Intel D975XBX (Bad Axe)

[pointreyes]

Originally Posted by Sandog

I don't believe AMD's move to DDR2 will be smooth at first...

That probably will not matter to the AMD crowd. Look at how picky DDR is with even the current AMD boards. It's almost an expectation that RAM will be a problem with an AMD system.

[lowrider007]

Originally Posted by Sandog

Anybody have the specs for the Conroe EE? I heard 3.0 or maybe 3.3GHz

tbh I can't see it being 3.0Ghz, I'd bet on 3.3Ghz myself as the conroe that was in the latest bench tests that whooped the 2.8 fx60 was 2.6 and that was rated as a low to mid range conroe, I think were looking at low range 2.3, mid range 2.6, high range 2.9/3.0 and 3.3Ghz for the EE.

[Sandog]

Hey Pointreyes,

My two buds comps can not go above 440MHz on their San Diego cores. I think AMD will have a time of it getting to 533MHz on the same chips plus I don't think it will do as much good as people think. You gain more latency with DDR2, something the onboard memory controller does not need. Plus I don't think it will be a smooth transition, it never is. There were also a few probs with the NF4 chipset. I have heard a few complaints of lousy chipset drivers and features that do not work like ActiveArmor. So now NF5 will be the new king? I am sure it will have more bugs than the last 3 NF chipsets. AMD has gotten to be a cocky lot. From the article link above they say they are going down a totally new road which is good for them. Thing is Intel will have the markets so locked up in 1yr that it won't matter, nobody will follow them (IMO). We will see. AMD is a tough company, most corporations chew the little guys up or buy them. AMD has really been able to take a bite out of Intel and it has all been good for everybody. Can't wait for Conroe.

[Sandog]

Originally Posted by lowrider007

tbh I can't see it being 3.0Ghz, I'd bet on 3.3Ghz myself as the conroe that was in the latest bench tests that whooped the 2.8 fx60 was 2.6 and that was rated as a low to mid range conroe, I think were looking at low range 2.3, mid range 2.6, high range 2.9/3.0 and 3.3Ghz for the EE.

I recall the U.K. or Europe getting EE chips before the States did. Maybe you will beat us to it.

[Intel_User]

Official JEDEC DDR/DDR2/DDR3 timings.

DDR: 3
DDR2: 3/4/5
DDR3: 5/6/7

Now if anyone thought CL5 was a problem, wait till you see CL7....

Here's what I think with AMD and DDR2. Remember Athlon 64's have integrated memory controllers right?? And that lowers memory latency??

Total memory latency=Memory Access latency+Distance between CPU and Memory controller(Let's say its Tcm)

Its easy to see that its true. CPU will take time to talk to the memory controller, and after that the memory module will take time accessing the memory.

If hypothetical AMD Tcm is 80, and Intel's 160, which CPU would be more significantly impacted by adding say, 5 to the number??

What I am saying is AMD will be more impacted by the high latency than Intel will.

[Intel_User]

That being said, Intel could have somehow put unfair optimizations in Conroe, looking how AMD's number in the IDF system looks about right. It was said that the ATI driver has "Conroe specific" optimizations. I wonder why a graphics driver needs changes to support a CPU??

[Belisarius]

Originally Posted by Intel_User

That being said, Intel could have somehow put unfair optimizations in Conroe, looking how AMD's number in the IDF system looks about right. It was said that the ATI driver has "Conroe specific" optimizations. I wonder why a graphics driver needs changes to support a CPU??

Mate this is a point of conjecture about Intel rigging the numbers, but in my eyes why would they risk it? If they release Conroe and it don't perform like that on the street, it'll be goodnight Intel, they are already losing market share, albeit only retail, but it's still a kick in the pants. Maybe AMD would get it to a 50/50 market share, imagine that and the cut throat price cuts to undermine each other

Maybe the doubters are right though?

[GFORCE100]

Originally Posted by Intel_User

Official JEDEC DDR/DDR2/DDR3 timings.

DDR: 3
DDR2: 3/4/5
DDR3: 5/6/7

Now if anyone thought CL5 was a problem, wait till you see CL7....

Here's what I think with AMD and DDR2. Remember Athlon 64's have integrated memory controllers right?? And that lowers memory latency??

