Should Intel release unlocked E5 Xeons sometime in the next 3 years?

[escrow4]

How about a locked high clocked X99 chip? I'd consider upgrading to a 5950 non K clocked at 4.5GHz all six cores, no turbo for 600 or so.

 

[MongGrel]

Here is a product from a few years back:

http://www.evga.com/articles/00537/

The EVGA Classified Super Record 2. It could overclock two LGA 1366 hexcore Xeons, but this is the sort of product that would not be purchased primarily for gaming.

http://forums.anandtech.com/showthread.php?t=2335636&highlight=l5639&page=88

http://www.ebay.com/itm/13145619354...PageName=STRK:MEBIDX:IT&rmvSB=true&rmvSB=true

:whiste:

 

[cbn]

How about a locked high clocked X99 chip? I'd consider upgrading to a 5950 non K clocked at 4.5GHz all six cores, no turbo for 600 or so.

It might be Intel could make the chip, but I'll bet the TDP even on 14nm would be over 160 watts.

If the TDP is over 160 watts (or whatever Intel rates LGA 2011-3) this makes extra work for Intel.

Then of course the question is how many people want this higher than stock TDP chip vs. the costs incurred to Intel to make the separate SKU and do the certification work on the socket.

If Intel feels the costs to make the chip are too high relative to the what they can expect to sell the chip then they won't do it.

This is most likely one reason Intel includes unlocked multipliers on certain chips. It lets them effectively offer something with the potential for higher performance and higher TDP (in order to satisfy certain parts of the market) without the extra work, cost and risk associated with creating extra SKUs beyond a certain power level.

P.S. This same reasoning is why I favor unlocked multipliers on certain E5 Xeons. The LGA 2011-3 large die processors are likely capable of a much higher level of performance (when appropriately cooled) than what we see Intel stocking them at, but maybe Intel feels target market is too small to justify the risk, energy and effort of making something with a higher TDP.

 

[DrMrLordX]

Great, look at my overclocked, "Ferrari" !

Fuzzy seats make it go faster!

That aside, I voted "yes" because it would be pretty funny, and the enthusiast crowd would love it despite there being few practical reasons why Intel should actually do such a thing.

If Intel released some unlocked, top-end skus, what would happen is that shady OEMs would make up their own Xeon skus and sell overclocked Xeon parts under that name at a markup. It'd be a <insertnameofOEMhere> "exclusive" part or what have you.

So sure you can't just OC an 8-core chip and try to pass it off as an 18-core chip but there are still questionable things OEMs could do to misrepresent the product being sold and cover it up with an overclock.

 

[cbn]

Great, look at my overclocked, "Ferrari" !

Fuzzy seats make it go faster!

Some of the E5 Xeons are pretty fast in single thread, but the major topic of this thread is the high core count Xeons and those are nothing like a "Ferrrari". More like a Turtle with the motto "slow and steady wins the race".

And regarding that "slow and steady" focus I expect that to become even more pronounced on the future top SKUs if Intel continues to increase core count. (re: If TDP stays the same and the power drop at each node is not enough to compensate for the increase in core count the base clocks will drop)

 

[cbn]

.

If Intel released some unlocked, top-end skus, what would happen is that shady OEMs would make up their own Xeon skus and sell overclocked Xeon parts under that name at a markup. It'd be a <insertnameofOEMhere> "exclusive" part or what have you.

So sure you can't just OC an 8-core chip and try to pass it off as an 18-core chip but there are still questionable things OEMs could do to misrepresent the product being sold and cover it up with an overclock.

If Intel thought there was a market for higher TDP chips they would spend the extra money, time and effort to create those SKUs. (See post #53 ---> http://forums.anandtech.com/showpost.php?p=37283253&postcount=53)

But if allowing overclocking on the top SKU gets more people buying more of those expensive processors I don't see the harm to Intel. (Actually boutique desktop OEMs like CyberPower do this already --> http://www.anandtech.com/show/7233/cyberpowerpc-gamer-xtreme-5200-desktop-review, but the processor is marketed as a pre-overclocked i7, etc)

 

[SOFTengCOMPelec]

Some of the E5 Xeons are pretty fast in single thread, but the major topic of this thread is the high core count Xeons and those are nothing like a "Ferrrari". More like a Turtle with the motto "slow and steady wins the race".

