Tri DDR3 on i7 - scam ??!!

D

drogos

Registered

​

Joined

Dec 2, 2008

• Dec 11, 2008

• #1

According to a very detailed and comprehensive benchamark linked below there is no difference what so ever when you use 2 memory chips vs. 3. The article is in Polish but you should be able to read graphs easily. What are your thouths ?
http://www.egielda.com.pl/?str=art&id=4601-20

 

Alexhk

Member

​

Joined

Nov 30, 2004

• Dec 11, 2008

• #2

In general use its not like the Core i7(at stock) is bandwidth starved with only dual channel. You incur a slight latency penalty going to tri-channel so if dual channel was already providing adequate bandwidth you'll see a small performance decrease going to tri channel. Going to depend on the application whether the additional bandwidth pays off or not.

Now once you start overclocking you may hit a point where tri-channel wins every time or you may not, i don't know, but I do know tri channel will be a must when octo core chips hit the market.

 

ozzlo

Senior whose title will pier

​

Joined

Mar 17, 2002

• Dec 11, 2008

• #3

It's not a scam but an excuse for me to get 6 gigs of ram vs 4 gigs .

the theoretical amount of bandwidth for Tri channel DDR 3 is insane. It is in theory 50% faster. The benefit I see currently because DDR3 prices are so high, Tri channel allows you to purchase 6 gigs of cheaper slower ram that has similar bandwidth to 4 gigs of faster more expensive ram.

Example.
http://www.newegg.com/Product/Product.aspx?Item=N82E16820231223
$165
3x2GB DDR 1333 Tri channel = 32GB/s

VS

http://www.newegg.com/Product/Product.aspx?Item=N82E16820227310
$250
2x2GB DDR3 2000 dual channel = 32GB/s

That's $80-100 difference in price for very similar bandwidth.

sure the dual channel option will be a minute amount snappier but with the tri channel option, you get 2 more gigs.

 

Last edited: Dec 11, 2008

N

no1stunna

Disabled

Joined

Dec 10, 2008

• Dec 11, 2008

• #4

dang you beat me to it ozzlo very well said

 

Albuquerque

Member

​

Joined

Mar 26, 2003

Location

North America

• Dec 11, 2008

• #5

For it to be such a sham, it sure looks to me like it helped performance in pretty much every case - even to include various latency tests.

Sensationalist topics FTL

 

jason4207

Senior Member

​

Joined

Nov 26, 2005

Location

Concord, NC

• Dec 11, 2008

• #6

The testing I've seen on this forum so far suggests dual-channel is faster than tri-channel. Dual-channel also puts less stress on the IMC.

I wonder if future i7 revisions will have better IMC's that can better utilize the tri-channel configuration.

If I had a current i7 rig I would definitely consider going w/ just a 2x2GB kit. I don't need more than 4GB at this point in time. If I did I could get another 2x2GB kit and be at 8GB which should hold me over for the next couple of years. Getting 12GB seems ludicrous to me at this time, but others may have a use for it.

 

Shiggity

Member

​

Joined

Dec 16, 2007

Location

Chicago, IL

• Dec 11, 2008

• #7

Once the Xeon Nehalems start coming out and you see tri-channel in the enterprise / server / virtualization environment, you'll see that it makes a difference.

 

Automiketic

Member

​

Joined

Jun 21, 2008

• Dec 11, 2008

• #8

i believe it just another case of the hardware out running the software. nothing properly utilizes as of yet so of course theres no increase. kinda like quad core vs dual core in gamming.

 

th3

Member

​

Joined

Aug 19, 2007

• Dec 11, 2008

• #9

Did a quick test a few days ago, tri is a win imho.
Sandra latency / int bandwidth:
Dual channel: 57ns / 22.06GB
Tri channel: 59ns / 29.58GB

The increased latency is very small (Bios F4j), the bandwidth increase is massive, 6GB is better than 4 (Linux 32bit PAE kernel), and no extra voltage needed to run tri, only advantages in my case.

 

jason4207

Senior Member

​

Joined

Nov 26, 2005

Location

Concord, NC

• Dec 12, 2008

• #10

th3 said:

Did a quick test a few days ago, tri is a win imho.
Sandra latency / int bandwidth:
Dual channel: 57ns / 22.06GB
Tri channel: 59ns / 29.58GB

The increased latency is very small (Bios F4j), the bandwidth increase is massive, 6GB is better than 4 (Linux 32bit PAE kernel), and no extra voltage needed to run tri, only advantages in my case.

Click to expand...

Well that's better than I've seen. Most of the stuff I saw Brollocks and others do had the bandwidth near equal b/n dual and tri, w/ a big latency hit for tri.

 

kayson

Member

​

Joined

Jan 5, 2005

• Dec 12, 2008

• #11

Just because there's not an enormous performance increase doesn't mean the technology isn't working. And it definitely doesn't mean it's a scam.