H[]R3NDUS

my brother inlaw sent this along to me so i thought i would pass it along...


IT EMERGED yesterday, here under the swaying old Taipei Hyatt, that Intel's high-end desktop darling with dual X16 PCI-E v2 graphics slots - the X38 chipset - is just slightly better off than the P35 when it comes to memory performance at a given FSB and memory frequency combo.
How much is that "slightly better off"?, we're sure you're dying to know.

We ran Sandra XI memory bandwidth and latency benchmark on two high end X38 pre-production boards, both based on the A0 chipset stepping (the A1 one, around any time now, should be a bit better, though). The vendor, whose name is protected for the obvious reasons, put forward two configurations for us: one based on DDR2 memory, the other based on DDR3. Both used the dual core Intel E6850 3GHz / FSB1333 CPU clocked to 3.33GHz / FSB1667, the favourite clock speed setting. The DDR2 one used Geil memory running at 833 CL3-3-3-5 setting, while the DDR3 config used 1667 CL7-7-7-14 setting.

How did they fare? Well, at around 7.5 GB/s Sandra memory bandwidth for DDR2, and 8.2GB/s for DDR3, the boards are a percent or two above the best P35 scores I could get, but still some 10% and 2% respectively behind similar setting scores on Nforce 680i DDR2. Obviously, the BIOS tuning isn't a done job yet, but it seems that there really is no drastic improvement compared to the P35, and that the same "high bandwidth, high latency" optimisation like in the P35 still stands, compared to Nvidia's focus on matching in sync bandwidth, low latency setup. After all, X38 will have even higher DRAM vs FSB multiplier settings to enable some really fast RAM running, but with FSB barely able to make use of it.

The joke, though, is that the twice-as-expensive DDR3 setup still can't match a year-old one with cheaper DDR2 memory. I guess we got to give more time for DDR3 to mature. Of course, the final X38 stepping, combined with BIOS improvements, should give even better memory results

eva2000

Hmmm your brother in law probably got that info from some site as it sounds familiar

dinos22

OMG i hate these rubbish writeups

do some proper tests and report back you dimwiths and none of that sandra ****

H[]R3NDUS

Yes he did Eva im not sure where though...he just emails me Daily about Intel eheh
this sounded interesting though..as X38 does not look to be much Faster than the current chipsets.

wazzer96

More teaser blabber about X38 from Gary Key, AT's motherboard "guru" - heh
now he says a "respin" of the chipset was mandated due to bugs which will push intro out a bit further

http://forums.anandtech.com/messagev...&enterthread=y

Hi,

Intel had to respin the X38 due to compatibility issues with a few PCIe 2.0 cards and some other items of interest. The latest chipsets have been delivered to the manufacturers and most are in full production at this point. A few of the boards required component changes so those designs are being retested at this point. It appears now that retail boards will probably show up by the 9/24 release date although it is looking more and more like supply and additional designs will not be in plentiful supply until October.

Also, early tests on final silicon indicate that the chipset performs better with DDR3 than DDR2 on like boards, hint DDR2-1066 at 4-4-4-12 and DDR3-1066 5-5-4-12 perform within a percent of each other on this chipset which is something we could not say on the P35. We are running overclocked benchmarks at this time and seeing DDR3 showing significant gains over DDR2 up top with the X38, however, still waiting on final BIOS releases so it could change. I say this, as the last BIOS spin for the DDR3 boards improved performance up to 9% in several benchmarks but we have not received updated DDR2 code in over a week.

T_M

Benny in Indo has tested two X38's in the last couple days that both bugged

dinos22

yup he just got the new ones in

moloko

That is the most rubbish thing I've read in some time (the OP).

Totally agree with Dino.

Test max FSB, max staple memory OC's, just do something legitimately worth reading about. It was nothing but a pathetic vomit of a couple paragraphs. Not having tested it in direct comparison myself, I'm still very confident that the difference between P965 and P35 was fairly marginal. However we still all relished in some of the offerings from P35 - higher dividers, stronger dividers, stronger FSB tweaking, generally higher max FSB... the list goes on. I'm confident that X38 will be much the same deal as the P965>P35 relationship was. Additionally as P35 brought with it DDR3 support, X38 will bring with it native support for 45nm chips and 1600FSB (400MHz) chips, so the progression of technology continues and the release is a positive in the grand plan.

I'll be getting the media samples shortly (i.e. not full commercial grade silicon) and the commercial samples should be landing on the doors of some of au's best in due time also. The media sample is shipped so my estimate is it should be another week or possibly two until we see the commercial samples being put through their paces. That's when the true testing will occur.