And regarding that "slow and steady" focus I expect that to become even more pronounced on the future top SKUs if Intel continues to increase core count. (re: If TDP stays the same and the power drop at each node is not enough to compensate for the increase in core count the base clocks will drop)

I guess a better analogy, would have been a Train or Bus (lots of passengers, lots of cores) vs Car (few passengers, few cores).

I.e. You are fitting a "Turbo/Super" charger, to a "TRAIN". How many teenage train enthusiasts, do you see fitting turbo/super chargers, to trains, over the weekend ?
(Joke, to illustrate the point, of reliability/practicality vs messing about)

AMD (in a small way), have tried the "too hot TDP" method, somewhat recently with the FX-9590/9370, at 220 watts TDP.
I rarely hear about anyone who has bought one, despite its 8 cores, surprisingly low price and partial performance improvement over the usual FX-8350. Part of the reason is the performance is STILL not that good, especially with single threaded, applications.

The market for it is very limited, partly because lots of people hate the idea of 220 W TDP. Also some/many opt for overclocking a cheaper FX-8350 (or similar), themselves.

Server cpus are intended to last at least 5 years (probably higher design life, e.g. 10 years), while running 24/7 at full load.

Widely overclocking (unlocked by Intel) Xeons, may well considerably shorten the life expectancy of the cpus. Perhaps to the point of ruling that out, as a market segment.
When we overclock 4790K's, we probably only use the computer for a few hours each day. Spending lots of that time, idling or NOT at 100% utilization as well.

But servers can be 24/7 at 100% load. Which could "burn out" a Xeon, in a lot less than its design life. Don't forget that it is not just the temperature, but also any increase in cpu voltage and other factor(s), such as electromigration.

But I don't know Intels (probably secret), life expectancy vs temperature/voltage/etc. So I don't know for sure, how much of a problem that would be.

 

[VirtualLarry]

Fuzzy seats make it go faster!

That aside, I voted "yes" because it would be pretty funny, and the enthusiast crowd would love it despite there being few practical reasons why Intel should actually do such a thing.

If Intel released some unlocked, top-end skus, what would happen is that shady OEMs would make up their own Xeon skus and sell overclocked Xeon parts under that name at a markup. It'd be a <insertnameofOEMhere> "exclusive" part or what have you.

So sure you can't just OC an 8-core chip and try to pass it off as an 18-core chip but there are still questionable things OEMs could do to misrepresent the product being sold and cover it up with an overclock.

"HP Hexa-Core", redux?

 

[Lepton87]

My understanding, is that a huge number of Xeons, are sold to server manufacturers, such as Dell, HP, IBM, etc.

They want their servers to be reliable, trouble free, quality, stable, accurate etc.

They are NOT going to touch overclocking with a 1 mile long barge pole. (My opinion).

The words, Overclocking and Xeons, should NEVER appear in the same sentence. (Joke, but I hope you can see, that the concept, is true, nevertheless).

.

BS. Back in the day, before SB, when every Xeon was overclock-able by changing FSB or whatever it was called, people very often bought Xeons instead of regular Core i7s , the same was true with Opterons. They were used in 1P systems and they were just regular i7, athlons with different binning and names though. The only exceptions were that EVGA mobo and Quad Father from AMD. NUMA was quite a pain in the ass in the desktop.

 

[SOFTengCOMPelec]

BS. Back in the day, before SB, when every Xeon was overclock-able by changing FSB or whatever it was called, people very often bought Xeons instead of regular Core i7s , the same was true with Opterons. They were used in 1P systems and they were just regular i7, athlons with different binning and names though. The only exceptions were that EVGA mobo and Quad Father from AMD. NUMA was quite a pain in the ass in the desktop.

Yes, in the distant, old days of computing, cpus were not locked.

But as time went on, they became increasingly locked down (except K, Black editions etc).

The rest is history.

EDIT:
BUT, strangely (in my view), some of the older (but not that old), Xeon series, WERE highly overclockable.

E.g. The X5650/5660/5670 etc ones.

Rightly or wrongly, Intel seems to have prohibited overclocking on Xeons, after the X5670s (etc). As far as they could readily do. From a technical point of view. E.g. Multiplier limited/locked.

But in those days, they only went up to a max of 6 cores per processor. Some articles seem to think that, that is part of the reason (if I remember, correctly).

