|
Early Havendale benchmark figures |
|
|
|
Written by Lars-Göran Nilsson
|
|
Wednesday, 29 April 2009 09:09 |
|
Could be why it won't launch
One of our sources has provided us with some early Havendale benchmark figures and we're starting to get a clearer picture of why Intel is skipping Havendale and moving straight to Clarkdale, which will hopefully perform better.
As we told you on Friday last week, Havendale and Clarkdale feature an upgraded G45-chipset type memory controller, unlike Lynnfield which has a cut-down Nehalem memory controller. However, it seems like Intel bodged something on Havendale, as the memory controller is having some serious performance issues. The Everest memory read tests on the G45 chipset is about 8,000 while the Havendale sample our source has been testing manages an underwhelming 2,600.
If this wasn't bad enough, the CPU itself doesn't seem to perform as well, as our source also has access to a Lynnfield CPU which had two cores disabled and its clock speed reduced to match that of the Havendale CPU for some comparative testing. The Lynnfield CPU managed to run SuperPI 32M in about 16 minutes, while the Havendale CPU took 5 minutes longer at about 21 minutes.
This isn't good news for Intel, but we're confident that the final Clarkdale CPU's will have been fixed, as Intel couldn't possibly release a new CPU that performs this poorly. At least it goes to show that even the mighty Intel can run into some major problems when it comes to designing a new CPU. It does also, at least in part, explain why the Core i5 CPUs with integrated graphics got pushed back compared to those without.
Also read
Clarkdale to use upgraded G45 memory controller
|
