Are Haswell Xeon CPUs overclockable?

[curioser]

And specifically, can they be OC with these boards? (which do support xeon CPUs)
Asus H97M-PLUS
Asus Z97M-PLUS
I'm looking at the Xeon E3-1230v3 CPU for either of these boards

Thanks!

 

[atp1916]

The H variant board does not support overclocking features at all, but the Z version does. You may also want to check that the Z97M board has bios support for the Xeon.

The real issue here though is that Xeons are multiplier locked, so your only OC option is to raise the BCLK... but that'll mess with the PCIe clock and blow stuff up real fast. Do you have a particular need for the server-oriented features that the 1230v3 provides?

 

[curioser]

The H variant board does not support overclocking features at all, but the Z version does. You may also want to check that the Z97M board has bios support for the Xeon.

The real issue here though is that Xeons are multiplier locked, so your only OC option is to raise the BCLK... but that'll mess with the PCIe clock and blow stuff up real fast. Do you have a particular need for the server-oriented features that the 1230v3 provides?

Nope. Don't have any needs for it. And as far as I've read those features are probably disabled. The CPU seems cheaper than an i7 4790k and I could probably get away with using ECC unbuffered ram. Which is slightly cheaper. Though the blowing stuff up part doesn't sit well with me. Overclocking may be out of the question. It seems I'd better forget about OCing altogether

 

[Tsumi]

You don't blow stuff up, but things get unstable very quickly. SATA, PCI-E, and USB clocks do not like being adjusted much, and all three of those are tied into the BCLK. Typically you can get a 105 mhz BCLK, if you're really lucky you can get to 108-110.

 

[bonehead123]

I have the same CPU on an Asus Maxiumis Hero VI Z87 board w/PC-2133 ram, and I am currently running it at 3.8ghz right now. Just remember that it runs a little less $$ because it does NOT have an integrated GPU....and a few degrees cooler than a comparable i7-4xxx

It has however taken alot of fiddling with various settings in the bios and AI Suite, but it's been stable for a week now. This weekend I will be stress-testing it to see how it handles that...

So far nothing has "blown up"....all my stuff works just fine ie USB/SATA/PCIe etc...

Perhaps with a Z97 board you might get a little more out of it, or nOt

 

[atp1916]

Jeez guys, when i said "blowing up" i wasn't meaning it in the literal sense.

I was simply implying that system stability goes from stable to unstable very, very fast.

Edit: stable to unstable very fast when messing with the BCLK...since i need to spell things out apparently.

 

[bonehead123]

Jeez guys, when i said "blowing up" i wasn't meaning it in the literal sense.

I was simply implying that system stability goes from stable to unstable very, very fast.

Edit: stable to unstable very fast when messing with the BCLK...since i need to spell things out apparently.

Well D U H.....

 

[schizrade]

This derailed fast. lol

So no, OP, do not buy a Xeon if you want to OC, it wont work very well, and by that I mean it wont work at all like you intend it to.

 

[curioser]

This derailed fast. lol

So no, OP, do not buy a Xeon if you want to OC, it wont work very well, and by that I mean it wont work at all like you intend it to.

Lol. Yeah I knew you weren't literal about the blowing up part.

Well, it could get hot and unstable fast. I don't have a water cooler setup, and it will be installed in a cramped mATX case (SilverStone TJ08B-E)

I was looking only for a negligible amount of OCing. But I'm forgetting about OCing altogether at this point.
I initially had plans to build it around the i7 4690k but I slowly started to realize the advantages of running a xeon CPU. Aside from the xeon being cheaper, it also runs cooler and like I mentioned before I can use ECC unbuffered ram (cheaper).

 

[ssnyder28]

I don't think any xeons have been overclockable since X58 chipset...

Yea you might get lucky and squeeze a few hundred mhz out of the chip, but thats it.

 

[curioser]

I have the same CPU on an Asus Maxiumis Hero VI Z87 board w/PC-2133 ram, and I am currently running it at 3.8ghz right now.

It has however taken alot of fiddling with various settings in the bios and AI Suite, but it's been stable for a week now. This weekend I will be stress-testing it to see how it handles that...