 

[ClockHound]

I mentioned this in another thread, partly in jest. But, if they can't succeed in the ruthless low margin mobile market, they might want to learn from other established companies surviving in shrinking markets with a boutique offering. The Intel Custom Shop.

Need to send your poorly optimized software a serious statement? Want a 16 core 488W 4.88Ghz monsterXeon? Not a problem. Sign on the MOQ order form below for a limited edition starting at just $49,999 each with a full 72 clock cycle factory warranty. 2 core turbo available for extra charge.

Your initials on the IHS? Laser-etched personalized options begin at $4,999.

Want your pet's likeness preserved forever in breathtaking 14nm lithography? The Intel Custom Shop's team of highly paid engineers can make your big barking dreams small while shrinking your will to live.

Visit the Intel Custom Shop for the fastest, most exclusive processors with the biggest epeen ever. All at at price to match.

The Intel Custom Shop - where Moore's Law is bent, broken and violated.

 

[DrMrLordX]

"HP Hexa-Core", redux?

Ugh, yeah, that.

I mentioned this in another thread, partly in jest. But, if they can't succeed in the ruthless low margin mobile market, they might want to learn from other established companies surviving in shrinking markets with a boutique offering. The Intel Custom Shop.

Need to send your poorly optimized software a serious statement? Want a 16 core 488W 4.88Ghz monsterXeon? Not a problem. Sign on the MOQ order form below for a limited edition starting at just $49,999 each with a full 72 clock cycle factory warranty. 2 core turbo available for extra charge.

Your initials on the IHS? Laser-etched personalized options begin at $4,999.

Want your pet's likeness preserved forever in breathtaking 14nm lithography? The Intel Custom Shop's team of highly paid engineers can make your big barking dreams small while shrinking your will to live.

Visit the Intel Custom Shop for the fastest, most exclusive processors with the biggest epeen ever. All at at price to match.

The Intel Custom Shop - where Moore's Law is bent, broken and violated.

Don't forget the diamond-encrusted IHS. After all, diamond has a great k-value right? And it captures that "alluring" blue LED light just so . . .

 

[cbn]

AMD (in a small way), have tried the "too hot TDP" method, somewhat recently with the (and a FX-9590/9370, at 220 watts TDP.
I rarely hear about anyone who has bought one, despite its 8 cores, surprisingly low price and partial performance improvement over the usual FX-8350. Part of the reason is the performance is STILL not that good, especially with single threaded, applications.

The market for it is very limited, partly because lots of people hate the idea of 220 W TDP. Also some/many opt for overclocking a cheaper FX-8350 (or similar), themselves.

Even in multi-threaded those 220 watt FX 9 series AMD chips were not that great compared to a stock i7-4770K and i5-4690K:

http://www.anandtech.com/bench/product/1289?vs=836

http://www.anandtech.com/bench/product/1289?vs=1261 (I would have linked the i5-4670K, but the bench was incomplete for that one. So I used i5-4690K instead)

Then, of course, the Intel chips being much faster in single thread made it the choice for mainstream gamers. All this with a TDP almost 140 watts lower than the FX-9590.

Probably a better example of success with high TDP and larger die sizes is to look at Nvidia where the top Teslas, Quadros and gamer cards are around 250 watts for a single die chip. (Now granted the gamer cards don't run 24/7 full load, but I bet those Teslas get a workout and probably the Quadros to a lesser extent too.)

 

[SOFTengCOMPelec]

Even in multi-threaded those 220 watt FX 9 series AMD chips were not that great compared to a stock i7-4770K and i5-4690K:

http://www.anandtech.com/bench/product/1289?vs=836

http://www.anandtech.com/bench/product/1289?vs=1261 (I would have linked the i5-4670K, but the bench was incomplete for that one. So I used i5-4690K instead)

Then, of course, the Intel chips being much faster in single thread made it the choice for mainstream gamers. All this with a TDP almost 140 watts lower than the FX-9590.

Probably a better example of success with high TDP and larger die sizes is to look at Nvidia where the top Teslas, Quadros and gamer cards are around 250 watts for a single die chip. (Now granted the gamer cards don't run 24/7 full load, but I bet those Teslas get a workout and probably the Quadros to a lesser extent too.)

I (especially), and to an extent this thread, may be combining too many different concepts, at the same. I think there are 3 basic concepts, here:

(1)...There is the issue of if Xeons, should be locked down, or allow easy overclocking.

