Xeon W3520 & I7-920

[spylake]

I am trying to figure out what the differences are between a W3520 and an I7-920.

Comparing their data sheets gave me the following pin differences

PIN I7-920 -- Xeon W3520

A36 RSVD -- DDR0_ECC[1] CMOS I/O
A37 RSVD -- DDR0_ECC[5] CMOS I/O
B34 RSVD -- DDR0_ECC[6] CMOS I/O
C33 RSVD -- DDR0_ECC[3] CMOS I/O
C34 RSVD -- DDR0_ECC[7] CMOS I/O
C36 RSVD -- DDR0_ECC[0] CMOS I/O
C37 RSVD -- DDR0_ECC[4] CMOS I/O
D34 RSVD -- DDR0_DQS_P[8] CMOS I/O
D35 RSVD -- DDR0_DQS_N[8] CMOS I/O
D36 RSVD -- DDR1_ECC[0] CMOS I/O
E29 RSVD -- DDR2_ECC[2] CMOS I/O
E30 RSVD -- DDR2_ECC[3] CMOS I/O
E33 RSVD -- DDR1_ECC[2] CMOS I/O
E34 RSVD -- DDR1_ECC[6] CMOS I/O
E37 RSVD -- DDR1_ECC[4] CMOS I/O
F30 RSVD -- DDR2_ECC[7] CMOS I/O
F31 RSVD -- DDR2_ECC[6] CMOS I/O
F32 RSVD -- DDR0_ECC[2] CMOS I/O
F33 RSVD -- DDR2_ECC[1] CMOS I/O
F36 RSVD -- DDR1_ECC[1] CMOS I/O
F37 RSVD -- DDR1_ECC[5] CMOS I/O
G29 RSVD -- DDR2_DQS_P[8] CMOS I/O
G30 RSVD -- DDR2_DQS_N[8] CMOS I/O
G33 RSVD -- DDR1_DQS_P[8] CMOS I/O
G34 RSVD -- DDR1_DQS_N[8] CMOS I/O
G35 RSVD -- DDR1_ECC[7] CMOS I/O
G36 RSVD -- DDR1_ECC[3] CMOS I/O
H32 RSVD -- DDR2_ECC[0] CMOS I/O
J30 RSVD -- DDR2_ECC[5] CMOS I/O
J31 RSVD -- DDR2_ECC[4] CMOS I/O

Both use the X58 Express chip set. Both have only one QPI bus and both are for single CPU systems.

The added pin assignments are obviously to support EEC or Error Correcting Code memory.

The next question is whether a W3520 will work in a MB intended for the I7-920 or if the Memory pathways and memory cards must be ECC DDR3 RAM.

The W3520 is the same price as the 920 so the only extra cost would be the MB and the ECC RAM.

Anyone have more data to fill in my gaps? Is ECC DDR3 on the market yet? Are any MB ECC configured?

 

[ilkhan]

physically theyd be identical, they're both bloomfields. You'd see some microcode differences for ECC (I think the xeons support it) but not much else.

 

[Rick James]

The Xeon has 2 QPI's. The other is too talk to a 2nd processor

 

[Idontcare]

Originally posted by: spylake
I am trying to figure out what the differences are between a W3520 and an I7-920.

Comparing their data sheets gave me the following pin differences

[snip]

Both use the X58 Express chip set. Both have only one QPI bus and both are for single CPU systems.

Originally posted by: Rick James
The Xeon has 2 QPI's. The other is too talk to a 2nd processor

As ilkhan and Rick James have mentioned, all Nehalem-based products on the market today physically have two QPI's on the die and only the 5500 series Xeons have the second QPI validated and activated for use on 2S systems.

See page 13 of Intel® Xeon® Processor 5500 Series Datasheet Volume 1

Originally posted by: spylake
The added pin assignments are obviously to support EEC or Error Correcting Code memory.

The next question is whether a W3520 will work in a MB intended for the I7-920 or if the Memory pathways and memory cards must be ECC DDR3 RAM.

The W3520 is the same price as the 920 so the only extra cost would be the MB and the ECC RAM.

It is ECC compatible, but ECC is not required.

Think AMD's X4 955...you could put it in an AM2+ socket and run it with DDR2 but you also have the option of dropping it into an AM3 socket and running it with DDR3, something you cannot do with its younger sibling the X4 940.

For the LGA1366 Nehalem-based xeons they will all function with standard non-ECC DDR3.

The ability to plug a 55xx Xeon into a desktop grade LGA1366 mobo will depend entirely on the mobo manufacturer updating the BIOS to support the CPU, nothing more complicated than that.

Originally posted by: spylake
Is ECC DDR3 on the market yet?

shopping results for "ECC DDR3"

Originally posted by: spylake
Are any MB ECC configured?

shopping results for Xeon 5500 motherboards with ECC support

 

[spylake]

Originally posted by: Rick James
The Xeon has 2 QPI's. The other is too talk to a 2nd processor

You are thinking of the 5500 series.

The W3520 is a i7-620 with ECC added, It is single chip only and has only one QPI or at least only one that works

 

[spylake]

Originally posted by: Idontcare

Originally posted by: spylake
I am trying to figure out what the differences are between a W3520 and an I7-920.

Comparing their data sheets gave me the following pin differences

[snip]

Both use the X58 Express chip set. Both have only one QPI bus and both are for single CPU systems.

Originally posted by: Rick James
The Xeon has 2 QPI's. The other is too talk to a 2nd processor

As ilkhan and Rick James have mentioned, all Nehalem-based products on the market today physically have two QPI's on the die and only the 5500 series Xeons have the second QPI validated and activated for use on 2S systems.

See page 13 of Intel® Xeon® Processor 5500 Series Datasheet Volume 1

Originally posted by: spylake
The added pin assignments are obviously to support EEC or Error Correcting Code memory.

The next question is whether a W3520 will work in a MB intended for the I7-920 or if the Memory pathways and memory cards must be ECC DDR3 RAM.

The W3520 is the same price as the 920 so the only extra cost would be the MB and the ECC RAM.

It is ECC compatible, but ECC is not required.

Think AMD's X4 955...you could put it in an AM2+ socket and run it with DDR2 but you also have the option of dropping it into an AM3 socket and running it with DDR3, something you cannot do with its younger sibling the X4 940.

For the LGA1366 Nehalem-based xeons they will all function with standard non-ECC DDR3.

The ability to plug a 55xx Xeon into a desktop grade LGA1366 mobo will depend entirely on the mobo manufacturer updating the BIOS to support the CPU, nothing more complicated than that.

Originally posted by: spylake
Is ECC DDR3 on the market yet?

shopping results for "ECC DDR3"

Originally posted by: spylake
Are any MB ECC configured?

shopping results for Xeon 5500 motherboards with ECC support

So if I understand correctly then a Xeon W3520 is a Bloomfield CPU with one active QPI and ECC enabled. It will operate in a standard X58 MB with or without ECC memory even though the MB makers won't certify it for such use?

 

[Idontcare]

Originally posted by: spylake
So if I understand correctly then a Xeon W3520 is a Bloomfield CPU with one active QPI and ECC enabled. It will operate in a standard X58 MB with or without ECC memory even though the MB makers won't certify it for such use?

Does the W3520 exist yet?

If yes then as you state it is how it would exist, 3000 series XEONs are single-socket and in the case of Nehalem EP they are LGA1366 compatible and x58 chipset compatible pending proper BIOS update exists.

For some reason my mind read your repeated usage of the SKU W3520 and kept translating that to the existing W5580 SKU.

When was the W3520 released?

 

[ilkhan]

As far as I know Intel only mints one die of nehalems as of yet, bloomfields; and they'll all work on the same socket, s1366.

 

[spylake]

Originally posted by: Idontcare

Originally posted by: spylake
So if I understand correctly then a Xeon W3520 is a Bloomfield CPU with one active QPI and ECC enabled. It will operate in a standard X58 MB with or without ECC memory even though the MB makers won't certify it for such use?

Does the W3520 exist yet?

If yes then as you state it is how it would exist, 3000 series XEONs are single-socket and in the case of Nehalem EP they are LGA1366 compatible and x58 chipset compatible pending proper BIOS update exists.

For some reason my mind read your repeated usage of the SKU W3520 and kept translating that to the existing W5580 SKU.

When was the W3520 released?

Seems to exist

http://www.provantage.com/inte...0601w3520~7ITEP31R.htm

 

[Idontcare]

Originally posted by: spylake

Originally posted by: Idontcare

Originally posted by: spylake
So if I understand correctly then a Xeon W3520 is a Bloomfield CPU with one active QPI and ECC enabled. It will operate in a standard X58 MB with or without ECC memory even though the MB makers won't certify it for such use?

Does the W3520 exist yet?

If yes then as you state it is how it would exist, 3000 series XEONs are single-socket and in the case of Nehalem EP they are LGA1366 compatible and x58 chipset compatible pending proper BIOS update exists.

For some reason my mind read your repeated usage of the SKU W3520 and kept translating that to the existing W5580 SKU.

When was the W3520 released?

Seems to exist

http://www.provantage.com/inte...0601w3520~7ITEP31R.htm

Yeah I'm just going by Intel's official processor spec website and they don't list any W35xx series processors yet, hence my question regarding the release timeline.

http://processorfinder.intel.c...ProcFam=528&SearchKey=

 

[Devildoll]

Originally posted by: Idontcare

Yeah I'm just going by Intel's official processor spec website and they don't list any W35xx series processors yet, hence my question regarding the release timeline.

http://processorfinder.intel.c...ProcFam=528&SearchKey=

Look harder , a google on w3520 brings up this on hit nr 4

W3520

and also if you go onto intels website and click Products > proccessors > Intel® server processors > Intel® Xeon® processor 3000 sequence ( click the logo ) > specifications tab

you will find w3520 on the third row.

The w3520 is supported by a couple of consumer mobo's and work with regular ddr3.

heres a thread on the asus forum where a guy shows he's w3520 running on a P6T Deluxe V2 mobo ( last post i think )




Asus forum

CPU-Z Image


also ... first post hi everybody

 

[ListedFirstcom]

Well Hello my Anandtech friends!

I have some results from a recent test of mine that I think you may find useful...

I'm a Mac guy, and I'm also a bit of a hack, (You can find me doing crazy stuff like this) MAcBook Air EVDO Hackhttp://www.engadget.com/2008/04/29/macbook-air-hacked-with-internal-ev-do-finally-live-up-to-its/


Anywho, I just got myself one of those spiffy new Quad core Mac Pros, with the W3520 2.66GHz CPU in there. The question you ask is the question I had. So I figured what better way to find out than to try?

Bought myself an i7 Extreme CPU, cracked open the Mac, Voided any and all warranty and slapped the little ah heck in.

The system works flawlessly! I bought some OCZ Platinum triple channel memory, just in case the ECC memory was an issue. Turns out, it's not. The i& Extreme works just fine with the ECC memory in there (I'm using the OCZ Platinum as it's much faster though). The mac even recognized the CPU as an Intel Xeon... I sure wasn't expecting that.

Anyhow, if any of you have any questions for my, or would like me to run any particular tests, feel free to give me a shout!


I DO have one question for you all though... Is there any way of Hardware overclocking the CPU? The Macs aren't exactly known for their ability to overclock. I overclocked the previous generation of Mac Pro (a 2.4GHz CPU to 3.0GHz) with the x5300 Series CPUs, using the BSEL mod (Which was tape on pin #26). If anyone of you know if this type of overclock is possible with the i7, Please let me know!

Thanks guys!

 

[Devildoll]

Originally posted by: ListedFirstcom

I DO have one question for you all though... Is there any way of Hardware overclocking the CPU? The Macs aren't exactly known for their ability to overclock. I overclocked the previous generation of Mac Pro (a 2.4GHz CPU to 3.0GHz) with the x5300 Series CPUs, using the BSEL mod (Which was tape on pin #26). If anyone of you know if this type of overclock is possible with the i7, Please let me know!

Thanks guys!

if you got a regular consumer motherboard , and apple didnt put a fancy bios with locks on it .

you should be able to just hit delete during the POST ( power-on self test) ( bootup)
and yeah , you'll find all the settings in there .

if you got a manual for that motherboard , they usually do a very good job of explaining every single option in there.

if apple did put a custom bios on , check out what model the mobo is , get a regular bios for it and flash , at your own risk you should read up on flashing if you are not familiar with it. but as you said you like tweaking with stuff , i think this might be up your alley

 

[ListedFirstcom]

Unfortunately Apple doesn't use a motherboard as we all know it. It's a "Logic Board" and runs EFI, rather than bios. There is nothing, and nowhere to go into. It's all controlled through the OS and flash updates. As far as I know, there isn't a known way to flash the EFI firmware either. This is why the only option for overclocking the Xeon CPUs in the previous generation was the tape, as it made the Mac EFI firware think it had a 1333Mhz bus CPU in there...


Thanks for the post though!

 

[bwanaaa]

newegg has delisted the 3520. is it going the way of the i940?

 

[spylake]

Originally posted by: bwanaaa
newegg has delisted the 3520. is it going the way of the i940?

I think Delisting is their way of saying "Out of Stock" the i7-920 was delisted for a long time when it was in short supply.

