Dell has announced Ivy EP workstations

[zaphoyd]

It still seams like a bargain compared to BOXX so perhaps we're way low on Apple's expected margins.

Some things to consider about boxx: it includes a giant liquid cooled case stuffed with extra ports of all sorts including ones that need extra controllers to go beyond chipset limits (more 6gbit SATA). It includes SATA drive bays and the associated hardware, cabling, cooling, and power supply headroom they entail. Boxx includes a 3 year warranty standard with the first year at the premium 24x7 phone, NBD on site support level (this is probably +$500 alone given apple's typically anemic base warranty/support).

The boxx alone with a 1620v2, a 500GB HDD, 8GB ram, and a single 1GB K600 is already $3500. I think it may be BOXX's margins we should be worrying about. A dell precision T3610 with the 1620v2, 8GB ram, 500GB HDD, K600, and 3 years of NBD onsite support is $1339.

 

[Umbongo]

Primarily because it is the stealth offering. Not even Intel is really promoting it.

Intel barely promote the Xeon UP line as is.

First, that $700 is primarily driven by what Intel charges. There isn't much of a way around that. If have a more price sensitive, cost effective solution present it. The price is hinged primarily on having a higher dynamic range 3.3-3.9 than 3.2-3.8.

I know where the price came from, are you incapable of following what people previously say on this forum? The E5-1660 V2 just isn't worth it to anyone but the bone-headed. Of the hundreds of Sandy Bridge-E and EP workstations I've configured, built, ordered and consulted on since it's introduction only twice have people gone with that price point, one i7-3960X and one i7-3970X because they wanted bragging rights. It is a terrible investment.

A $1,000 more is going to filter off a significant number of users. Their budgets just aren't going to go that high. The other factor you are sweeping under the rug is that Apple puts a 30-35% markup on these CPUs. The bigger the price the bigger the bump. 35% on $700 is $245. On $1,700 it is $595. That $350 additional won't knock off as many as the $1,000 additional but it will be some.

Totally disagree. People have paid similar values for that sort of performance for years. Despite what I said about the P2 price point rarely being purchased by my clients, dozens opted to go with single E5-2687W systems because, although not nearly as good value as an E5-1650 or i7-3930K, they were better than a DP solution. The E5-1680 is in response to vendor/customer demand. Intel would rather people used the E5-1680 V2 than E5-2600 V2s as it strengthen's the branding. I merely floated the E5-1680 V2 as a possibility after being informed of why it exists by someone in the know, although I expected that was why.

25% over what? The 1650 v2 ? The core counts aren't the same. The higher percentage increase from core count increase would come from a 4 core jump rather than a 2. If additional cores help the jump by 4 can be worth another $1K. If have a "good enough" number of cores with respect to the workload then $1K cheaper and faster is better.

Yes, an E5-1680 V2 will be capable of 25% performance over an E5-1650 V2 for some multi-threaded processing workflows, certainly wouldn't be my choice for my personal use. Nor would any of the 10-cores in a single CPU system and I'd rather run dual E5-2643 V2s than a 12-core, even with the extra expense. Whether they use the E5-2670 V2 to keep it lower clocked than the 12-core option, E5-2680 V2 as something substantial over a 6-core, or the E5-1680 V2 as that is what Intel will push; it's only for a small subset. We know it is because "no one" was buying DP Mac Pros. Even the 12-core is going to sell low compared to the number of 4 and 6-core systems.

Chuckle. So Intel "top edge additional tax" pricing doesn't work for the 1660 v2 but does work for the 1680 v2?

The E5-1660 V2 is so similar in performance to the E5-1650 V2 that the premium, from Intel and Apple, makes it just not worth consideration. Even if there was no Apple tax it is poor value. The E5-1680 is at least something different with the potential for big gains with the right workload.

The same priced 10 core has better bang-for-buck than they edge priced 1680. Same reasons why the 12 core doesn't make alot sense in dual (or single) set ups. 8 and 12 (single and dual respectively) weren't even on Intel roadmap for v2 about a year ago. Those were Haswell era kinds of target core counts and the current pricing is indicative of that kind of early generation arrival.

Yes it does and if I had to choose something from that pricing area I'd go with the E5-2680 V2. This is Apple though. In an ideal world we wouldn't even be discussing this; they'd all be available. Thankfully CPU ugprades still look possible on the user level..

End of the day, in the real world, an E5-1660 V2 offers no real productivity gain over an E5-1650 V2. An 8, 10 and 12-core, regardless of the price, does and gives you functionality too in with VMs being so useful. Just remains to be seen what Apple choose which could be 8, 10 or neither.

 

[clamnectar]

The 1680 v2 is out in a zone where it doesn't make sense because there is a 2600 v2 offering with approximately the same speed and a higher core count for exactly the same price. The speed difference can't really dig out from the core count deficit unless purely focused value evaluation on single threaded drag racing. The dynamic range is larger than anything else but the throughput is not there at the price point.

So deconstruct60... you laid out your reasoning for why the 1680 won't be used, and now that it's showed up in geekbench I would love to see you make a post saying "I was wrong!"

 

[theSeb]

Nah, I expect him to slice your post into at least 4 and find a way to disagree.

The geekbench score does show a 1680. This could be a good option for me, but it will be between that and the 1650

 

[beaker7]

Nah, I expect

Please provide a logical basis for your expectations.

him

Please provide proof that you know beyond question the gender of this poster.

to slice

technically its referred to as a multi quote

your

In technical internet parlance it's referred to as "the" post as ownership has not been established and independantly verified.

post

See above.

into at least 4

6 and counting so far.

and find a way to disagree

Guaranteed.

 

[theSeb]

Haha.

 

[Tutor]

Nurse! Scapel! Sponge! Suture! Pick me up off the floor!

...

Which doctor(s) from these lists [ http://en.wikipedia.org/wiki/List_of_fictional_doctors ] assisted you in the surgery? Did House play a role?

 

[deconstruct60]

So deconstruct60... you laid out your reasoning for why the 1680 won't be used, and now that it's showed up in geekbench I would love to see you make a post saying "I was wrong!"

If there is a limited number of BTO CPU options then yes it looks like it is in (and higher GHz 10 core options out). It still doesn't make alot of sense from a multicore preformance perspective. The same folks making snarky comments above this post are all over that 1680 Geekbench thread clowning that the Geekbench scores are lower than the 2012 about equal to the 2010 machine offerings. With 10 cores the score would have been higher. So at this point if want to go faster than 2012 offerings you are now stuck picking Apple's $3,000+ CPU option at the very top end.

The win with the E5 1680 is only on single CPU Geekbench scores versus the 2010-2012 options.... which basically doesn't even appear in that forum thread. To be such a highly relevant factor for the targeted consumers it is also particularly odd that no one over there even mentions it.

So sure, cheer and clap for less multicore performance for just as much money as could have bought higher.

Neither the E5 1680 v2 nor the E5 2697 v2 have $/performance that looks good to wider markets. That these two are the only ones that Apple wants to dribble out looks like that Apple may be pricing the whole Mac Pro line up in looney toon land. I didn't think this was going to be a Cube repeat either, but it looks now to be drifting that way.

 

[wallysb01]

If there is a limited number of BTO CPU options then yes it looks like it is in (and higher GHz 10 core options out). It still doesn't make alot of sense from a multicore preformance perspective.

It seems to really make zero sense, the E5-2680V2 is about the same price, only drops .2 off the total GHz, but has 10 cores. And 3*8 < 2.8*10.....

It really seems the 1680 should be priced around $1500 to really make any sense. It what other spot in the intel line up does the 2600 out preform the 1600 in a single socket system at any price point?

I didn't think this was going to be a Cube repeat either, but it looks now to be drifting that way.

I'm not sure how much this will effect the low end, which is what will make or break the new Mac Pro, but the top end was already going to be unattractive without a DP configuration. Without well placed performance/$ choices, it is looking like Apple is doing their best to kill the top end.

 

[clamnectar]

If there is a limited number of BTO CPU options then yes it looks like it is in (and higher GHz 10 core options out). It still doesn't make alot of sense from a multicore preformance perspective. The same folks making snarky comments above this post are all over that 1680 Geekbench thread clowning that the Geekbench scores are lower than the 2012 about equal to the 2010 machine offerings. With 10 cores the score would have been higher. So at this point if want to go faster than 2012 offerings you are now stuck picking Apple's $3,000+ CPU option at the very top end.

The win with the E5 1680 is only on single CPU Geekbench scores versus the 2010-2012 options.... which basically doesn't even appear in that forum thread. To be such a highly relevant factor for the targeted consumers it is also particularly odd that no one over there even mentions it.

So sure, cheer and clap for less multicore performance for just as much money as could have bought higher.

Neither the E5 1680 v2 nor the E5 2697 v2 have $/performance that looks good to wider markets. That these two are the only ones that Apple wants to dribble out looks like that Apple may be pricing the whole Mac Pro line up in looney toon land. I didn't think this was going to be a Cube repeat either, but it looks now to be drifting that way.

Hey, I'm not saying I'm happy with Apple's choice on this, I'm just saying that it would be nice for you to acknowledge your eminent fallibility for a moment before resuming your omniscient pose. But whatevs...

As I wrote somewhere else, the only rational explanation for this CPU choice is that they REALLY care about cache. This processor has substantially more cache per core than any other choice in the 1600 line. Or almost any other CPU, period. That's the only thing that makes it stand out, as far as I can see. I'm having a really hard time coming up with what the other options will be though. There will most likely be a cheaper option, like a 6 core... but which one? And what about a 10 core? I'm a bit puzzled at this point.

----------

It seems to really make zero sense, the E5-2680V2 is about the same price, only drops .2 off the total GHz, but has 10 cores. And 3*8 < 2.8*10.....

It really seems the 1680 should be priced around $1500 to really make any sense. It what other spot in the intel line up does the 2600 out preform the 1600 in a single socket system at any price point?



I'm not sure how much this will effect the low end, which is what will make or break the new Mac Pro, but the top end was already going to be unattractive without a DP configuration. Without well placed performance/$ choices, it is looking like Apple is doing their best to kill the top end.

You know what I'm thinking now... it's possible that with extra heat, and with the reduced cache per core, and with the imperfect scaling of most software to cores, that the 1680 actually performs better than the 2680. Or maybe the performance is almost perfectly equal multicore, and therefore the better single core score puts it over the top. I can only find one geekbench score for a 2680v2, and it's a 20 core machine that scored 48000. Divide that in 2 and what do you have?

 

[deconstruct60]

As I wrote somewhere else, the only rational explanation for this CPU choice is that they REALLY care about cache. This processor has substantially more cache per core than any other choice in the 1600 line.

L3 cache comes with the core.

From Anantech's 12 core performance review http://www.anandtech.com/show/7285/intel-xeon-e5-2600-v2-12-core-ivy-bridge-ep


The 1680 is quite likely a 2680 with two cores (and QPI link) flipped off. Since it is the exact same thing you pay the exact same price. Pay for the privileged to have 2 present cores turned off and and broader dynamic Tubro range ( in part because the cores are turned off. )

If spend 40% time in Adobe CS 5 and 60% doing something that scales it makes some sense.

Or almost any other CPU, period.

In this price zone. Several years ago for Power, Itanium , etc.

You know what I'm thinking now... it's possible that with extra heat, and with the reduced cache per core,and with the imperfect scaling of most software to cores, that the 1680 actually performs better than the 2680.

How may workloads fit in a the L3 cache? Don't see the reduced heat comes from though. There are fewer cores in a 1680 but the clock range scales higher. The "extra" L3 cache is there so that scaling to the higher clock range is actually effective as oppose to increasing the number of no-op wait states about as much as throughput.

Or maybe the performance is almost perfectly equal multicore, and therefore the better single core score puts it over the top. I can only find one geekbench score for a 2680v2, and it's a 20 core machine that scored 48000. Divide that in 2 and what do you have?

Almost equal mulitcore scores? Wouldn't be surprising. Higher 1680 on real-world scalng computational workloads would be.

Single core score is far more effective achieved with the 1650 or even 1660 though. You have to got to a population that does A, B, and C to find a fit where this works. Generally, the longer the list the smaller the population.

 

[deconstruct60]

I'm not sure how much this will effect the low end, which is what will make or break the new Mac Pro, but the top end was already going to be unattractive without a DP configuration.

You are only looking at the CPUs. If Apple cranks up the mark-up on the GPUs ( under assumption that the buyers are price insensitive so $/performance isn't a priority. ) then lower cost ones can rise even though those the E5 1620/1650 are still in same price zone as previous generations. Same issue with the rest of the components (SSD , etc. ). Uniformly applied, these systems generally could take on that lower $/perf attribute. That generally the same quicksand the Cube rolled out on.

 

[Umbongo]

I'm not sure how much this will effect the low end, which is what will make or break the new Mac Pro, but the top end was already going to be unattractive without a DP configuration. Without well placed performance/$ choices, it is looking like Apple is doing their best to kill the top end.

I don't think that it is Apple trying to kill the top end so much as that market (DP workstations) being so tiny, and Apple now being so big, that it just isn't worth catering to.

 

[theSeb]

If there is a limited number of BTO CPU options then yes it looks like it is in (and higher GHz 10 core options out). It still doesn't make alot of sense from a multicore preformance perspective. The same folks making snarky comments above this post are all over that 1680 Geekbench thread clowning that the Geekbench scores are lower than the 2012 about equal to the 2010 machine offerings. With 10 cores the score would have been higher. So at this point if want to go faster than 2012 offerings you are now stuck picking Apple's $3,000+ CPU option at the very top end.

The win with the E5 1680 is only on single CPU Geekbench scores versus the 2010-2012 options.... which basically doesn't even appear in that forum thread. To be such a highly relevant factor for the targeted consumers it is also particularly odd that no one over there even mentions it.

So sure, cheer and clap for less multicore performance for just as much money as could have bought higher.

Neither the E5 1680 v2 nor the E5 2697 v2 have $/performance that looks good to wider markets. That these two are the only ones that Apple wants to dribble out looks like that Apple may be pricing the whole Mac Pro line up in looney toon land. I didn't think this was going to be a Cube repeat either, but it looks now to be drifting that way.

Are these people trying to compare a geekbench 2 score to a version 3 score?

 

[VirtualRain]

I could see Apple actually not offering a 10-core and going with the E5-1680 V2 (3GHz 8-core and same price as the 2680 V2) as it is a workstation CPU for uni-processor systems, and then the only E5-2600 V2 being the 12-core option. Which also gives it a big jump over the 8-core.

BTW, kudos... You called it (on the 1680v2)!

Although this CPU choice is not as compelling as the 2680v2, its of no concern to me as both would be out of my price range (and I can't even fully leverage a quad core most of the time - unless I'm using handbrake).

 

[deconstruct60]

Are these people trying to compare a geekbench 2 score to a version 3 score?

1680 v2 versus 2 CPU set-up with 2680 v2 ? No. There aren't any geekbench 2 scores for either one. It is a bit Apples to Oranges on systems and configuration. The 2680 v2 float and stream scores are bit oddball. And the reported clock is also oddball. If that entry is someone's homegrown memory/cpu clock setting concoction then it isn't all that helpful differentiating between the two.


P.S. If talking about E5 2697 v2 scores from June that were posted those really aren't vying for same price point that the x680 2 is. Oddly though there are 4 other "AAPLJ90,1" Geekbench 2 scores that showed up that didn't pop up on the rumor blogs. Some i7-3930 and a E5 1660 (v1) . Perhaps because those are pretty obvious just engineering lab configs that aren't going to ship (or just faked entries).