(2)...There is the issue, of if Intel should release much higher TDP Xeons, for particular niche usage situations. Such as very high end workstations.

(3)...If Intel should release cpus, with more than 8 cores (up to 18 even), which ALLOW overclocking, but are marketed, much like the 5960X (Extreme Haswell-E Unlocked 8 Core Socket 2011-3 Processor), are.

My opinions, are that:

(1)...Xeons should remain "locked" parts, and NOT encourage overclocking of them. This is important, to ensure stability, power efficiency, long life and reliability.

(2)...I don't have the facts and figures that Intel have almost certainly got. Without them, it is difficult to make informed decisions. Maybe the market for higher TDP parts, is seen as too small, and not profitable. Maybe the technical problems of cooling such huge TDP parts, especially in a low height 1u (server rack) environment, limit the marketability of such parts.

They may also get so hot, so quickly, as you start to raise the frequency and possibly voltage, that the gains are rather limited. Even a small rise in watts used per core, will be a much bigger number, when you multiply it by 18 (cores).

Also they may hold high computing efficiency (i.e. more instructions per watt), as important enough, to rule out such chips. I.e. The market place would not like the excessive heat, or the expensive of it.

(3)...Yes, I'm all for this. It would be great, to play around with 18 core versions of the '5960X' cpus. But being realistic. The 18 core cpus are approaching $4800. So it would have to be an affordable (to extreme hobbyists), to succeed. E.g. 10 or 12 cores, is probably more affordable/realistic, for Haswell generation stuff, at least.

 

[cbn]

I think there are 3 basic concepts, here:

(3)...If Intel should release cpus, with more than 8 cores (up to 18 even), which ALLOW overclocking, but are marketed, much like the 5960X (Extreme Haswell-E Unlocked 8 Core Socket 2011-3 Processor), are.

(snip)

(3)...Yes, I'm all for this. It would be great, to play around with 18 core versions of the '5960X' cpus. But being realistic. The 18 core cpus are approaching $4800. So it would have to be an affordable (to extreme hobbyists), to succeed. E.g. 10 or 12 cores, is probably more affordable/realistic, for Haswell generation stuff, at least.

If Intel were to make that kind of decision, I'm thinking it would be a balancing act of 1.) cost associated with creating a new Extreme edition SKU (at some lower price point) vs. any extra sales they might get for that new SKU 2.) Simply having an unlocked multiplier on an existing or future higher priced Xeon.

Which one (creating new higher core count extreme edition SKUs or converting existing or future locked E5s SKUs to unlocked) makes more sense for Intel and the community?

For certain I can't see any detriment to the existing community by having locked multipliers changed to unlocked multipliers on the top existing or future Xeon E5 SKUs. Major OEMs will use the chip in the same way the used the unlocked Xeon W3680, so no harm done there.

As far as Intel goes, the risk is basically zero and they may entice some people with that unlocked multiplier into buying up. (2.3 GHz is such a low clock on that E5 2699 v3 and scaling to 3.6 Ghz (for example) should still be quite favorable on the frequency voltage curve, though it will still require water cooling). Maybe the biggest risk to overclockers is that one of the cores is weak and can't keep up with the others. Though 18C @ 3.6 Ghz (for example), I'd imagine would be much less of a problem than trying to push all eighteen cores to 4.4+ Ghz or beyond.

Maybe the major questions begin to arise when we consider adding unlocked multipliers to SKUs other than the top E5 Xeon SKU?

Does changing the multiplier from locked to unlocked on the 12C/24T E5 2690 v3 (for example) cause some people to buy that instead of the more expensive E5-2699 v3? Or is having unlocked E5 2690 v3 still a good idea because some people who would have bought the lower priced i7-5960X will buy up instead?

I am of the opinion unlocked multipliers should also be on some of the E5s below the top SKU.

So at this point the question then becomes what is a better overall idea for midrange core count (~12) LGA 2011-3 processors? Extreme edition 12 core or unlocked E5 12 core Xeon? I think the unlocked 12c core Xeon makes more sense because Intel already charges a massive premium moving from 6C to 8C? What would they charge on 12C Extreme edition compared to what the 12C E5 Xeon already costs? This coupled to the risk of creating a new Extreme edition SKU at a higher price point?

