Feeds

back to article Intel set for server chip blitz

High-end server chip rivals Intel and IBM have picked the same day - next Monday, February 8 - to launch their respective quad-core "Tukwila" Itanium and eight-core Power7 processors. As El Reg previously reported, IBM is getting ready to launch the initial Power7-based servers in New York next Monday. The Power7 chips will have …

COMMENTS

This topic is closed for new posts.
Bronze badge

quad socket and up?

I read one other article that seemed to indicate that the new 8 core Nahalems would not be available in a dual socket configuration, is this the case?

If so it seems like Intel will be losing out on quite a bit of stuff as the new 12 core opterons seem geared towards dual socket configs, which are of course by far the most popular server and blade design, and significantly lower cost. Perhaps the 8 core Intel chips are so expensive to make that it doesn't make sense to put them in dual socket systems(yet?), but that would be surprising to me if that was the case. Perhaps intel is trying to lure customers to upgrade to the more profitable quad socket systems while AMD is going the other direction, letting customers stick with dual sockets.

I don't recall the 6-core chip intel released for quad socket systems a year or two ago being such a big hit.

0
0

Tukwila

For those of you who don't live in Washington state, its pronounced TUK-will-AH.

And it is only fitting that they name it that since Tukwila is home to some of the biggest colocation facilities on the west coast.

1
1
Silver badge
Happy

No, no, no....

It's pronounced "Tuk-ZIL-LA", as christened here on the Reg forums by Chris Morley.

0
1
Megaphone

Tukwila is boring

A chip that is only used for HP-UX and Non-Stop OS. I hope they will be taking questions.

1) Will Unisys, SGI or Groupe Bull have Tukwila systems or did they drop support?

2) Why only four cores when Nehalem and POWER7 are both 8 cores.

3) Why less than 2GHz when Nehalem is 3+ and POWER7 is 4+

4) Why announce a chip and not announce actual systems, what is the point?

5) Wasn't tukwila announced two years ago?

http://arstechnica.com/hardware/news/2008/02/intel-shows-off-tukwila-first-2-billion-transistor-cpu.ars

6) Why is HP requiring a fork lift upgrade to Tukwila systems

7) Why is Tukwila finally getting to 65nm when all other chips are 45nm or better?

8) Why is there no core performance increase when software costs are so high and priced per core?

9) How are the Linux sales on Itanium? I hear they are so bad RedHat has dropped support.

10) Will Tukwila finally have hardware virtualization? Itanium seems to be the only processor besides SPARC64 to not provide at least some hardware virtualization.

11) I hear Oracle is going to increase the core price to .75 just like SPARC or higher what is the factor?

12) Is HP going to have a BI offering now that Exadata was canceled and Neoview is not selling?

13) Besides having twice the cores in the same footprint what is the value of Tukwila? The same core and software license performance.

14) Oracle still does not have support for the eBiz suite applications on Itanium..will that ever change?

15) I hear HP is giving away free upgrades to Tukwila, but I don't see the point since applications will not get any more performance.

2
0
Happy

You missed the most important question

16) As Nehalem provides more cores and more performance per core than Tukwila, is this the final chance for customers to migrate to x86-64 before the Itanium architecture is end of lifed? And saying "we have a roadmap" doesn't count.

0
0
Alert

Argh..Oracle pricing

Why do they have to call it a "Itanium 9500"?

Oracle only charges .5 for Itanium 91xx and earlier chips.

The new pricing says Itanium 9500 will be 1 license per core as it now falls under

"All Other Multicore chips"

So the price of Oracle and Weblogic is going to be 2X per core and the core performance is not even going to increase.

http://www.oracle.com/corporate/contracts/library/processor-core-factor-table.pdf

I need to hear about this Power7 stuff.

0
0
WTF?

"A chip that is only used for HP-UX and Non-Stop OS."

Plus VMS (though obviously HP would prefer to keep that quiet). OK VMS on AMD64 would be more attractive in general than VMS on IA64, people might actually buy VMS on Proliant, but HP haven't worked that out yet.

0
0
Silver badge
Happy

RE: Tukwila is boring

Tukwilla is boring? I'm sorry, but I wasn't expecting a CPU to be a great coversationalist. But there are some errors in your post, Allison, and you need to do a little research on this one. "A chip that is only used for HP-UX and Non-Stop OS...." And Windows, OpenVMS, and Linux. Oh, and that's NonStop which is chewing into IBM's mainframe base. Please compare to Power7 when it launches, which will support just AIX and maybe Linux.

"1) Will Unisys, SGI or Groupe Bull have Tukwila systems or did they drop support?" Don't know and tbh, don't really care! I only want hp's Itanium kit unless Unisys comes up with something interesting. As it stands, all three mentioned are still signed up to the Itanium Alliance so they would still seem interested. How many companies will be making Power7 servers this year? One. And no dates yet on which Power7 servers when, but probably initially only the low-end ones where IBM can get away with putting in Power7 chips with duff cores that failed the process. So, no real octo-core Power7 with all cores working until when, October?

"2) Why only four cores when Nehalem and POWER7 are both 8 cores." Because the cores are better, maybe? After all, dual-core Itanium had no problem competing with quad-core Power in the marketplace, so maybe the overall package of Itanium, hp-ux (and/or Windows, OpenVMS, NonStop, Linux, OpenVMS) and hp support is just more appealling than IBM's offering. We certainly find it compelling, though we do buy Power when it is the better option.

"3) Why less than 2GHz when Nehalem is 3+ and POWER7 is 4+" Oh dear, I can't believe you're using the old frequency angle! So, how did that not work when Intel used to compare faster Xeon chips to Power5? Oh, I remember now - IBM claimed that the Power did more with each clock cycle than the Xeon. So let's just say Tukzilla will also "do more", just as the current range seem to be doing compared to Power6/6+. Besides, you do remember that the CPU clock is just one factor on performance, and that keeping a nice fat pipe to the CPU is much more important. Ah, now I see how the current dual-core Itaniums can beat Power6/6+.

