It's feeling a little

[Jonathan]

Nehalem-y in here today!

http://www.extremetech.com/article2/0,2 ... 764,00.asp
http://anandtech.com/cpuchipsets/intel/ ... spx?i=3448
http://enthusiast.hardocp.com/article.h ... VzaWFzdA==

Launch is in a few weeks.

[Jonathan]

http://www.guru3d.com/article/core-i7-m ... ce-review/

[Jonathan]

In short, modest or no gains on older games, larger gains on new games, huge gains in multi-GPU configurations, and nice performance across the board in anything threaded. General single-threaded integer is perhaps 10-15% faster, assuming it wasn't memory bandwidth constrained. Significantly more memory bandwidth. The cheapest part is $285 and all three are 130W. The motherboards are expensive. I wouldn't expect budget motherboards for Socket 1366.

[Jonathan]

I am miffed all my favorite sites favor Everest's memory bandwidth test, which isn't threaded.

[Jonathan]

And I am sort of flabbergasted that sites report "memory latency" without measuring the cache hit rate. Hint: it's hard to fool our prefetchers.

[Jonathan]

http://www.theinquirer.net/gb/inquirer/ ... s-counting

NHM review roundup.

[VLSmooth]

Which review(s) (if any) do you prefer?

I ended up reading Anand's last night when it popped into my RSS reader.

Also, woo! (as you and others mentioned):

Thermalright 1366 Bolt Thru Kit
http://www.thermalright.com/new_a_page/news.html

Now to see how fast prices drop (specifically, cpu, motherboard, ddr3 ram). If the drops are slow, I might as well build a new system asap. I'll want to research 6 memory slot motherboards too (hoping the ASUS is flat-out the best to save time).

[VLSmooth]

I recently skimmed this
AnandTech: Nehalem: The Unwritten Chapters
and wondered... since Nehalem used the "2% performance boost per 1% of increased power consumption", can I significantly (increase power efficiency)/(decrease temperature) by under-clocking a Nehalem chip? Or is it not worth it?

[Jonathan]

Yes, but I believe that's true for any chip?

[Jonathan]

Just to be clear, power drops as the cube of voltage and perf drops linearly with frequency. And voltage and frequency aren't quite linear, but they're close. So underclocking pretty much anything will net you
1. an increase in power efficiency
2. a decrease in temperature
3. a decrease in absolute perf

[Jonathan]

What you really want to know is how significantly can you underclock your Nehalem while maintaining levels of perf that you demand on your workloads (i.e. HD anime). The answer is, I dunno.

[skanks]

All of these reviews are for the desktop/server variants of Nehalem. Do you have any recommended reviews for the mobile Nehalem architecture?

[Jonathan]

Mobile NHM products have not been launched.

Edit: you want either a Auburndale (dual-core) or Clarksfield (quad-core). Both are not scheduled to be released for a while.

[VLSmooth]

Sorry, I had to run to a meeting. I was supposed to mention:

The Q9450 can operate at voltages down to 0.85V and as high as 1.3625V, while the Core i7-920 currently appears to be limited to a minimum of around 1.137V. Power consumption of a CPU at a fixed clock speed is proportional to the cube of the voltage, so despite whatever power efficiencies Intel has included in Nehalem they will not outweigh a Penryn running at a lower core voltage.

which you covered (at least the relation part).

Hence, there's a voltage minimum which is separate from the clock speed. I thought about lowering both voltage and clock speed, while still maintaining acceptable performance, but I'm not sure how much that would help, hence your:

What you really want to know is how significantly can you underclock your Nehalem while maintaining levels of perf that you demand on your workloads (i.e. HD anime). The answer is, I dunno.

is right on target. I think this also means I should disable Turbo (once I figure out how) since my temperature and power requirements are probably tighter than the standard.

[skanks]

Not to be a huge ignoramus, but wasn't the Nehalem supposed to be all about the Mobile computing? I thought the point was slightly better performance with significantly fewer transistors to get seriously boosted performance per watt.

[Jonathan]

http://anandtech.com/cpuchipsets/intel/ ... i=3461&p=2

Mo' info about mainstream parts.

[VLSmooth]

It'd feel a lot more nehalem-y nearer me if there were more 1366 motherboard reviews! (Most likely will order parts while visiting family)

[Jonathan]

Pardon me, sir, but your server scores are showing.

The SAP numbers are absolutely astonishing, as Intel's dual socket is able to outperform quad socket opteron machines.

Martin, I'm looking at you.

[Jonathan]

http://www.anandtech.com/cpuchipsets/in ... i=3513&p=4

A nice terse summary.

Core i5 will be launching in 2H 2009 and is probably the part anyone on this board would want. It will be available in desknote form factors for those who like that sort of thing. Two socket server will launch 1H 2009 (i.e. soon), for those (Martin) who want 16T systems. The truly mobile Arrandale, built on 32nm, will be on sale in Q4 2009 (note this is different than Core i5 launch).

[Jonathan]

Pardon me, sir, but your server scores are showing.

The SAP numbers are absolutely astonishing, as Intel's dual socket is able to outperform quad socket opteron machines.

Martin, I'm looking at you.

http://www.anandtech.com/weblog/showpost.aspx?i=554

Well, the SAP numbers are showing a dual 2.93 GHz (or 3.2 GHz) Xeon beating the only quad AMD 8384 (Shanghai at 2.7 GHz) score of 22000 we have so far. Granted, a blade server is most of the time a bit slower. But four AMD 8384 2.7 GHz will be in the same league as a dual Xeon X5570, which will be out very soon now.

Confirmed!