 

[SOFTengCOMPelec]

If Intel were to make that kind of decision, I'm thinking it would be a balancing act of 1.) cost associated with creating a new Extreme edition SKU (at some lower price point) vs. any extra sales they might get for that new SKU 2.) Simply having an unlocked multiplier on an existing or future higher priced Xeon.

Which one (creating new higher core count extreme edition SKUs or converting existing or future locked E5s SKUs to unlocked) makes more sense for Intel and the community?

For certain I can't see any detriment to the existing community by having locked multipliers changed to unlocked multipliers on the top existing or future Xeon E5 SKUs. Major OEMs will use the chip in the same way the used the unlocked Xeon W3680, so no harm done there.

As far as Intel goes, the risk is basically zero and they may entice some people with that unlocked multiplier into buying up. (2.3 GHz is such a low clock on that E5 2699 v3 and scaling to 3.6 Ghz (for example) should still be quite favorable on the frequency voltage curve, though it will still require water cooling). Maybe the biggest risk to overclockers is that one of the cores is weak and can't keep up with the others. Though 18C @ 3.6 Ghz (for example), I'd imagine would be much less of a problem than trying to push all eighteen cores to 4.4+ Ghz or beyond.

Maybe the major questions begin to arise when we consider adding unlocked multipliers to SKUs other than the top E5 Xeon SKU?

Does changing the multiplier from locked to unlocked on the 12C/24T E5 2690 v3 (for example) cause some people to buy that instead of the more expensive E5-2699 v3? Or is having unlocked E5 2690 v3 still a good idea because some people who would have bought the lower priced i7-5960X will buy up instead?

I am of the opinion unlocked multipliers should also be on some of the E5s below the top SKU.

So at this point the question then becomes what is a better overall idea for midrange core count (~12) LGA 2011-3 processors? Extreme edition 12 core or unlocked E5 12 core Xeon? I think the unlocked 12c core Xeon makes more sense because Intel already charges a massive premium moving from 6C to 8C? What would they charge on 12C Extreme edition compared to what the 12C E5 Xeon already costs? This coupled to the risk of creating a new Extreme edition SKU at a higher price point?

Intel are rather secretive, about many things, including, WHY they have done things, the way they have. So we have been left to fend for ourselves. This has led to theories and even conspiracy stories, as to what is really going on.

I get the suspicion, that the reason why the 5960X/4690K/4770K/4790K DON'T support ECC ram, and don't necessarily support all of the feature set, such as one or more visualization options.
Is to minimize/stop attempts for users to either try and save money on much more expensive parts, by overclocking and/or attempting to produce something which is faster, than the standard Intel parts.

tl;dr
I think Intel have made the firm decision to block overclocking on Xeons and minimize the chance/risk that users will try and use overclocked parts in servers or workstations.

I have to admit that when the unlocked and highly overclockable, X5650, X5660, X5670, L5640 etc were current, I did not come across any stories of it causing problems, in the server world.
So in theory, the same thing, could happen again. I.e. That if the current Xeons were unlocked, computer assemblers/users could be trusted to NOT overclock mission critical equipment.

If Intel unlocked the Xeons and things went badly. Intel would be blamed. Intel could lose profits on their very profitable server line. E.g. Because people buy the much cheaper, low frequency, higher end parts. Then overclock them, so that they act like the higher frequency part, that they saved lots of money, by not buying it.

Also if it made their cpus (Xeons) unreliable/unstable/shortened-life-time. It would be Intel who would get the bad reputation and blame. Additionally they may also have to pay, indirectly with (dishonest in real terms) warranty claims, on wildly overclocked and overvoltaged, high end Xeon cpus.

The higher end chips, are probably very expensive to make. So if Intel has to keep on giving them out for free (warranty replacements), the costs will add up.

There is little incentive for Intel to change their policy on this matter. Because I can see few ways in which, allowing the Xeons to be overclocked, would benefit Intel.

You can't even argue, that it would make people, more likely to buy Intel Xeon parts, as it would make them better than the competition. Because at the moment, Intel are, arguably the main suppler (by a long way) of very high end server cpu chips.

 

[VirtualLarry]

The higher end chips, are probably very expensive to make. So if Intel has to keep on giving them out for free (warranty replacements), the costs will add up.

