Intel Core 2 Extreme QX9770 in Real-Life Apps

[anticupidon]

Digit-Life - Intel Core 2 Extreme QX9770 in Real-Life Apps - Page 1: Introduction, tests
I stumbled upon it ..seems a fair nice article.

[MagicLess]

Interesting read, thanks for finding it.

[tool_462]

Odd timing to put out an article on the QX9770.

[JumpingJack]

Odd timing to put out an article on the QX9770.

Not really, he was looking at how memory impacts performance between various configurations.

[1Tanker]

I was surprised at the Encoding and especially the Web Server results. I figured they would have enjoyed the greater memory bandwidth. Nice article.

[beerandcandy]

its a monster
i wonder if nehalem is better?
lol what a loaded question

[anticupidon]

Not really, he was looking at how memory impacts performance between various configurations.

True.
After reading an article about the DDRAM3 performance impact versus DDRAM2 and reading this article i was more enlighted and more satisfied about not buying the creme de la creme of today's hardware.
Still it was an article i was craving after...

[JKflipflop]

its a monster
i wonder if nehalem is better?
lol what a loaded question

I heard nehalem is 30% slower in single threaded apps and 45% slower in multithreaded apps than a smithfield.

[beerandcandy]

awsome! giving amd a chance to live...

[kg4icg]

How quickly some people forget the results someone got last week about Nehalem which were posted on this site.
Intel Core i7 940 in Real TestbyUneIT

[tool_462]

Not really, he was looking at how memory impacts performance between various configurations.

Shows you how far I read into it

[monst0r]

I would have thought that memory would play a bigger role in rendering and the likes. Good article

[JumpingJack]

I would have thought that memory would play a bigger role in rendering and the likes. Good article

This is the advantage (applicability) of a large cache with aggressive prefetchers. Intel's current platform is not very sensitive to memory conditions, since the FSB is generally an albatross Intel has architected in features that work around it, and does a good job in most all cases.

Jack

[tool_462]

This is the advantage (applicability) of a large cache with aggressive prefetchers. Intel's current platform is not very sensitive to memory conditions, since the FSB is generally an albatross Intel has architected in features that work around it, and does a good job in most all cases.

Jack

But Nehalem should see a good performance gain when using faster/tighter memory? (Such as AMD chips saw with X2's etc)

[JumpingJack]

But Nehalem should see a good performance gain when using faster/tighter memory? (Such as AMD chips saw with X2's etc)

This is a case where it is hard to say architecturally ... not until we see some numbers and have controlled samples (on my bench anyway) to draw conclusions.

The cache hierarchy is much different in Nehalem over Penryn. The largest cache pool is one level down, and while the access to that memory pool will be faster than that to main memory, it is slower than the enormous L2 memory cache pool found in Penryn.

How to differentiate and comprehend what is IMC driven, cache driven, and core driven will be difficult.

Most all of Nehalems architectural tweaks are focused on multithreading, even the front end buffers were built with SMT in mind, as such you cannot really compare memory access when looking at multithreaded benchmarks because there is so much more going on in the front end....

Similarly, in single threaded benchmarks/performance, Intel has de-allocated the hard sharing rules for the re-order buffer, the dispatch station, and reservation stations -- as such, a single threaded app will have a much wider execution window compared to Penryn, which will be a sizable boost unrelated to memory access....

All in all, my take is such --

On the DT
- Nehalem will have varying degrees of improvement on single threaded performance, some case 0 to none, others as high as 15 or 20% IPC wise, it will vary.
- Multithreaded desktop apps will shows major gains, such as Cinebench/rendering and encoding/transcoding. Some games will benefit but the gaming performance will not be overwhelmingly stellar .. but will be better than the 'ho-hum' down play statements you see from AMD fanboys.

On server --
- Nothing short of spectacular. AMD does not stand a chance, Intel will blow by them in every workload bar none.