Intel X58 Express Chipset B-2 to B-3 Stepping Change

[003]

i think x58 has a problem with pci-e interface. think that core i7 is a very powerful cpu as we can all see from cpu intensive tests but when it come to games we cant see its performance. now we know cpu has the power and we know what gpus can do so only thing that might cause this situation is x58.

Umm. Is that not called being GPU bottlenecked?

IOH is the ICH10 component...

So if I've been giving more voltage to the IOH in the BIOS, that is not the chipset, but the SB? Then would the chipset be the ICH? What the hell? Then why is the SB called ICH10R?

[generics_user]

Umm. Is that not called being GPU bottlenecked?



So if I've been giving more voltage to the IOH in the BIOS, that is not the chipset, but the SB? Then would the chipset be the ICH? What the hell? Then why is the SB called ICH10R?

I/O Controller HUB

maybe just another short term

[sholvaco]

Well... according to Intel's own chipset diagram the X58 NB is reffered to as X58 IOH so it is in fact the northbridge getting the new stepping.

[dirtypop]

Do you think this update will allow the X58 to support upcoming hexacore procs?

That would be worth waiting to upgrade for....

[STaRGaZeR]

IOH is not the southbridge

[sholvaco]

Do you think this update will allow the X58 to support upcoming hexacore procs?

That would be worth waiting to upgrade for....

This slide from last week's 32nm presentation confirms that X58 is going to remain as the platform for westmere hexes (Gulftown), so technically any X58 motherboard should do (probably requires a bios update though).

[Revv23]

That would be a nice upgrade cycle for people spending a lot on this platform right now.

[saaya]

Unless you get a crap cpu that can't even run at 3,9 stable without atleast 1.425V....

yeah, true... but the newer i7 chips seem to clock VERY nice, and on water bclock is already limiting quite some 920s and even one air some chips are limited.

and yes, pciE improvements on x58 would be nice

qpi multi has nothing to do with the chipset, its the cpu that has the hardcoded multis on.

how can you claim this and sound oh so sure when you are clearly wrong? you cant possibly know this since you dont have the tools and information available to KNOW it, so how can you make a bold claim like this?

we managed to force a qpi multiplier of 12x on BloodRage, but unfortunately its only the cpu that can actually run that multiplier, the IOH remains at 18x since it doesnt support a lower multiplier.
since both ioh and cpu have to run at the same speed, the IOH pulls the QPI speed of the cpu up, but if you read the cpu registers you can see that the cpu qpi multiplier is set to 12x, which is reported correctly by cpuz as well btw.

this was hinted to me at idf by the nehalem chief architect as well, when i asked him if it was possible to run a lower than 18x qpi multiplier he said "the cpu can run any qpi mutiplier..."

Well, it might be the CPU or the chipset hitting the QPI wall. I wonder which one.

depends on what you mean, the bclock/qpi clockspeed or the qpi multiplier. the ioh has the qpi multiplier stuck at 18x or higher, as for bclock/qpi clockpeeds, im not sure but i think its the ioh. newer cpus can reach high bclocks and qpi clocks with less vtt, but they are still stuck at the same 4ghz/8gts which points towards the ioh limiting.

and if you remember x38 and p35 and p45 etc, they all cr4p out at almost the same pciE speed, memory speed and fsb speed, there is almost no variation. just like with qpi where almost all boards stop at the same qpi/bclock speed.

IOH is the ICH10 component...

that makes as much sense as saying the cpu is the gpu component? :P

The X58 PCIe/QPI/DMI Switch chip is NOT updated. Only the ICH10R component chip that handles I/O. No change in BCLK or the actual X58 chip. Only thing affected is its I/O companion chip that goes from B2 to B3.

then the quote is wrong? i didnt see the release note at work, but if he quote it properly then your wrong.
cause it says the IOH is updated from B2 to B3 and the IOH is X58, NOT the ICH southbridge...

[Loque]

way to burn those who got the first x58 boards..

[mikeyakame]

way to burn those who got the first x58 boards..

Chipset is getting a revision. Nothing new about this. This happens all the time. Nothing to do with burning people. It's just part of the maturity process.

[003]

Chipset is getting a revision. Nothing new about this. This happens all the time. Nothing to do with burning people. It's just part of the maturity process.

That.. and I still don't know what was the issue with it in the first place? The 222 bclock limitation? I highly doubt intel would be releasing a new revision that would make overclocking the non EE i7s any easier, and hurt the 965 sales.

[mikeyakame]

That.. and I still don't know what was the issue with it in the first place? The 222 bclock limitation? I highly doubt intel would be releasing a new revision that would make overclocking the non EE i7s any easier, and hurt the 965 sales.

It's as saaya and bingo13 have hinted. Minor bugfixes on X58 IOH, possibly to do with PCIE or QPI. BCLK limitation would be on the CPU itself.

What 222mhz BCLK limitation by the way? If you check the Overclocking forum, 241mhz BCLK has been reached.

Heres Errata update for B2 stepping. Fixes will most likely be some of the higher priority bugs found in this list. No errata update for B2 -> B3 stepping is publicly available yet.

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

[posershadow]

It's as saaya and bingo13 have hinted. Minor bugfixes on X58 IOH, possibly to do with PCIE or QPI. BCLK limitation would be on the CPU itself.

What 222mhz BCLK limitation by the way? If you check the Overclocking forum, 241mhz BCLK has been reached.

Heres Errata update for B2 stepping. Fixes will most likely be some of the higher priority bugs found in this list. No errata update for B2 -> B3 stepping is publicly available yet.

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

222bclk is around 4ghz qpi or 8gt/s. That is where saaya says the IOH runs out of steam. 241 has been reached, but more likely by luck and a good board. You don't see many boards that want to get that high. The DFI seems to be one of the consistent top performers, but if you look at hwbot, boards like the gigabyte extreme and asus r2e really don't have bclk records above 220.

[saaya]

Do you think this update will allow the X58 to support upcoming hexacore procs?

That would be worth waiting to upgrade for....

that has nothing to do with the ioh, just needs a bios update and in theory some boards might have a too weak pwm, but even the cheap boards from ecs and biostar and asrock have decent pwms, so they should work with hexacores at stock and even some overclocking without any problems

way to burn those who got the first x58 boards..

how so?
and did 965 owners get burned now that the 975 is out?
did gtx280 owners get burned after the gtx285 came out?
and those are releases that actually impacted performance quite a bit, while the stepping change seems to be nothing but a weekend bugfix bbq

That.. and I still don't know what was the issue with it in the first place? The 222 bclock limitation? I highly doubt intel would be releasing a new revision that would make overclocking the non EE i7s any easier, and hurt the 965 sales.

yeah, not on purpose at least...
but everytime they improve yields or tdp they tend to improve overclocking as well, obviously not on purpose

unfortunately this update wont bring any ocing benefits it seems...

What 222mhz BCLK limitation by the way? If you check the Overclocking forum, 241mhz BCLK has been reached.

thats just beeing picky, sure with tweaking you can get a tad higher, but the highest stable is still around 230 which is less than 10% above 222.

the real limitation is qpi clockspeed, which newer cpus can run slightly higher, but still, its limited and will always be, since its running at 4ghz with 222bclock, which is way way high and notable speed boosts are unlikely.
the only way to get notably higher bclocks in the 300 range is if the qpi clock can be reduced or kept the same with higher bclock speeds, and thats what you need lower qpu multipliers for.

maybe intel will add it in x68, lets hope so...

222bclk is around 4ghz qpi or 8gt/s. That is where saaya says the IOH runs out of steam. 241 has been reached, but more likely by luck and a good board. You don't see many boards that want to get that high. The DFI seems to be one of the consistent top performers, but if you look at hwbot, boards like the gigabyte extreme and asus r2e really don't have bclk records above 220.

its not really the board per-se, its the bios that needs some tweaking to reach those slightly higher bclock speeds.

in the end those high bclocks dont really make all that much sense, even the latest 920s dont hit 4600mhz (220x21) on air so bclock isnt limiting them, even on water thats not a limiting factor. only on phase change and ln2, but if your running that kind of cooling you will most likely be running a 940 or 965 anyways since you invested quite a bit in cooling to have a fast rig.

anyways, lets keep praying that intel will allow lower qpi multis soon, id love to see 300 bclock just for the heck of it, even if its not needed and means no perf boost over higher multi and 133 bclock

[FuryPT]

So, basically, this new stepping is already on the mobos at this moment?

I wanna buy a X58 UD5 but i want this new stepping XD

[Zorlac]

Not sure if anyone knows of a mobo with this chipset stepping yet. Intel hasnt even released a PCN for a new DX58SO revision yet... =/

I bet we will have to wait for Computex in June.

Hoping Intel will have to TRIM based SSDs as well

[Rasamaha]

well, ES i allready done
You can find it on asus p6t7ws

Other photos
http://photofile.name/users/rasamaha...60/?mode=small

[Cliff Burton]

I hope EVGA Classified adopt it too on a new PCB revision.

[003]

Guys if I ordered a Rampage II Gene in mid-July, it should have a B3 revision X58, correct?

[Anemone]

I think we are waiting to see folks come back and report that B3's are in the channel.

[AuDioFreaK39]

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