The actual per-chip mfg and testing cost is actually quite low, as I understand it. It mostly has to do with mm2 of silicon area, and yields. Overall yields can be improved by salvaging dies, by disabling defective portions, and selling down-binned or down-featured chips.

 

[JDG1980]

The top part, which seems to be a Xeon E5-2699 v3, seems to cost $4783.95 Source.

Assuming Intel continue to NOT allow overclocking (K) parts, until the very top part (this is NOT always the case, e.g. I5 4690K etc).

Then it would probably cost about $5000, for the unlocked top part.

It wouldn't make much sense for Intel to allow actual Xeon parts to be overclocked, for the reasons specified here. Xeons are what people choose when reliability is more important than the absolute top performance.

That said, if there was enough demand and high enough yields, I could see Intel binning the 18-core E5-2699 v3 die into a consumer processor (i7-5990X?) It wouldn't have the Xeon branding, and wouldn't support ECC RAM or multi-processor configurations. That said, very few software packages outside of servers even need 8 CPU cores (which is one reason why AMD's FX chips don't fare too well in many benchmarks), and the niche of people who need even more cores than that but aren't willing to pay professional prices is probably too small for Intel to care about.

 

[VirtualLarry]

That said, very few software packages outside of servers even need 8 CPU cores (which is one reason why AMD's FX chips don't fare too well in many benchmarks), and the niche of people who need even more cores than that but aren't willing to pay professional prices is probably too small for Intel to care about.

The distributed-computing crowd would love them!

 

[SOFTengCOMPelec]

The actual per-chip mfg and testing cost is actually quite low, as I understand it. It mostly has to do with mm2 of silicon area, and yields. Overall yields can be improved by salvaging dies, by disabling defective portions, and selling down-binned or down-featured chips.

That makes sense. A long time ago, I heard that the cpus (mainstream), only cost something like $5, each to make. Most of the extra "cost", was to pay back the huge development (and other) costs.

I thought, because they were 18 core, rather than 4 core, that it would cost a lot more. But a lot of die space is saved, because the quad cores, have a lot of junk slapped on them have the integrated gpu as well.

So it is more like going from 8 cores (estimate, without bothering to look up die area saved, with no igpu), to 18, which does not sound too bad. Especially (as you mentioned), they salvage dies, and separate them into three different die types. Which probably makes the yields better, as well.

Assuming, they also get good yields. Which they should, because by the time the big high end Xeons are put into production, the process node and architecture, have been in production and out on the market place, for a long time (relative to the typically rapid technological development pace).

There would still be hidden costs, to replacing burnt out, overclocked parts. Such as stocking-chips/admin/checking-testing-faulty-cpus etc etc.
Also they would be losing some potential production capacity, which would be better utilized, for making parts, which they can sell for lots of money. Rather than making ones, which they get nothing back for.

There is also the risk, that it could partly tarnishing the name of Intel. Because people would go on the internet (and other communication methods), and rather than being 100% honest, and saying I wildly over-voltaged my heavily overclocked chip from Intel, and blew it up.
They might say that their Intel cpu chip, broke on them again. Bad bad Intel.

There is also the risk that if yields are not too good yet, or have dropped, in a new plant, or for other reasons. The 18 core Xeons, will be especially that much harder to manufacture, and correspondingly more expensive to replace.

 

[cbn]

If Intel unlocked the Xeons and things went badly. Intel would be blamed. Intel could lose profits on their very profitable server line. E.g. Because people buy the much cheaper, low frequency, higher end parts. Then overclock them, so that they act like the higher frequency part, that they saved lots of money, by not buying it.

I don't think the low TDP versions of a particular core count E5 should have the unlocked multiplier.

For example, there should not be an unlocked multiplier on the cheaper, slower 12C/24T E5 Xeons so a person can make them into more expensive, faster 12C/24T E5 Xeons.

Example of cheaper, slower 12/24T E5 Xeon: $1329 65 watt 12/24T E2650L v3 --> http://ark.intel.com/products/81903/Intel-Xeon-Processor-E5-2650L-v3-30M-Cache-1_80-GHz

Example of a more expensive, faster 12/24T E5 Xeon: $2094 135 watt 12/24T E2690 v3 --> http://ark.intel.com/products/81713/Intel-Xeon-Processor-E5-2690-v3-30M-Cache-2_60-GHz

Only the top bin in a particular core count category should have unlocked multiplier.