 

[spylake]

New reason to only buy a DO stepping. The current issue of a magazine, that shall go nameless, benched a 920 CO vs the Phenom and a top Core 2 dual.

The 920 was well ahead in all benchmarks except for "Fear" where the 920 was around 57 fps and the phenom was over 200 fps. The text stated that Intel explained the performance discrepancy was due to a USB bug. They didn't specifically identify whether the bug was in the CPU or the X58 support chip.

The bug below was corrected in the DO stepping of the I7-920 and the W3520

AAJ44 USB 1.1 ISOCH Audio Glitches with Intel® QuickPath Interconnect Locks and Deep C-States
Problem: An interrupt directed at a Core in C3 or C6 that collides with an Intel® QuickPath Interconnect (Intel® QPI) Lock sequence may delay ISOCH transactions to DRAM long enough to underrun USB 1.1 buffers.

Implication: USB 1.1 Audio devices may have audio glitches.

Workaround: None identified.

It is hard to tell if this is the bug they were referring to but it could be.

There are also USB related issues with the current stepping of the X58 MB chips. Unfortunately I think the B3 stepping is a ways off and I can't think of any way to know which stepping a particular MB has. The boxes won't likely say and the chip itself is under a heat sink.

The update data for the X58 chipset is here

http://www.intel.com/Assets/PDF/specupdate/320839.pdf

The update data for the D0 CPU stepping is here

http://download.intel.com/desi...or/specupdt/320836.pdf

 

[sgrinavi]

Originally posted by: spylake

Originally posted by: bwanaaa
newegg has delisted the 3520. is it going the way of the i940?

I think Delisting is their way of saying "Out of Stock" the i7-920 was delisted for a long time when it was in short supply.

Well... usually when something is out of stock they give you the opportunity to get an email alert. This item has been deactivated. I'm glad I got mine when I did. This is a sweet processor, too bad the mother board doesn't support it properly. Pffft...

 

[Nemesis 1]

Well These are all DO chips so there very good . The 920 is unreal . Check this out. I guess we haven't seen anything yet . All I can say is I want that 32nm 2 core and I want it now . I believe now what brother inlaw said. If the 45nm DOs can do this and better I am hyped for 32nm. Guys are hitting 5ghz AIR not stable . Water maybe stable.

http://valid.canardpc.com/show_oc.php?id=571685


Now this was done by coolar - so take it from there nor best source.

 

[Nemesis 1]

This is probably more correct.

http://valid.canardpc.com/show_oc.php?id=542221

Have to wait and see , Fugger is said to have monster chip . well see.

I guess the 975 is going to kick ass as it should for price

 

[Idontcare]

Unfortunately for whatever reason cpuz validation does not capture the Vcore used to attain the clocks, so we really have no idea how much of a "quality" overclock we are looking at without being told the rest of the story.

If it is 24/7 voltage or suicide voltage kinda makes a difference on the level of "wow" it generates.

Also I see that 975 is an ES, takes the "wow" down a notch as well.

At some point in time the enthusiast community is going to need to have a serious discussion amongst itself regarding what constitutes viral marketing versus entertainment. Every time they try to have the discussion over on XS Fugger just goes thru and bans the people from the thread. Hardly an uncensored situation over there.

Seeding recognized individuals with special binned silicon so they can parade around with the WR's isn't exactly a grass-roots hobby/enthusiast sport any more IMO.

I'm sure they and their sponsors would beg to differ, but I just see this as having degraded the world-record situation to the point of viral marketing.

 

[Nemesis 1]

Your kidding me right . Over ayear ago I was bitching abour Xtreme Spam. I got plowed under.

Now I have seen a few of coolars post threw the years . I can't ever recall him being called out like DR. WHo did. I guess Intel tring to stop leaks. The water cooling section at XS is all spam. The test run are a joke . Basicly they are selling products from, Coolers to Ram to MB.

The only thing xtreme about xtreme now is the trolling and flamebaiting. Befote C2D was best site now it sucks.

 

[Tempered81]

 

[Idontcare]

Yeah I saw our boy DrWho in action there again. We all know there is no such thing as a leak, they are all official leaks intended to be viral marketing per usual. DrWho jumping in like that is just the latest twist to make it seem authentic again as if renegades are running amok and railing against Intel's lawyers. Hardly.

If it was a legitimate concern by DrWho it would be handled by pm to the poster and then followed up by a mod removing the posts and threads as deemed necessary. Silly nonsense all this viral crap.

 

[Nemesis 1]

Well I like getting facts Viral or not as long as there TRUE facts.