears. Gunship sounded incredible.


RE: Turbo
By initialised on 8/20/2008 9:31:09 PM , Rating: 2
Turbo = Inverted speed step, that is all.
Powersave by default, full power when needed only on cores as well as clocks this time round. Suggests to me that the Nehalem will be a powerwhore like previous HT enabled CPUs.


RE: Turbo
By Ctsephion on 8/20/2008 2:28:35 PM , Rating: 2
I wonder if we will have manufacturers that treat this as the new way to capitalize on the "retro-cool" aspect that is the turbo button. I don't think it would be too far fetched to think that manufacturers like Thermaltake or Antec would do a specialized button on a Nehalem oriented computer case, considering they had a "turbo" button on MSI's 9600GT.


RE: Turbo
By Sulphademus on 8/20/2008 3:10:22 PM , Rating: 3
While I like the idea of a hardware switch to toggle performance modes, I hate the name. Turbo just sounds sooo... juvenile, like red flame decals make it go faster too. Perhaps something like "Performance Mode" (assuming low power is default) or "Power Saver" (if full throttle is default).

It would be very nice if this could downclock/upclock my graphics card as well.


RE: Turbo
By Schrag4 on 8/20/2008 4:05:50 PM , Rating: 2
The button on the front of the case will toggle between 'Performance Mode' and 'Tree-Hugging Mode'

Sorry, couldn't resist. Seriously, though, if the CPU has a Turbo Mode, the user should be able to turn it on, and leave it on, if he or she likes. I also think software should be smart enought to kick it on and off when needed, which would be fine most of the time. However, sometimes I get a little impatient with my machine and would like to know that it's not slowing down in the name of power efficiency.

I know this sounds ridiculous, but perhaps instead of a button on the case, there would be a timer that you would set. That way if you know you're going to bed in an hour, you can just set the timer for an hour, and know that it will be in turbo mode while you're still up, and not in turbo mode while you're asleep.

I got the idea from a house that my wife and I were considering buying. It had timers on the shower fans (up to 30 min). I can't tell you how much electricity that would save in our house if we had these, between running the fan (not much electricity) and running the air-conditioning to make up for all the cool air we just accidentally pumped outside for the last 2 hours!!!! I've been meaning to get these for our house but our local Home Depot doesn't have them, and I've simply forgotten all about it since I looked.


RE: Turbo
By MrBlastman on 8/20/2008 4:40:05 PM , Rating: 2
The new Nvidia cards (200x series) downclock automatically when not running 3d applications to save power.


RE: Turbo
By MrBlastman on 8/21/2008 10:28:21 AM , Rating: 2
I love getting downrated for a factual comment.

My EVGA Precision tool shows it being downclocked clear as day. What more is there to dispute?


RE: Turbo
By AlexWade on 8/20/2008 3:15:31 PM , Rating: 4
I had a turbo button computer. Later on, I found out that it did nothing but change the LED lights to read 66 instead of 33. So I modified the jumpers so that the TURBO would read 200.

The funny thing is I still play those old DOS games.


RE: Turbo
By Ringold on 8/20/2008 9:31:26 PM , Rating: 2
I just whipped out VGA Trek yesterday. Games might be 5000x the size, but sadly, not really all that much better in some cases.

I still think Stars! may well be my favorite game of all time, and think I had more fun in some old text-based adventure games than, say, DooM 3. In fact, when "my boat comes in," I swear I'll personally finance a modern Stars! clone, minimal emphasis on graphics, just mild improvements that technology allows over the original. Once it's done, I'd try to sell it for a year or two, then I'd make it open-source so it avoids the same fate of Stars!; incompatibility with future OS's.


RE: Turbo
By phorensic on 8/25/2008 5:18:58 PM , Rating: 2
The button on my 386 case did the same thing...and sure enough I did exactly the same thing as you. I think I made it say 999, though, hehe.


RE: Turbo
By walk2k on 8/20/2008 4:40:02 PM , Rating: 2
When will we start seeing these chips and the motherboards for them? Oct?


So how does this work?
By voodooboy on 8/20/2008 3:14:01 PM , Rating: 2
quote:
More power can be transferred to that single core as well when performing simple operations that don’t require four cores blazing away.


Sounds good, although it is something they should have implemented much earlier even on their dual cores. But I have a question; Is there a predesignated "turbo" core or will the "turbo" duty be cycled across the 4 cores using some kind of algorithm?




RE: So how does this work?
By Diesel Donkey on 8/20/2008 5:12:31 PM , Rating: 2
It would be nice if you could have anywhere from 1 to 4 cores active at any given time, based on the workload. I hope it's not just two options: 1 core or all 4 cores.


RE: So how does this work?
By benx009 on 8/20/2008 6:05:08 PM , Rating: 2
Well, when are you ever going to need anything else. You either want your PC running at its fastest for gaming performance, or at its slowest speed for lighter tasks, as long as it can handle them comfortably. This is a Nehalem CPU. Slow for this thing is probably encoding a 30 minute video in like 2 minutes (probably not accurate, but you get my drift).


RE: So how does this work?
By Ringold on 8/20/2008 9:33:03 PM , Rating: 2
Well, there are some multi-threaded apps that can load 2 cores but not 4, or at least don't load the 3rd and 4th to any big degree.