"4) Why announce a chip and not announce actual systems, what is the point?" Intel make the CPU, it is the vendors such as hp that make the servers. Duh!

"5) Wasn't tukwila announced two years ago?" Not sure on the exact date but probably. However, it was decided to include DDR3 in the spec, which meant the memory controllers had to be redesigned. The rest of the design, though tweaked, would seem to have been in place for a while, which means there are likely to be better yields and less bugs than IBM's rushed Power7, which IBM have rushed out the door to meet the Tukzilla launch date.

"6) Why is HP requiring a fork lift upgrade to Tukwila systems" Because the current Integrity range has offered in-box upgrades since 2001. It's about time for a change. Sorry, but how many Power upgrades have resulted in anything other than a box-swap? Face it, hp have done much better job of designing ahead and making their servers good for several CPU upgrades, whereas IBM haven't. It seems to be a common failing amongst IBM server products. You wouldn't want me to mention the fun IBM had with blade system power bricks changing every time there was a Xeon upgrade, about half-a-dozen times in the space of two years, do you? Hopefully, IBM can take a look at the new hp Integrity servers and maybe learn something about proper design. :P

"7) Why is Tukwila finally getting to 65nm when all other chips are 45nm or better?" Agreed, the 65nm bit is a problem, but not a big one. The design has been held back to add the new memory controllers for DDR3. As it was originally designed at 65nm, it made sense to take the least disruptive path and stay there. The problem for IBM is that Intel do this a lot - get a chip working at one size die, then rapidly shrink the die to reduce power draw and ramp both performance and density. Just look at the Xeon roadmap. Intel has a 45nm process already and soon a proven 25nm process, which bodes well for Itanium's future. And, to go back to your question about octo-cores, when Itanium hits 32nm or 25nm it can use the proven Nehalem processes to go hex- and octo-core. And what do we have on the Power roadmap after 45nm Power7? Nothing but "future" - no indication of a 25nm Power8.

"8) Why is there no core performance increase when software costs are so high and priced per core?" You want to complain about increased core counts increasing licence costs when Power7 is going octo-core!?!? So Tukzilla's clock is going up but you claim there will be no per core gain? What, are you saying clock makes no difference, which kinda blows a big hole in your argument above? So, the core will be faster, and it will have greater bandwidth to the faster DDR3 memory, wanna hand with the obvious conclusion? It will perform better per core than the current Itaniums with the same apps. And, as the servers can cram more cores per socket, the customers can use smaller boxes to replace current boxes, saving on rackspace and power. So, faster, smaller, cheaper to run - you getting there yet?

"9) How are the Linux sales on Itanium? I hear they are so bad RedHat has dropped support." No idea. But we do have a commitment from RedHat to continue current support of RHEL5 on Itanium. What will harm RedHat's chances on Itanium is they have decided not to support KVM (their virtualisation tool) on Itanium. But, we still have Integrity Virtual Machines, or Xen for if we need another way to virtualise Linux on Integrity. The fun bit will be if Larry releases Oracle Linux on Tukzilla.

"10) Will Tukwila finally have hardware virtualization? Itanium seems to be the only processor besides SPARC64 to not provide at least some hardware virtualization." Bit confused hear as Integrity has had a better partitionig options than Power since 2001. With Integrity Virtual Machines, soft partitioning and hard partitioning, PRM, WLM and GWLM, all from hp and fully supported, what is the need for "hardware" virtualisation? Cough*features-sell*cough.

"11) I hear Oracle is going to increase the core price to .75 just like SPARC or higher what is the factor?" I hear Jordan is going to get bigger boobs. Unless it happens it's just talk, or FUD in your case. And seeing as Larry has been very carefully targetting IBM in his attacks, not hp (seeing as the majority of his new Oracle installs go onto hp servers), it is much more likely he will be jacking up licence costs on Power rather than annoying hp.

"12) Is HP going to have a BI offering now that Exadata was canceled and Neoview is not selling?" I have no idea on Neoview sales but I bet they're at least as good (or as bad) as Exadata. And didn't hp and Microsoft recently announce a joint project on BI and CRM? As I predicted they would, should Larry get too jumped up (as he has).

"13) Besides having twice the cores in the same footprint what is the value of Tukwila? The same core and software license performance." Dear, you're repeating yourself. Please see the answers to point 8 above.

"14) Oracle still does not have support for the eBiz suite applications on Itanium..will that ever change?" No idea, ask Larry. But isn't the number one platform for new eBiz installs (along with ALL Oracle software offerings) hp, namely ProLiant? If ProLiant is the top seller for eBiz then I can't see hp being too worried, especially as I can't recall ever hearing of an eBiz install on Power6.

"15) I hear HP is giving away free upgrades to Tukwila, but I don't see the point since applications will not get any more performance." LOL! "I hear a competitior is giving away free upgrades, which means customers can cash in old servers and get new ones which consume less power and need less rackspace AND perform better" - yeah, I can see why that would be a problem for you!

/SP&L, new object of amusement though.

0
1
Go

@Matt

Welcome back, you've been missed.

"isn't the number one platform for new eBiz installs (along with ALL Oracle software offerings) hp, namely ProLiant? If ProLiant is the top seller for eBiz then I can't see hp being too worried"

Indeed.

So when are HP going to stop wasting customers money on IA64? When are they going to rationalise the remaining IA64-specific stuff over to AMD64-based Proliants? Presumably HP are picking up the full costs of IA64 now as no other vendor really cares.

Meanwhile, Proliant hardware in various shapes and sizes (blade, classic rack, whatever) is already capable of doing pretty much anything you might want from an IA64 box, with the exception of a tiny quantity of huge-memory massive-SMP systems. And as you rightly point out, Proliant has the huge advantage of massive market acceptance. You might sell a Proliant (with some added software uniqueness) into a Dell shop, you'd get nowhere with IA64.

There is one little snag though; HP haven't yet ported their relevant OSes (HP-UX, NonStop, VMS) to AMD64. But the money HP would save on chip+system development specific to IA64 (by sharing costs with the AMD64 world) should more than cover the cost of porting the OSes and layered products, and then there'd be a much bigger addressable market for NonStop and VMS.

Summary: People don't buy IA64 systems because of IA64, they buy them despite IA64, they buy because of the features they get from the OS (features which they will also get on a decent AMD64 box, once HP and Intel can find a way of abandoning IA64 without looking too embarrassed).

0
0

Nehalem and Opteron != Itanium and Power7

Itanium and Power7 are RISC processors. Higher clock speeds are not needed.

Opteron 12 core chips are a game changer and and shame on Intel for not being able to do dual 8 core Nehalem, but its probably a controller problem and not a limitation of the chip.

A 24 core CISC machine will those RISC machines a run for their money. I wouldn't be rushing out to purchase the RISC machine.

Oracle??? who can afford Oracle?

0
1

RE: Allison Park & AC

Copy that down from an IBM sales battlecard then? Very insightful, seriously though is that the best they can come up with? I know HP have a similar thing "10 things to ask IBM about Power",

PS Oracle E-Business 12 is supported on Itanium with HP-UX 11i V3 has been for about 8-9 months - I know cause I've just done a project with it.

However future Linux support is a bit of a kick in the nuts, in the same project we migrated an Oracle instance on IA-64 Linux to HP-UX so we could deliver HA to the applications (as HP themselves have now dropped Serviceguard Clustering on Linux) - being objective at the minute I wouldn't suggest that a customer deploys Linux on Integrity for a new project at the current time.

AC - Do HP really waste customers money on IA-64, are you really comparing apples with apples - does that mean IBM are also wasting customers money with Power (when they have X-Series) and SUN with Sparc (when they have Galaxy) etc etc

simple fact is you can't scale high enough for things like large SAP instances (which aren't that uncommon) on a Proliant, you can scale out comfortably but thats not always the best thing to do as you end up with more OS's to manage and a higher operational costs. Funnily enough a lot of very well known online gaming websites run their backends on large SQL databases on Itanium for this very reason.

0
0
Happy

Tukzila is no new news

"1) all three mentioned are still signed up to the Itanium Alliance so they would still seem interested.

==> SGI went bankrupt after Rick Belluzzo came over from HP and forced them to move from MIPs to Itanium. Now Rackable bought SGI and is only interested in the current Itanium chip.

===> Unisys is not interesting in supporting Tukwila as the volumes are too low to justify R&D or even sales.

How many companies will be making Power7 servers this year?

===>Groupe Bull and Hitachi

One. And no dates yet on which Power7 servers when, but probably initially only the low-end ones where IBM can get away with putting in Power7 chips with duff cores that failed the process. So, no real octo-core Power7 with all cores working until when, October?

====> Will you admit you are wrong on Monday?

"2) Why only four cores when Nehalem and POWER7 are both 8 cores." Because the cores are better, maybe?

==>Because 65nm does not allow enough space four 4 cores

"3) Why less than 2GHz when Nehalem is 3+ and POWER7 is 4+" Oh dear, I can't believe you're using the old frequency angle!

===> Well no one is saying EPIC architecture is the future and everything I have seen says Nehalem will replace Itanium

"4) Why announce a chip and not announce actual systems, what is the point?" Intel make the CPU, it is the vendors such as hp that make the servers. Duh!

===> Not sure why its meaningful for customers and I doubt there will be any new news. The chip was supposed to be out in 2006 to compete with Power6.

"5) Wasn't tukwila announced two years ago?" Not sure on the exact date but probably.

==> yes announced as the first 2 billion resistor chip

However, it was decided to include DDR3 in the spec, which meant the memory controllers had to be redesigned.

==>There had to be more than DDR3 otherwise why not do a Tukwila+

The rest of the design, though tweaked, would seem to have been in place for a while, which means there are likely to be better yields and less bugs than IBM's rushed Power7, which IBM have rushed out the door to meet the Tukzilla launch date.

==> Why would Power7 be rushed to meet a Tukzilla launch date when it was designed to compete with Power6 and the Tukzilla systems wont be available till May? Maybe Power7 had such good yields the announce was moved in three months...that does happen for some chips.

"6) Why is HP requiring a fork lift upgrade to Tukwila systems"

====> Well the HP smooth upgrade path (you only get to keep the rack) will not be true any more

"7) Why is Tukwila finally getting to 65nm when all other chips are 45nm or better?" Agreed, the 65nm bit is a problem, but not a big one. The design has been held back to add the new memory controllers for DDR3. As it was originally designed at 65nm, it made sense to take the least disruptive path and stay there. The problem for IBM is that Intel do this a lot - get a chip working at one size die, then rapidly shrink the die to reduce power draw and ramp both performance and density. Just look at the Xeon roadmap. Intel has a 45nm process already and soon a proven 25nm process, which bodes well for Itanium's future. And, to go back to your question about octo-cores, when Itanium hits 32nm or 25nm it can use the proven Nehalem processes to go hex- and octo-core. And what do we have on the Power roadmap after 45nm Power7? Nothing but "future" - no indication of a 25nm Power8.

====> Let's not confuse Xeon with Itanium. Tukwila is 65nm and Poulson while Intel says it plans for it to be 32nm does not have a date....and that is tock of Tukwilas tick

"8) Why is there no core performance increase when software costs are so high and priced per core?" You want to complain about increased core counts increasing licence costs when Power7 is going octo-core!?!? So Tukzilla's clock is going up but you claim there will be no per core gain? What, are you saying clock makes no difference, which kinda blows a big hole in your argument above? So, the core will be faster, and it will have greater bandwidth to the faster DDR3 memory, wanna hand with the obvious conclusion? It will perform better per core than the current Itaniums with the same apps.

====> Everything I read about Tukwila says "twice the performance" nothing says "more than twice the performance or four times the performance" and since it has twice the cores then the core performance is the same.

"9) How are the Linux sales on Itanium? I hear they are so bad RedHat has dropped support." No idea. But we do have a commitment from RedHat to continue current support of RHEL5 on Itanium. What will harm RedHat's chances on Itanium is they have decided not to support KVM (their virtualisation tool) on Itanium. But, we still have Integrity Virtual Machines, or Xen for if we need another way to virtualise Linux on Integrity.

====> So unless HP buys RedHat RHEL6 will not be supported on Itanium. So it is just HP-UX, VMS, Non-Stop and some Windows.

"10) Will Tukwila finally have hardware virtualization? Itanium seems to be the only processor besides SPARC64 to not provide at least some hardware virtualization." Bit confused hear as Integrity has had a better partitionig options than Power since 2001. With Integrity Virtual Machines, soft partitioning and hard partitioning, PRM, WLM and GWLM, all from hp and fully supported, what is the need for "hardware" virtualisation? Cough*features-sell*cough.

=======> Partitioning is not virtualization. IVM is a high overhead HP-UX host and is more for test and development. I figured Tukwila would have hardware virtualization like Power or Xeon...it it doesn't then Itanium has a real problem.

"11) I hear Oracle is going to increase the core price to .75 just like SPARC or higher what is the factor?" I hear Jordan is going to get bigger boobs. Unless it happens it's just talk, or FUD in your case. And seeing as Larry has been very carefully targetting IBM in his attacks, not hp (seeing as the majority of his new Oracle installs go onto hp servers),

====> As pointed out by Pony Tail, Oracle's core factor table has Tukwila at 1 license per core which is up from .5. Who is Jordan?

"12) Is HP going to have a BI offering now that Exadata was canceled and Neoview is not selling?" I have no idea on Neoview sales but I bet they're at least as good (or as bad) as Exadata. And didn't hp and Microsoft recently announce a joint project on BI and CRM?

=====> Never heard of SQL Server being used for BI

"13) Besides having twice the cores in the same footprint what is the value of Tukwila? The same core and software license performance." Dear, you're repeating yourself. Please see the answers to point 8 above.

====> I guess I am just shocked at how poor Itanium has done so poorly vs. the expectations.

"With twice as many cores, Tukwila will allow HP to put the same compute capacity into a system half the size, Bartlett said. For example, a fully loaded SuperDome with 64 processors will occupy one server cabinet with Tukwila, whereas today it uses two." VP HP.

===> I hope they come out with a better value proposition than just two to four cores.

"14) Oracle still does not have support for the eBiz suite applications on Itanium..will that ever change?" No idea, ask Larry. But isn't the number one platform for new eBiz installs (along with ALL Oracle software offerings) hp, namely ProLiant? If ProLiant is the top seller for eBiz then I can't see hp being too worried

====> Confusing Xeon and Itanium again. eBiz will never be supported on Itanium and is fully supported on all Power systems.

"15) I hear HP is giving away free upgrades to Tukwila, but I don't see the point since applications will not get any more performance." LOL! "I hear a competitior is giving away free upgrades, which means customers can cash in old servers and get new ones which consume less power and need less rackspace AND perform better" - yeah, I can see why that would be a problem for you!

====> Annyong-hi kashipshio....or for you Cheers.....and see you Monday

/SP&L, new object of amusement though.

1
0

Only twice the performance with double the cores?

It sounds to me that all the new optimizations, i.e. DDR3, 30MB cache are there purely to help it scale to twice the performance for twice the cores, I wonder if having an ondie Memory controller (aka Nahalem, Opteron) might have been a better idea. It doesn't seem like a big enough step in the right direction given all of the delays.

I would have liked to have seen higher clocks as well as single threaded performance will still be the same and Itainium in recent years has dragged behind x86 in a performance per core basis

Sun was very much on the right track in porting Solaris to x86 alas it didn't get the required tracktion. Now HP-UX on x86-64 would be a very nice proposition indeed.

0
0

Hardware virtualisation vs partitioning

Allison,

Can you enlighten us as to what your definition of Hardware virtualisation is on a CPU and why not having it on a chip is a disadvantage. What is the key benefit and ROI that a user will get

"Partitioning is not virtualization. IVM is a high overhead HP-UX host and is more for test and development. I figured Tukwila would have hardware virtualization like Power or Xeon...it it doesn't then Itanium has a real problem."

Just curious to know

Thanks

0
0
Happy

"you can't scale high enough ... on Proliant"

I'll happily agree that there is an application niche where the need for single system image massive memory massive SMP currently rules out Proliant. So when today's Proliant DL785 G6 (?) isn't big enough (8 sockets/48 processors, 512GB of memory) isn't enough hardware, there's still IA64 (and soon there'll be the next Proliant, which will hopefully be on time, on budget, working right, and affordable, as they generally have been, unlike IA64 and IA64-based systems).

I'm sure there are at least a couple of dozen customers where 48 processors and 512GB of memroy isnt enough, and you kindly name a few examples.

But if the cost of chip and system development is only spread between those few dozen customers because the vast majority are buying "industry standard" product from the same company, the niche customers see prices going up sooner or later, and the shareholders see money being burned that could get a better return elsewhere. Neither IBM nor Oracle/Sunhave exactly that problem.

Most IA64 customers care more about using the OS they know and love (be it HP/UX, NonStop, or VMS) and they see no advantage in IA64, in fact it's a hindrance rather than an advantage.

But if your employers want you to believe otherwise, so be it.

1
0
Bronze badge

Vendor-Speak...

There is a lot of supposition going on here. I can help a little on the Power 7 stuff:

- IBM had advised that Power 7 servers were a March intro for several months. Moving the announcement up to February is a surprise.

- The allegation the IBM will bring out small machines with defective cores FIRST does not match with IBM open announcements. IBM set expectations that mid-high servers would come earliest.

- IBM's 4 GHz RISC core will well-outperform INTC's 2 GHz RISC core. Since IBM packs 8 on a chip, it would be reasonable to expect that 8 *4GHz is four times the throughput of 2*2GHz from an Itanium chip. We will see.

- IBM is announcing SYSTEMS and INTC is announcing CHIPs. There isn't much doubt about that. Itanium built-out Systems should be announced quickly after INTC's announcement, though.

- 32nm or 25nm Itaniums have not yet been announced by INTC. There are IBM and INTC announcements for 32nm chips. INTC will be first (6-15 months). IBM will be later (24-36 months). This is only relevant if it affects high-core offerings or high-GHz offerings that deliver important performance advantages.

- Itanium hardware virtualization hasn't appreciably improved in over six years. There needs to be INTC guidance for that. IBM and INTC 32-bit hardware virtualization has increased significantly/dramatically in the past five years.

- Oracle matters. We will have to see what they do. They are 'playing defence' with all the Sun stuff that they bought, though.

- Even Itanium fans are on edge with this delayed intro. Itanium needs a smooth launch to have any chance at a further 5 year's existence in the CPU-wars.

1
0
kit

Additional comparisons

The speed of the chip i.e. 1.6+ vs. 4+ G is not relevant if the jobs they handle are parallel in nature. However for serial computation , higher the speed ,greater the advantages.

What makes Power7 out run Itanium, are the design and technologies which IBM has used, namely

SOI to reduce current leakage

eDram to reduce number of transistor count by 2 thirds w.r.t. Sram, It would mean Power7 consuming less electricity even with a bigger L3 cache (32M) as compares to (30M) Tukwila.(transistor count of the chips are 1.2B to Power7 as to 2B to Tukwila). In theory Power7 can handle bigger jobs than Tukwila and faster because of higher CPU speed and bigger cache, while spending less on the electricity bills.

Oct cores , 4 threads/core of Power7 vs. Quad cores, 2 treads/core of Tukwila

32 treads/Power7 : 8 threads/Tukwila,

It is significant to their chip virtulisation performances. More threads the better.

0
0
Unhappy

Tukwila does have two on-die memory controllers

I don't get why the cores do not have any higher performance than Montvale either.

We tested some Montvale SX2000 systems vs. Power6 and found the IBM gear to be 2.8X better performance per core. Even with Oracle pricing the IBM gear was a better price and lower cost.

Not sure if there will be any benchmarks released on Monday but core performance is important to us not how many chips happen to be in the system which has the cores.

We also tried Integrity Virtual Machine in the lab but the eight thread limit eliminated it for any of our databases or applications. The overhead was about 35% also since there is not hardware virtualization in the chip. Is this going to be fixed with Tukwila? We still have a lot of HP-UX systems which are going somewhere now that HP stopped selling PA-RISC.

If the Intel "announce" on Monday does not have any systems or benchmarks than they should be chastised as fakers. "Volume shipements" to HP without systems does not an announcment make.

0
0
Silver badge
Troll

RE: Tukzila is no new news

Well, if Tukzilla is "no new news", maybe that's because Intel and hp have been keeping the market well-informed of what was coming, whereas with IBM it's all up-in-the-air guesswork. But let's have fun with the rest of your ramblings:

"...SGI....bla,bla,bla....Unisys....bla,bla,bla...." Erm, so what? I don't have any of either of those vendors' systems. I don't have any Bull boxes either. I doubt if many hp customers do, so why should hp or their customers worry? The reason other vendors (including IBM) haven't been as good at selling Itanium servers is because hp had such a massive advantage, having been in from the base design, and simply got first to market with better designs. And the sales of hp Integrity alone seem to be making enough money to keep both hp and Intel happy.

".....Partitioning is not virtualization....." No it's not, but then both are simply ways to run more than one OS image on the same server, usually for consolidation projects or maximising utilisation, and seeing as both attain the same goal I'd say they both count. Or do you have some funky, one-in-a-billion case where it just has to be hardware virtualisation or nothing? Maybe in the IBM salesguide? Probably not.

"....IVM is a high overhead HP-UX host...." All virtualisation has overheads, honey, even hardware virtualisation. It's kind of a law of reality that you can't get anything for nothing. Otherwise it wouldn't be virtualisation, it would be magic! Please do not try and tell us that hardware virtualisation has zero overheads as such amusing statements will only get you the moniker of Mrs Kebabfart.

".....and is more for test and development....." Sorry to disappoint, but we have it in production, with VMs in clusters, running biz crit apps. The only time I have seen a similar level of belief in any IBM virtualisation product it was on mainframes, not pSeries.

".....Because 65nm does not allow enough space four 4 cores...." Maybe, but then hp has a history of innovating where others said it couldn't be done. Remember when IBM had dual-core Power5, and was shouting to all that Intel didn't have a dual-core Itanium, and that it was too big a die to make a dual-core with? I remember an IBM salesgrunt saying IBM could put twice as many cores per socket and there was "no way hp can do that with Itanium". Well, hp was quite happy to prove him wrong when they cobbled two Itanium chips onto a card that plugged into the single socket, effectively doubling the core count. It was called mx2, and it really put a hole in the IBM salepitch of the day. Agreed, it wasn't dual-core, but for those customers with CPU-intensive apps it got them to the same place.

"....Let's not confuse Xeon with Itanium....." Ah, but then you don't want to admit the two are linked. For years, Intel have been taking RAS features from Itanium and putting them into Xeon, and the economies of scale of Xeon have allowed Intel to fund Itanium and cut the costs of Itanium systems. IBM doesn't have a similar partner CPU for Power with the same dynamics of scale (the Cell products are completely different and shares virtually nothing with Power products). In the same way, hp took RAS features from Integrity and fed them into the ProLiant range, making the ProLiant the leading x64 range. At the same time, the cross-pollination allowed hp to reduce the cost of each Integrity server as more and more parts became common to both. With QPI, the motherboards for Xeon and Itanium now share many components, further reducing the cost for Itanium. IBM just doesn't have the same going for Power. Which is why the new Integrity range will still be cheaper than IBM's new Power range. And then you have to think that hp actually sell ProLiant and Integrity in tandem and are happy to make a mixed solution, whereas IBM will try to sell you a mainframe first, if that fails then they push a Power server with AIX, and if that fails they try Xeon last.

".....Never heard of SQL Server being used for BI...." <Whispers so as not to alert others to Allison's lack of knowledge > SAP BI 7.0. Comes from a rather large and well-known German company. Goes rather nicely with MS SQL, especially on hp ProLiant (SAP's most common platform). Now look like you knew that all along before the other posters notice! ;)

"....eBiz will never be supported on Itanium and is fully supported on all Power systems...." Considering Adam's statement from above, it would seem that you are far more aware of Power than Itanium. Maybe you should try reading up on your opposition before jumping in with both feet.

/SP&L, yawning a bit though as it's already been a long week.

0
2
Grenade

Re: RE: Tukzila is no new news

I really shouldn't... but Matt makes it too easy. I hat defending others... It seems funny that Matt is using Sun's arguments for CMT now. It's about the system not the core speed! LMOA! Intel admits that the core speed will remain the same and they are just increasing the core count and Matt can defend this? They promised much much more. Come on Matt, even a Fanboy such as you can scream foul.

"Well, if Tukzilla is "no new news", maybe that's because Intel and hp have been keeping the market well-informed of what was coming, whereas with IBM it's all up-in-the-air guesswork."

Over 2 years late, lower MHz, and fewer cores than promised... Hmmm... Who's keeping who up to date?

""...SGI....bla,bla,bla....Unisys....bla,bla,bla...." Erm, so what? I don't have any of either of those vendors' systems. I don't have any Bull boxes either. I doubt if many hp customers do, so why should hp or their customers worry?"

Who cares? Maybe Intel? Intel sells volume. One vendor does not equal volume, or HP would have just developed the chips themselves. Intel cares very much that no one besides HP uses Itanic. That's why HP has to keep pumping bribes into Intel to keep developing this stillborn chip.

""All virtualisation has overheads, honey, even hardware virtualisation. It's kind of a law of reality that you can't get anything for nothing. Otherwise it wouldn't be virtualisation, it would be magic! Please do not try and tell us that hardware virtualisation has zero overheads as such amusing statements will only get you the moniker of Mrs Kebabfart."

The original poster was reaching, really... However, HP does have more overhead in their virtualization than all of the other vendors, and IBM has a LOT of overhead, so that's saying something. Of course, x86 virtualization does not have much to be proud of either. Really,

when it comes down to it, who cares. As long as you can squeeze more out of less, the better.

"Sorry to disappoint, but we have it in production, with VMs in clusters, running biz crit apps. The only time I have seen a similar level of belief in any IBM virtualisation product it was on mainframes, not pSeries."

I almost didn't comment here 'cuz it was such nonsense. Saying that HP's virtualization was trusted more than IBM's? Come now... You're really serious? They're really about the same in my opinion...

"Maybe, but then hp has a history of innovating where others said it couldn't be done. "

HA HA HA HA HA!!! Are you serious? HP hasn't innovated anything since selling off PA-RISC. HPUX has some advantages, but come on now... Your example of HP innovating is an Intel chip being put on a card?

""....Let's not confuse Xeon with Itanium....." Ah, but then you don't want to admit the two are linked. For years, Intel have been taking RAS features from Itanium and putting them into Xeon, and the economies of scale of Xeon have allowed Intel to fund Itanium and cut the costs of Itanium systems."

You insist on saying that Itanic will innovate because Intel innovates with Xeon... They are linked in that Intel wants Xeon to take over the Itanium market. That's it... That's the only way that they are linked. When Intel drops Itanic and HP is forced to move their customers once again, don't say that we all didn't say we told you so. Even SPARC has a bigger market than Itanic. Intel does not have as much patience as you think.

"Goes rather nicely with MS SQL, especially on hp ProLiant (SAP's most common platform)."

Who runs SAP on MS SQL? Now you're just being silly. Note to Matt - SAP is for running real companies, not the company you seem to want us to believe you work for. Once again, you're saying that Itanium is great because Proliant is successful. Interesting how your mind works.

HEY MATT!!! This subject is about ITANIC! Stop changing the subject!

""....eBiz will never be supported on Itanium and is fully supported on all Power systems...." Considering Adam's statement from above, it would seem that you are far more aware of Power than Itanium. Maybe you should try reading up on your opposition before jumping in with both feet."

Avoiding the subject are we Matt? Hmmm... Really I don't think it matters if eBiz runs on Itanic or not, but your avoidance of the subject speaks volumes. Itanic has fewer and fewer apps. Note to Matt - Applications matter. If you don't run 'em, then no one will run on your hardware. HP and Intel will have to pump more money than they are now into vendors to keep developing for their platform.

Moving forward there will be three platforms - X64 (Intel volume) - Power (IBM) - SPARC (Oracle)

Of course, SPARC definitely has less legs than x64 and Power, but they all have more chance than Itanic.

1
0
Welcome

welcome back

"hp took RAS features from Integrity and fed them into the ProLiant range"

Examples? Preferably ones that weren't already in AMD64/Opteron (they may well not have been in Xeon pre its transformation into an AMD64 clone, but those don't count, as the inspiration came from AMD64/Opteron, not from IA64).

"making the ProLiant the leading x64 range"

Proliant has been the industry standard IA32 server since the dinosaur era, and already was AMD64 market leader as a natural consequence of that. Again, what exactly did IA64 contribute to that? The short-lived and largely irrelevant IA64-based Proliant DL590/64 [1] didn't bring much to the Proliant range, except perhaps embarrassment.

"At the same time, the cross-pollination allowed hp to reduce the cost of each Integrity server as more and more parts became common to both. With QPI, the motherboards for Xeon and Itanium now share many components, further reducing the cost for Itanium."

You need to be a bit careful with that argument. The more components and architectures there are in common between IA64 systems and AMD64 systems, the less uniqueness IA64 has. The less uniqueness IA64 has, the less justification there is for continued spending on separate IA64 chip and system product lines, in what is a distraction to both players' mainstream (AMD64) business.

It doesn't matter anyway, this decision isn't about commercial and technical value for money for customers and shareholders (and even channel partners, employees, etc). It's about the preservation (for now) of certain egos in Intel and HP HQs.

[1] http://h18000.www1.hp.com/products/quickspecs/10907_div/10907_div.html

0
0
Stop

Oracle eBiz suite APPLICATIONS are not supported on Itanium

The database can be on Itanium just like any other Oracle database but the application servers will never be supported on Itanium. You are forced into a linux/x86 and HP-UX/Itanium

0
0

Random oberservations

#1 SGI isn't famed for it's decision making - remember they bought Cray Research and immediately sold off Cray Superservers to Sun Micro just as project Starfire was about to be launched ... thus handing the E10K directly to a competitor.

#2 HP still sell PA-RISC and Alpha too ... through the Renew programme - OK so they aren't building new systems but that's not the point.

#3 Most customers keep running UX, Solars or AIX for one of two reasons - they need the scalability or they need the stability of the current platform and the costs of migration of a critical application outweigh the savings.

Only time will tell if IA64, Power or Sparc have a long term future. I hope so because the IT world will be a boring place if all we have is x86.

0
0

E-Business R12 on Itanium

Just double checked - been available since August 2007 - on HP-UX Itanium

Or are you saying that because R12 is not the legacy Oracle E-Business Suite (that Oracle are discontunuing) it doesn't count???

PS this was from a presentation at an Oracle User Group forum by

Kelton Keller – Senior Technical ConsultantSolution Alliances Engineering, Global Oracle Alliance

FUD disproven.....

0
0

AC - SAP and SQL on x64/86

I think you'll find there are a shed load of organisations that run SAP on Windows with a SQL Backend out there, quite a lot of it in UK local government as one example,

0
0
Silver badge
Happy

RE: Re: RE: Tukzila is no new news

"I really shouldn't....." Going on previous Sunshiner efforts, you probably shouldn't for your own good, but then that would rob us of a good laugh or two.

"....It seems funny that Matt is using Sun's arguments for CMT now...." Yes, you were right - you really shouldn't. CMT is Sun's acknowledgement that Sun couldn't design a server where they could get the bandwidth wide enough and keep it flowing well enough to even feed the feeble UltraSPANKed chips. CMT is a design that doesn't try but simply accepts failure, instead having lots of weiner cores and stop-start-stop failure happening at once in the hope that the average throughput would somehow match a proper chip design. It didn't. Niagara is pathetic outside the webserving niche and is living on borrowed time until Nehalem EX arrives, at which point it will die. Rock, supposedly the real enterprise deal, has already died because the technologies Sun tried to use to keep the cores spinning didn't work. Itanium, especially in Tukzilla, is the complete opposite idea - large, powerful cores with massive bandwidth. That is why Intel delayed it for DDR3, because Tukzilla's large and powerful cores benefit from even faster memory. Isn't it about time you Sunshiners just admitted you got it very wrong and you should all go try something else, as computing really doesn't suit you?

"....Over 2 years late, lower MHz, and fewer cores than promised..." Still a lot more cores and MHz than Rock!

"....Intel sells volume...." And currently Itanium is outselling SPARC by a mile, so there's plenty of volume there. Especially as that volume is also large SMP systems, not the one-core weiner systems where Niagara is selling (I'm told it was an 8:1 ratio of single-core to multi-core sales for Sun Niagara, and even worse when comparing all Niagara to Fujitsu SPARC64 single-cored M3000 sales).

"....HP hasn't innovated anything since selling off PA-RISC...." <Yawn> and more Sunshiner denial. Apart from innovations in areas such as servers and storage, hp has managed great advances in an area completely barren to Sun for many years - profits! Oh, and I think it's time to remind you of that hp printer biz you Sunshiners despise so much, mainly because it shows that hp really did innovate and diversified successfully into new areas, whilst Sun just curled up and died.

"....Your example of HP innovating is an Intel chip being put on a card?...." Yes, a solution that no-one else in the industry could do. Care to tell me what great innovations Sun made with Xeon after Ponytail admitted defeat and started making Xeon servers? Oh, I remember - they went and bought designs from a third party. Real innovation that is - not! Want to talk about Sun's many and hilariously poor attempts to break into the blades market? Surely such a great "innovator" as Sun could make so much better Xeon blades than anyone else that they'd be top of the blades market buy now? The reality is Sun's blades stunk almost as bad as Niagara, and they limp along far behind even Dell in the blades market. Now, do I have to remind you which two vendors are fighting it out to be top of the blades market? Yes, that's right, IBM and hp, because they innovate real advances in their designs. Even when Intel and AMD give Sun a working chips, Sun couldn't design the servers around them good enough to compete. No wonder Sun stuck with SPARC for so long, it was just a poor attempt to hide their failings by avoiding the level playing field of x64.

"...They are linked in that Intel wants Xeon to take over the Itanium market...." Earth calling planet Loon! If you hadn't noticed, and it probably is very hard to see with that Sunshiner Blindfold (TM) on, Xeon has been eating into the UNIX market for years. The whole enterprise UNIX market is shrinking, no-one denies that, and the reason is cheaper x64 systems running Windows and Linux (and BSD) getting more and more capable, and so taking roles away from low-end UNIX boxes. The problem for Sun was it's installed base was mainly low-end, and IBM and hp dominated the high-end, and hp and IBM did Windows and Linux. Seeing as Sun didn't do x64 (until far too late), they were the ones with most to lose. Yes, Xeon is eating into Itaniums market, but it has already eaten SPARC, and will continue eating into Power. As long as Itanium still has an scale-up SMP advantage over Xeon then Intel will continue to make it as they want a slice of the high-end enterprise pie.

"....Who runs SAP on MS SQL?..." <Sigh> I suppose it is a bit much to expect a Sunshiner to have any experience of anything outside the Slowaris on SPARC world. But then I suspect that will be changing soon! I hope you're still young enough to learn a lot of new stuff, because all that old SPARC and Slowaris is going to be gone inside ten years. I suggest you start with the SAP website, then try the Microsoft website. I'm sure you'll find a whole new World of information you never knew existed!

"....Avoiding the subject are we Matt?...." How is it avoided? Adam answered it quite succinctly, surely even clearly enough for a Sunblind drooller like yourself to understand? Maybe you'd better do some reading and comprehension lessons before you try reading the SAP and MS webbies.

".....Itanic has fewer and fewer apps....." A truly bizarre statement - how can the number of apps be getting fewer? All the evidence - and hp has plenty on their website - shows the number of apps growing. What evidence do you have, other than vague FUD from a Sun salesguide? None.

"...Note to Matt - Applications matter...." Note to moron - hp has all the common UNIX business apps already, especially those from Oracle, SAP, and Microsoft. And hp has been selling a lot more hp-ux servers running mission critical apps in the high-end than Sun SPARC for years. Some even with IBM software. Yes, DB2 even runs on Itanium - when was the last time you saw production DB2 on a Niagara server? I hear there was some effort about five years ago by IBM to get it working, but the market response was so tepid that IBM gave up.

Come on, just admit it - Snoreacle will soon be using only Intel and AMD chips. Sun lost, SPARC is dead, and the future of enterprise UNIX is a two horse race - Power and Itanium. Your team took the early bath because your product wasn't good enough, as demonstrated by the long slide in SPARC sales. How long Fujitsu can hold on in the face of their dwindling marketshare in a shrinking UNIX market is anyone's guess. IBM has a future in the enterprise with Power, and so does hp with Itanium, but Snoreacle will only have one putting Oracle apps on IBM pSeries or hp Integrity. Whilst you love to brag about the large installed base, that's all old SPARC, and it's all going to be replaced by non-Sun products. Bye-bye, TTFN, don't let the door hit you on the arse on the way out!

/SP&L.

0
3
FAIL

Tukwila is so last decade

a 2003 core produced with 2006 fabrication technology

0
0
Megaphone

@Re Re Re Re Re Tukwila (Matt 5 Feb 10:21)

Matt is quite possibly right that SPARC isn't relevant as an IA64 competitor; the real competition is elsewhere.

Matt hasn't yet explained how IA64 hardware manages to compete against AMD64 outside the current massive-SMP massive-memory niche which AMD64 soon will grow into.

E.g. Matt's 3 Feb 21:18 post refers to IA64-inspired RAS features in modern Xeon. Matt was asked for examples. This RAS claim is one that HP field folks have been told to make over a number of years and is one they have (afaik) never managed to substantiate when challenged; Matt hasn't yet done so either.

Matt also makes a related claim that there's increasing commonality between Xeon and IA64. Now this is a claim I'm more comfortable with (Quickpath etc) but what Matt doesn't do is show how that can be anything but BAD news for the lower-volume higher-priced decreasingly-unique (?) IA64 (chips and systems).

Can our Matt do any better this time?

0
0
Anonymous Coward

The details are out

Anyone still reading?

No sign of Matt yet, so let's have a look at some of the RAS claims in the "Technical White Paper" at

http://h20341.www2.hp.com/integrity/downloads/21568_Intel_Tukwila_Tech_WP_r06.pdf

Some of the RAS claims include:

Memory scrubbing (caches and main memory): sorry, been around for years on x86 and elsewhere. I'd quote an early whitepaper but work's Outlook Web Access (where I have a link) is more broken than usual and Google doesn't find the one I want.

Dynamic link rerouting: Hypertransport had that long before Quickpath came out.

Processor onlining/offlining: VMS had that when SMP was introduced (mid 1980s) though (as below) it's not *that* relevant as a RAS feature.

Processor hot plug/hotswap: Proliant had that years ago, though like onlining/offlining it's actually mostly a slideware feature as few (not none) OS/application combinations can sensibly use it.

"Chipkill" equivalents and variants (posher than ECC memory protection): initially from IBM, then in Sun and Proliant and elsewhere. Since the 1990s.

PCI (Express) Hot Plug: Proliant had PCI hot swap ages ago, can't say for PCI Express right now.

Quickpath RAS: Now the same across IA64 and Xeon Quickpath RAS? So where's the IA64 RAS advantage?

Etc.

Actually, that's basically it. Nothing new to see here. Best to just contact your HP rep and ask when they're introducing an AMD64 version of your favourite OS.

0
0
This topic is closed for new posts.