Guest
SiGfever
2008-03-02T12:10:57Z

 Nehalem will be very expensive at first so anice Maximus Formula and a 45nm chip could stave off the addict's hunger for a while at least. [:D]

Kamrooz
2008-03-02T12:48:01Z

I do agree that if you need a upgrade, best just go for it...Heck. If I could get a new rig now, I would in a heartbeat. My rig is way too dated to even considering waiting for Nehalem.

But if you do have the cash, best not to go too crazy. Nehalem will indeed be a monster....On some of the forums I visit, there are a few users who work at intel, and they assure me that my jaw will drop. So We'll have to wait and see how this sucker performs ^_^. In terms of pricing, always extreme edition first at around 1000 or so. Than Q1 2009, we'll have our 150-600 dollar range. 😛....I'm just praying Nehalem is a overclocking god just like Core 😛. 

Lithotech
2008-03-03T14:22:00Z

Couple of quick questions.

 I am very curious, what is the cause of the very high memory latency in the Phenom systems in Sandra's bench?

 Also, I am wondering exactly what air cooler you used in the overclocking tests?

 Thanks! Great parts coming down the line, hard to resist an upgrade with these results!

.

Kamrooz
2008-03-03T14:31:30Z

Lolthotech: Unfortunately, AMD took a backwards plunge when it came to memory performance on K10...Mostly due to the architecture and route they took with their l2/l3 cache. I'm not 100% sure on this one though, so if someone more informed on the barcelona architecture could drop their thoughts, I'd appreciate it. But the x2's have a much higher memory bandwidth then Barcelona when you compare them via synthetic benchmarks. I really do need to read more into barcelona...To get a bigger picture of why it's so flawed 

Crisis Causer
2008-03-03T18:18:44Z

I just know that Phenom's L3 cache runs at a fixed speed (1800Mhz iirc), unlike L2s which run at the CPUs speed.  This is why scaling isn't so good with Phenom's as other CPUs; less performance is gained per MHz.  I don't know why the synthetic benchmarks are weird thoug.  Just felt like pointing that out.

Lithotech
2008-03-03T20:33:19Z

Crisis Causer wrote:

I just know that Phenom's L3 cache runs at a fixed speed (1800Mhz iirc), unlike L2s which run at the CPUs speed.  This is why scaling isn't so good with Phenom's as other CPUs; less performance is gained per MHz.  I don't know why the synthetic benchmarks are weird thoug.  Just felt like pointing that out.

Yeah, I thought I understood what was going on with the Phenom, it's HT link and L3 cache... but some numbers I saw in a recent magazine (Maximum or CPU Mag -- dunno which) were not nearly that far off, they were getting 88-98ns for latency. I remember now that this is with gangmode off, latency went up. With gangmode on the numbers were only a 10ns or so up from what the same ram would get in an Intel system. So there is still a couple hundred ns that's not accounted for here!

I'll dig around for the issue and if I find it will report back to clarify.

 .