Additionally they may also have to pay, indirectly with (dishonest in real terms) warranty claims, on wildly overclocked and overvoltaged, high end Xeon cpus. The higher end chips, are probably very expensive to make. So if Intel has to keep on giving them out for free (warranty replacements), the costs will add up.

The warranty issue you are mentioning is something that definitely deserves concern, but this situation already exists with chips like i7 5960X (which if overclocked void the warranty.) Furthermore, we have to consider the number of folks that would overclock a E5 Xeon will be much smaller than the group of people overclocking i7 5960X, and Intel is making more money on each Xeon E5 chip.

You can't even argue, that it would make people, more likely to buy Intel Xeon parts, as it would make them better than the competition. Because at the moment, Intel are, arguably the main suppler (by a long way) of very high end server cpu chips.

For someone overclocking a i7 5960X to 4.4 Ghz, pretty much the only upgrade path they would have is an overclockable higher count chip (Eg, A stock speed 14C/28T E5 2697 v3 just barely beats 8C/16T @ 4.4 Ghz in MT, but lacks the single thread performance).

 

[Shmee]

Everything should be unlocked.

 

[cbn]

Maybe one factor to consider are the high TDPs on various discrete GPUs (gamer/professional/server) and how Intel high core count E5 CPUs might compete with those? (Supposedly Skylake E5 Xeon with have AVX 512)

Regarding that above question ^^^^ (which I asked in the opening post), here is a review on the HP Hewlett-Packard Z840 (Videomaker.com January 2105):

http://www.videomaker.com/article/17877-hewlett-packard-z840-review

During our &#8220;fair&#8221; testing we used Adobe Premiere Pro to render and export full HD video, while simultaneously watching two 4K UHD videos as well as two 1080p videos. During this process there was not one bit of lag and the videos played smoothly. The video rendering and export completed amazingly fast and every other direction we gave to the HP Z840 was met with immediate results. This was not only due to the Xeon processors, but also to the incredible NVIDIA Quadro K5200 GPU. At 8GB of GDDR5 256-bit memory and using the Kepler architecture with 2,304 CUDA cores, and memory bandwidth of 192 GBps, the K5200 was a perfect addition to the Z840.

It has two (2) Intel Xeon E5-2640 v3 processors for a total of 16C/32T @ 2.6 Ghz base clock/3.4 Ghz turbo) and a Quadro K5200 video card.

http://ark.intel.com/products/83359/Intel-Xeon-Processor-E5-2640-v3-20M-Cache-2_60-GHz

http://www.nvidia.com/object/compare-quadro-gpus.html

At this time the price of the two Xeon E5-2640 v3 ($944 each) are roughly the same as the single Quadro K5200: http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&IsNodeId=1&N=100008333 600531778

So at this point in 2015 the CPU to GPU cost ratio is 1:1. Going beyond simply adding unlocked multipliers to certain E5s I am interested to see what other things Intel will do to either keep that CPU to GPU cost ratio intact, or to shift the ratio more towards CPU (by giving the CPU greater capacity for parallel operations,etc.)

 

[videogames101]

I mentioned this in another thread, partly in jest. But, if they can't succeed in the ruthless low margin mobile market, they might want to learn from other established companies surviving in shrinking markets with a boutique offering. The Intel Custom Shop.

Need to send your poorly optimized software a serious statement? Want a 16 core 488W 4.88Ghz monsterXeon? Not a problem. Sign on the MOQ order form below for a limited edition starting at just $49,999 each with a full 72 clock cycle factory warranty. 2 core turbo available for extra charge.

Your initials on the IHS? Laser-etched personalized options begin at $4,999.

Want your pet's likeness preserved forever in breathtaking 14nm lithography? The Intel Custom Shop's team of highly paid engineers can make your big barking dreams small while shrinking your will to live.

Visit the Intel Custom Shop for the fastest, most exclusive processors with the biggest epeen ever. All at at price to match.

The Intel Custom Shop - where Moore's Law is bent, broken and violated.

You just made my day. Nominated for quote of the year!

 

[cbn]

So far this thread has been a discussion of unlocked multipliers on higher core count Xeons, but how do folks feel about having unlocked multipliers on the top bin of lower core count Xeons?

For example, two six core Xeons overclocked to 4.4 Ghz beats a single eight core Extreme edition CPU at 4.4 Ghz.