Total memory latency=Memory Access latency+Distance between CPU and Memory controller(Let's say its Tcm)

Its easy to see that its true. CPU will take time to talk to the memory controller, and after that the memory module will take time accessing the memory.

If hypothetical AMD Tcm is 80, and Intel's 160, which CPU would be more significantly impacted by adding say, 5 to the number??

What I am saying is AMD will be more impacted by the high latency than Intel will.

Modern graphics cards using DDR3 already have much higher latencies in the regions of CAS = 12 or 14 let alone the other values that also comprise the final memory latency number.

Latency is never a good thing to have too much off in any walk of life, even if it means delaying buying your wife flowers As a remedy, the IT industry just seems to go along the logic that if we make the memory bandwidth high enough we can sleep well with regards to the high latency as the former will assist the latter and all will be okie dokie. This doesn't mean a better latency wouldn't help things more, just that they do whatever's feasible, make the cash, move on and not feel bad about it. As processors have gotten faster though, more latency is just a nail in their coffin not in terms of death but rather performance. All this said and duster however, FB-DIMM's is a topic that hasn't yet been discussed here at present and what holds a certain answer

[GFORCE100]

Originally Posted by Intel_User

Official JEDEC DDR/DDR2/DDR3 timings.

DDR: 3
DDR2: 3/4/5
DDR3: 5/6/7

Now if anyone thought CL5 was a problem, wait till you see CL7....

Here's what I think with AMD and DDR2. Remember Athlon 64's have integrated memory controllers right?? And that lowers memory latency??

Total memory latency=Memory Access latency+Distance between CPU and Memory controller(Let's say its Tcm)

Its easy to see that its true. CPU will take time to talk to the memory controller, and after that the memory module will take time accessing the memory.

If hypothetical AMD Tcm is 80, and Intel's 160, which CPU would be more significantly impacted by adding say, 5 to the number??

What I am saying is AMD will be more impacted by the high latency than Intel will.

Don't forget L1/L2/L3 cache, it's size, access latency etc. Of course system memory access latency is important but a fair number of instructions can be executed from data already within any of these caches depending on the processor design. Cache is always faster than the main system memory and this holds true back when we first started using L2 cache using SRAM chips connected to the motherboards via a COAST socket (Cache On A STick). L1 cache was the Intel 486SX chip. Cache has helped keep performance rolling.

Enthusiasts in this thread should remember Intel's first Celeron 266 back in 1998 based around the Klamath core but no L2 cache. It was very slow for this very reason. Then Intel made the Medecino core at 0.25micron and adding 128KB L2 cache that was only 1/4th of the Pentium II's at that time but ran at full processor speed hence 300 and 333MHz at launch. Keen posters here will also recall this these were SMP enabled at that time making for a rather cheap SMP rig, but as this is a seperate topic it's time for a period.

In short, lets remember how cache plays a role in the system latency, it's not on every clock the CPU needs to access the main memory.

In my opinion the CPU that will lose out most if latency is increased is the CPU whose core(s) can perform most ticks (i.e. cycles) in the period of time = to that of the latency. This is because in computing terms, these cores have been waiting for the longest time if we stay by the rule that 1 cycle = a given set time penalty lost in productivty. For this to work out cycles must imply the highest IPC (Instructions Per Cycle) value from the compared processors.

As the Intel Core is faster clock for clock than the AMD K7, for now I'll point to Intel suffering most. I'll point but only slightly because Intel have got the cache subsystem right which makes me sit comfortably,

[GFORCE100]

Originally Posted by Intel_User

That being said, Intel could have somehow put unfair optimizations in Conroe, looking how AMD's number in the IDF system looks about right. It was said that the ATI driver has "Conroe specific" optimizations. I wonder why a graphics driver needs changes to support a CPU??

Revised SSE code handling and instructions in the CPU require recompiled code to take advantage hence even if the ATI driver used at IDF was tweaked so to speak, we can expect such things to happen anyway and that means in Windows Vista for example and DirectX 10 (not called WGF anymore).

Same theory as when the 57 MMX instructions came about, to use it (even thogh it only operates on Integer values hence is somewhat useless from a game point of view) one had to specifically code it in the code for it to work its magic.