RE: So how does this work?
By emboss on 8/21/2008 6:00:43 AM , Rating: 2
I assume it's using a mechanism similar to the "Foxton" technology that was supposed to be in Montecito (the dual-core Itanium 2). Basically, the maximum frequency allowed for a core is determined by how much power the chip as a whole is using.

Taking the simple relation of power usage being proportional to frequency times the square of the voltage, then if you disable 3 of the four cores you can run the remaining core at 4 times the power level. This would be about twice the frequency assuming a 40% voltage increase. Of course, this assumes that heat is the limiting factor (which it is for the Itanium, no idea for Nehalem).

Like the old "turbo" button, this is really a "throttle down" button. The electronics would throttle the chip back if it detects power usage is outside the TDP.

The other positive thing for Intel is that it'll make overclocking much harder. If you cannot exceed make the chip exceed its TDP, then it really puts a big limit on how far you can push it. Of course, the kilobuck extreme editions will have this disabled or adjustable or something.


Focus on Nehalems.....
By crystal clear on 8/21/2008 11:33:27 AM , Rating: 2
Now that we have Intel plans very crystal clear as mentioned below-

Of note was Gelsinger's revelation of the first eight-core Nehalem-EX silicon. The EX stands for expandable server market apparently and indicates that the chips will slot in large systems. Intel will sell these chips with 2, 4 or 8 cores and do a couple software threads per core.

The desktop - Core i7 - and standard server Nehalem chips will start arriving in the fourth quarter. Then the the EX server chips and additional desktop desktop (“Havendale” and “Lynnfield”) and mobile (“Auburndale” and “Clarksfield”) chips will appear in the second half of 2009.



http://www.theregister.co.uk/2008/08/20/intel_neha...

Success of the i7 in the market place depends on Intel's ability to bring out a rock solid product without any major issues,like the types that plagued barcelona.

Success depends on Intel's ability to bring the product into the market in time,to enable Tier 1 OEMs to plan & execute their product offerings & get them into the market in time.

Pricing & Performance being the major determining factor in the end.

Get me to the market in time is the message.

AMD is in a very critical stage after the Barcelona disaster,eventually being punished severely in the form of market share,profit share,revenues etc.

They cannot afford a repeat with the Shanghai if it does not deliver on time,then AMD can say goodbye to whatever hopes & dreams it has to break even or return to profitability or increase it marketshare or revenues etc.

It will play havoc with its share prices & its very existance.

That reminds me Intel's acquistions in the past of VMware & Havoc has paid off by intergrating these technologies into Intel products namely Nehalems & Larrybee for a starter.

"You can get about 2x the performance out of VMware thanks to VT-d, which will be in the Nehalem chips."

Sorry to say that D.T. coverage of Intel's IDF is very poor & lacks the contents so desperately needed.

Even something basic like providing links to Intel's IDF press releases ,which gives the relevant contents accurately.

This is constructive criticism & should be taken ina postive manner.




RE: Focus on Nehalems.....
By crystal clear on 8/21/2008 12:39:00 PM , Rating: 2
Intel Senior Vice President Patrick Gelsinger told eWEEK at the Intel Developer Forum here.

One reason for this gradual rollout of Nehalem-based processors, Gelsinger said, is to allow OEMs to build new systems and allow for extensive testing and a long validation process for IT shops interested in the new processors.


“With Nehalem, it is more work for our customers and our OEM partners as they build up systems,” said Gelsinger. “We are on track for what we said, and when you have major new system designs, there is a longer validation cycle and a longer ramp into the marketplace.”

http://www.eweek.com/c/a/IT-Infrastructure/Intels-...


Worries ?
By William Gaatjes on 8/21/2008 12:09:15 PM , Rating: 2
quote:
Unfortunately, Intel refused to divulge latency information, so the memory picture is still incomplete, though it’s looking bright on the bandwidth side of things.


This worries me, Intel always gives good information about a processor from the very first bootable chip.
But no information is given about the latency. I am wondering why.




RE: Worries ?
By crystal clear on 8/21/2008 12:26:35 PM , Rating: 2


“We have bettered the core and extended our cache architecture, and we also now have a dramatic system architecture upgrade with the integrated memory controller, three channels of DDR3 [double data rate 3 memory], [and] three times the memory bandwidth with half the memory latency,” said Gelsinger.

http://www.eweek.com/c/a/IT-Infrastructure/Intels-...


By Tywald on 8/21/2008 6:35:39 AM , Rating: 1
Most of us have probably read the short article from Anandtech about Nehalem that it was first and foremost designed to increase multi-threaded and bandwidth performance. It also reads that we shouldn't expect a huge gain in performance in current games and single-threaded applications.

To me this is an atempt to make the "Core i7" perform stronger and stand out as a much better choice over the "Core 2 Quad" to the average joe that doesn't care why it's soo, knowing that it does that by overclocking itself isn't important as long as it performs better than an equally clockrated processor.

Also i think it's an attempt form Intel to hinder AMD from gaining supporters from people that buys cheaper processors and overclocks them. If AMD manages to pull an Ace from their sleves, their "Deneb" could be just as fast the "Penryn" clock-for-clock. This means that whit the aggresive pricing strategy they're going to use as well as market it as overclocking friendly, it could pose a threat against the middle and lower end "Nehalems" that aren't going to be manually overclockable according to Intel. (unless they have changed that policy).

- Tywald




"DailyTech is the best kept secret on the Internet." -- Larry Barber














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki