32nm quad core Xeons

[Idontcare]

My "in the other camp" comment was intended to speak to the more vocal AMD fanboys who derided Intel's MCM integration approach as being a fake quad-core while hyping up the technical merits of the then absent-from-the-market monolithic quad-core Barcelona.

I was not aware of any AMD employees making official statements deriding MCM integration in the past and I was not considering that possibility when I made my post. I hope it did not come across as being a "call-out" in any capacity.

I appreciate you weighing in and clarifying what AMD's intended message was regarding the intrinsic disadvantages of Intel's implementation of MCM integration. The lackluster thread scaling of FSB-based hardware in the face of semi and fine grained applications was all the eating of the pudding I needed to convince myself of this:

 

[wwswimming]

Savvy product diversification for an artificially segmented market as a means to enrich the ASP topology and maximize shareholder equity?

I hear it is going to be a paradigm shift with synergistic effects between Intel accounting and Otellini's direct-deposit account.

stop talking like a Vice President. i'm eating breakfast.

 

[exar333]

Intel has ALWAYS been "Consumer unfriendly".

Edit: Just thinking of how "UN-green" Intel products are, because they disable SpeedStep on Celerons. Millions of laptops in use sucking down and wasting more power than they need, shortening battery life and inconveniencing people unneccesarily, just to prop up Intel's profit margins.

AMD really deserves a gold star here, because ALL of their chips, from the top to the bottom of their line-up, have power-saving features.

What a bunch of crap. AMD doesn't have as many product lines as Intel has in the consumer area. They do this to be competitive and sell chips, not to help the consumer. Fanboyism at it's best.

 

[cbn]

L5609 according to the table is 2.86 Ghz but in the text it mentions this CPU actually being 1.86 Ghz without HT.

Hopefully L5609 ends up being a true bargain SKU with the full 12Mb cache.

 

[soccerballtux]

What a bunch of crap. AMD doesn't have as many product lines as Intel has in the consumer area. They do this to be competitive and sell chips, not to help the consumer. Fanboyism at it's best.

Lol yeah this fanboyism is a little extreme Larry

 

[aigomorla]

Intel has ALWAYS been "Consumer unfriendly".

Edit: Just thinking of how "UN-green" Intel products are, because they disable SpeedStep on Celerons. Millions of laptops in use sucking down and wasting more power than they need, shortening battery life and inconveniencing people unneccesarily, just to prop up Intel's profit margins.

AMD really deserves a gold star here, because ALL of their chips, from the top to the bottom of their line-up, have power-saving features.

...

Larry... u really need to take this statement back.

AMD has not even been in the laptop market as long as intel.
AMD also does not have anywhere near the same amount of processors intel has.

Intel would rather offer enterprise first pickings at power saving cpu's (ULV) over the consumer, because enterprise tends to use there machines A LOT LONGER and harder.

But Intel OFFERS it, its just pricey, but they've always offered it.
But you pay for technology. This is common market business.

So... u need to take this statement back.

And my arguement... Intel is good enough that MAC even uses it.
<sarcasm.. please dont flame me on that last statement...>

 

[exar333]

Anyone know when the Xeons will be released? Maybe 2Q or 3Q?

 

[toyota]

Anyone know when the Xeons will be released? Maybe 2Q or 3Q?

Most of these chips are slated for release on March 16, 2010.

http://techpowerup.com/110567/Intel_Readies_13_Westmere-based_32_nm_Xeon_Processors.html

 

[Idontcare]

Wow, they really slipped their node cadence for 32nm all the way out to March!?

Aigo called this one, I was convinced we'd see 32nm debut in Nov just like the prior nodes but he said no way in hell (paraphrasing) despite all the PR released by Intel at the time regarding 32nm ramp and $7B investment, etc.

I tip my hat to Aigo, he nailed this one.

 

[Phynaz]

Wow, they really slipped their node cadence for 32nm all the way out to March!?

The 'Dales are supposed to launch in January.

 

[ilkhan]

Wow, they really slipped their node cadence for 32nm all the way out to March!?

Aigo called this one, I was convinced we'd see 32nm debut in Nov just like the prior nodes but he said no way in hell (paraphrasing) despite all the PR released by Intel at the time regarding 32nm ramp and $7B investment, etc.

I tip my hat to Aigo, he nailed this one.

We'll see westmere in january, so they slipped from....july (?) to january in 4 years? I can live with that.

The s1366 westmere cores wont be out until march, which is less than a year after s1366 bloomfield xeons debuted.

 

[VirtualLarry]

What a bunch of crap. AMD doesn't have as many product lines as Intel has in the consumer area. They do this to be competitive and sell chips, not to help the consumer. Fanboyism at it's best.

No, it's the truth.

How would you feel if the number one car maker made seat belts an optional feature, that required purchasing the deluxe package?

And the number two car maker including seatbelts in ALL of their cars?

That's the way I look at it here - power saving features like SpeedStep and CoolNQuiet/PowerNow are essential, and they should be manditory. I would even prefer that the gov't step in and mandate these features in all CPUs, when technically possible.

 

[aigomorla]

That's the way I look at it here - power saving features like SpeedStep and CoolNQuiet/PowerNow are essential, and they should be manditory. I would even prefer that the gov't step in and mandate these features in all CPUs, when technically possible.

I havent touched those features on my main machine for a VERY VERY LONG time. :\

 

[lopri]

They can make more of a given cpu that has already been designed and taped out (higher volume, leading to lower cost) and intentionally disable a certain percentage of them.

Those practices are usually possible when you're a monopoly. (Intel, Microsoft, etc.) I can't imagine computers, cars, TVs, refrigerators sold with built-in features disabled.

My "in the other camp" comment was intended to speak to the more vocal AMD fanboys who derided Intel's MCM integration approach as being a fake quad-core while hyping up the technical merits of the then absent-from-the-market monolithic quad-core Barcelona.

I don't recall such 'fanboys' especially when Barcelona turned out to be such a dud. I do, however, deride Intel's MCM quads for, combined with low multipliers, the misery they provided me with while overclocking.

P.S. And for the record, I haven't seen AMD fanboys in this forum quite a while. However, I can point to quite a few Intel fanboys without much effort. (and they know who they are. )

 

[exar333]

No, it's the truth.

How would you feel if the number one car maker made seat belts an optional feature, that required purchasing the deluxe package?

And the number two car maker including seatbelts in ALL of their cars?

That's the way I look at it here - power saving features like SpeedStep and CoolNQuiet/PowerNow are essential, and they should be manditory. I would even prefer that the gov't step in and mandate these features in all CPUs, when technically possible.

Are you joking? God I hope so. Why should the government even be involved in a company deciding what products to market?

Why are these features essential? What is next? The government should mandate that one company can't have a process advantage over another and Intel can't release 32nm until AMD does?

Your arguement is rediculous. AMD adds these features because their products are slower, plain and simple. It is a way to get people to buy their chips, nothing else.

 

[exar333]

Those practices are usually possible when you're a monopoly. (Intel, Microsoft, etc.) I can't imagine computers, cars, TVs, refrigerators sold with built-in features disabled.


I don't recall such 'fanboys' especially when Barcelona turned out to be such a dud. I do, however, deride Intel's MCM quads for, combined with low multipliers, the misery they provided me with while overclocking.

P.S. And for the record, I haven't seen AMD fanboys in this forum quite a while. However, I can point to quite a few Intel fanboys without much effort. (and they know who they are. )

You must not know a lot about cars, TVs, or other products. TV companies do this all the time, they often release a "different" model of a TV with stripped down features although they are commonly disabled (ports, etc). This is an industry practice.

You "deride" Intel's MCM quads because you have problems OCing them? Really? Seems a lot of people here didn't have a problem OCing them. BTW, what competition did AMD have to the Q6600 when it was launched? Oh wait, NOTHING.

 

[JFAMD]

No, it's the truth.

How would you feel if the number one car maker made seat belts an optional feature, that required purchasing the deluxe package?

And the number two car maker including seatbelts in ALL of their cars?

That's the way I look at it here - power saving features like SpeedStep and CoolNQuiet/PowerNow are essential, and they should be manditory. I would even prefer that the gov't step in and mandate these features in all CPUs, when technically possible.

Seat belts are a mandatory saftety feature required on all cars - bad analogy.

The government should not be in the business of dictating product features.

I actually prefer to see the mid stack parts being defeatured. As they keep the E series stripped down with lower memory speeds and lower qpi speeds, it artificially limits their ability to compete in the bulk of the market. Only ~5% of the market buys those X series parts with all of the features (I would have to pull the exact numbers from Mercury Research but the last time I looked it was definitely single digit). The mid-band parts are ~80% of the market because server customers want price performance more than raw performance.

 

[21stHermit]

That's the way I look at it here - power saving features like SpeedStep and CoolNQuiet/PowerNow are essential, and they should be manditory. I would even prefer that the gov't step in and mandate these features in all CPUs, when technically possible.

Theirs a difference between fancy sounding trademarked names and real world performance. With modern CPU's having upwards of a half-billion transistors, even modest differences in per transistor power consumption can quickly add up.

Any review I've read comparing CPU power consumption, consistently shows Intel CPU's use less power than comparable AMD CPU's. So while Intel lacks the trademarks, they win at the 120V wall outlet.

Yes theirs a place for government in mandating energy efficiency, but not at this level. Many people leave their PC's on 24/7, let's start there. An easy way to control power consumption, escalating KWH rates, the more you use the more you pay per KWH. Just like $4/gal gasoline.

 

[VirtualLarry]

Seat belts are a mandatory saftety feature required on all cars - bad analogy.

The government should not be in the business of dictating product features.

Perhaps a better analogy is CAFE, Corporate Average Fuel Economy. Why shouldn't CPUs be subject to a similar rule, Corporate Average Power Efficiency? If (hopefully when) this gets implemented, then Intel will be forced to offer some sort of SpeedStep on all of their models.

That's Ok, California will probably ban PC power supplies greater than 600W soon enough anyways.

 

[cbn]

Deleted

 

[JFAMD]

Any review I've read comparing CPU power consumption, consistently shows Intel CPU's use less power than comparable AMD CPU's. So while Intel lacks the trademarks, they win at the 120V wall outlet.

How about a review on Anandtech?

At both idle and full load our six core product (2435) beats the Intel Nehalem.

 

[dmens]

^

too bad the nehalem beat the crap out of istanbul in performance while consuming only 24 watts more. absolute power comparisons are meaningless.

 

[JFAMD]

Absolute power compares are extremely important in a variety of places:

1. Cloud computing
2. Dense data centers
3. Power constrained data centers

I once stood in the data center of one of the largest ISPs in Germany. The room was half full, but the CTO was explaining that they had run out of capacity - power capacity.

Reducing power by 20W per server would have given them 840 additional watts on the same power budget. That is ~3-4 more servers for every rack that they had installed.

Absolute power matters a lot in the data center, far more than absolute performance. I can say that in the past 3 years, outside of the financial sector, I have never had a customer demand performance over power, it is quite the opposite.

 

[Accord99]

Absolute power compares are extremely important in a variety of places:

Only when all other things are equal.

Reducing power by 20W per server would have given them 840 additional watts on the same power budget. That is ~3-4 more servers for every rack that they had installed.

But if the lower power server is slower, they lose all that power savings when they have to buy 6-8 additional servers to make up for the lost throughput.

 

[JFAMD]

Price:
Intel 5570 = $1386
AMD 2435 = $989

Intel is 40&#37; more expensive.

Anandtech has already shown that they consume more power

So, if you look at price/performance/watt, you see the comparison starts to look a lot more even.

The majority of server workloads are not running at 100% utilization. HPC is 80-90% utilization. Virtualization is 50-60% utilization. Typical data center workload is 15-20% utilization.

So, if servers are sitting at low utilization for a large percentage of the day, does it make sense to spend a 40% premium to get more performance for the small areas of peak utilization?

HPC would be one area where you might be tempeted to say yes, but generally speaking, they go for mid-bin parts because they have to deploy hundreds or thousands and the price premium/extra power consumption makes those parts a poor choice (both ours and our competitors).

So, who actually uses those top high performance parts? Typically workstations and reviewers doing benchmarks.

If you want to see how competitive the server market is, look at comparisons of where the bulk of the buying happens, in the middle. Intel's X series is ~5% of their total server shipments. Our SE series is ~1-3%. Comparisons of these products are great fodder for magazines and websites, but they don't really reflect a.) overall competitiveness or b.) what most customers are buying.

If you don't believe me, when you go into work on monday, take a look at your company's AVL for servers. You'll probably see 3-5 configurations that IT is allowed to buy. You probably won't see our SE or Intel's X series on the list.

Economics and power drive the decisions in the data cener, performance does not. There will be specialty bids where performance comes into play, but those are the exception, not the rule.

Chevy sells relatively few corvettes. It's mostly because the overwhelming bulk of car buyers are making decisions based on different factors other than raw performance. For the few that do, there is a shiny corvette waiting on the lot for you.

I don't drive the fastest car around, and I am betting that most here do not either.