The first game to use MMX back in 1997 with the launch of the Pentium MMX was POD, should ring a bell to some folks in here, Come on, it's only 9 years ago,

Programming can be a pain in the butt, that's why it's perfectly normal to expect programmers to be lazy or slow goers and why before dual or multi-core or even SSE2/SSE3 optimized programs appear as frequently as fresh daily bread, we'll be waiting a little longer, say 5 years before things mature into proper California wine quality

[pointreyes]

Originally Posted by GFORCE100

...say 5 years before things mature into proper California wine quality

Oregonian wine quality would be better.

BTW: Being a SMP (dual PIII, AthlonMP, Opteron, and Xeon) user for many years I'm one of those people that actually faced that frustration of lazy programmers not building software to work properly with the hardware. It was very annoying to see one proc being used for benchmarks or know that a game worked with only seeing one proc. Only now are we finally seeing games that use more than one proc.

[TQ]

Originally Posted by pointreyes

Oregonian wine quality would be better.

Depends on the vinter.

[emgarf]

My BadAxe system parts arrived yesterday... so far, the following parts are assembled:

BadAxe (duh) - BIOS updated to latest (0618, I think)
Pentium D950, stock cooler
(x2) OCZ Gold Series 1GB 240-Pin DDR2 SDRAM Unbuffered DDR2 667 (PC2 5400)
eVGA Geforce 7900GTX 512MB 256-bit GDDR3 PCI Express x16
PC Power & Cooling Turbo-Cool 850 SSI power supply
(x3) Caviar SE16 320GB 7200 RPM 16MB Cache SATA 3.0Gb/s
Benq DW1655 DVD-RW

...all in a Lian Li PC-G70B case.
This is a full tower case, here's a link - http://www.lian-li.com/Product/Chassis/S_C_PC-G70.htm
The case is set up with intake fans at the front bottom and on the rear panel next to the processor.
Exhaust fan is on the side centered over the card slots. Power supply also has its own exhaust fan.
Both intakes and the exhaust fan are all 120mm.

So far, so good as the system boots into BIOS and will boot from CD/DVD, but I have a couple concerns:

1) With the system displaying the BIOS hardware monitoring screen, temps stabilize at:

Processor - 67C
Zone 1 - 47C
Zone 2 - 42C

Fan speeds are:

Proc - 1500
Side exhaust (aux) fan - 1500
Front intake - 1000
Rear intake - 1000

This seems a little warm for a system just sitting in BIOS, although it meets Intel's thresholds of 75/65/65.
I'm concerned that these temps will climb during typical PC activity, although the fan speeds are apparently being kept fairly low by the MB.
Won't be able to evaluate 'typical activity' until after the OS and drivers are installed.

2) The RAM is 667MHz DDR2 CL 4-4-4-8, but the BIOS defaults to 533MHz 4-4-4-10 with that memory installed.
Manually changing the BIOS timing to OCZ's advertised timing seems to cause no ill effects,
but as noted above I'm just entering BIOS at this point and not really running anything. I think one of my
Linux bootable CDs has memtest on it; maybe I'll try that.

Anyone see a problem with the manual timing change?


Thanks in advance for any advice / comments,
emgarf

[pointreyes]

Originally Posted by TQ

Depends on the vinter.

Not when you're an Oregonian.

Oregonians hate California. It's like a quiet Civil War between the two states.

emgarf, see you have the 7900GTX. Those temps do seem high but maybe your bios is setup to make the proc run as quiet as possible? If so that could cause part of the issue. My 830 idles at 55 and I'm using the stock HSF but my case has that huge crossflow fan opening and 2 x 120mm exhaust fans.

[pointreyes]

From ComputerWorld about Apple computers:

Quad-core coming?

In the other big news from IDF, Intel unveiled more details regarding its plans for quad-core chips. Intel said it plans to release its first quad-core chip for desktop PCs, codenamed “Kentsfield,” in the first quarter of 2007. But this chip won’t really meld four central processing units into one chip — as rival AMD plans to do with its quad-core chips coming in 2007.

Instead, Intel’s Kentsfield will combine two dual-core chips into one package. This isn’t the most efficient design strategy, but it will help Intel get quad-core chips to market at the same time as AMD. Of course, Intel will improve the quad-core design as quickly as possible.

For Apple users, quad-core chips will mean a huge amount of horsepower compared to today’s highest-end Apple desktops.