I'd appreciate it if you could update the thread with your stress testing results. Thanks

 

[curioser]

I have the same CPU on an Asus Maxiumis Hero VI Z87 board w/PC-2133 ram, and I am currently running it at 3.8ghz right now. Just remember that it runs a little less $$ because it does NOT have an integrated GPU....and a few degrees cooler than a comparable i7-4xxx

It has however taken alot of fiddling with various settings in the bios and AI Suite, but it's been stable for a week now. This weekend I will be stress-testing it to see how it handles that...

So far nothing has "blown up"....all my stuff works just fine ie USB/SATA/PCIe etc...

Perhaps with a Z97 board you might get a little more out of it, or nOt

How did the stress-testing go?

 

[wirk]

Here I am: Xeon E5-1680v3 running @4.4 GHz OC/WC with 128 GB ECC RAM in Asus X99-E WS mobo. All single-socket Xeons are OC-ready. Stop spreading nonsense guys.

 

[curioser]

Here I am: Xeon E5-1680v3 running @4.4 GHz OC/WC with 128 GB ECC RAM in Asus X99-E WS mobo. All single-socket Xeons are OC-ready. Stop spreading nonsense guys.

Running stable? at what blck are yo urunning it at. for how long did you stress test it. yadda, ydda. It's not applicable to the chipset I'm running right now. But I guess it could help to those running their xeons on an x99 motherboard
Thanks

 

[schizrade]

Here I am: Xeon E5-1680v3 running @4.4 GHz OC/WC with 128 GB ECC RAM in Asus X99-E WS mobo. All single-socket Xeons are OC-ready. Stop spreading nonsense guys.

Not the one OP is looking at smart guy.

 

[Riccochet]

Someone looking at multi-socket Xeon's isn't going to want to overclock. If you're crunching data or running a VM host do you really want to risk stability?

Single socket Xeon's basically i7's with more cache, ECC support and no GPU. If you don't mind using a 2011 mobo they overclock nicely.

 

[curioser]

Someone looking at multi-socket Xeon's isn't going to want to overclock. If you're crunching data or running a VM host do you really want to risk stability?

Single socket Xeon's basically i7's with more cache, ECC support and no GPU. If you don't mind using a 2011 mobo they overclock nicely.

That's interesting. Most people are saying that even xeons running on x99 don't OC that well. That I can expect 103 blck at best.

 

[drescherjm]

BCLK is not how you overclock E5-16XX xeons. For these you use multiplier overclocking.

 

[curioser]

Not the one OP is looking at smart guy.

Well. let's encourage sharing even if they're slightly OT. I'm curios anyhow. I'm still open to the idea of moving to the x99 chipset. Even if It means saving for another 2-3 month to get there. I'm just too impatient to wait another month.
So far I've bought an el cheapo mtax q87 board for $20. It's not going to OC, period. So i'd definitely get the xeon e3-1230 v3 for the time being. If I'm satisfied with it during a 15-day trial period, I'll keep that rig.
OC would only be an option if gaming isn't that good with the stock speeds of the hw.
In that case I'd replace the mobo with a z97 mobo. And If OCing the xeon on a z97 doesn't work out, I'd then sell the xeon and get an i5-4690k. If none of that performs to my expectations then an upgrade to the next best chipset (the x99) is in order

 

[curioser]

BCLK is not how you overclock E5-16XX xeons. For these you use multiplier overclocking.

Ahh ok. That's definitely some misleading info I'm getting then.

 

[Bill1024]

Someone looking at multi-socket Xeon's isn't going to want to overclock. If you're crunching data or running a VM host do you really want to risk stability?

Single socket Xeon's basically i7's with more cache, ECC support and no GPU. If you don't mind using a 2011 mobo they overclock nicely.

I would overclock dual xeons for crunching BOINC or Folding at home.
The EVGA SR-2 still is a popular board for overclocking two xeons.

 

[XxDaRkReAp3rxXdOtCoM]

Here I am: Xeon E5-1680v3 running @4.4 GHz OC/WC with 128 GB ECC RAM in Asus X99-E WS mobo. All single-socket Xeons are OC-ready. Stop spreading nonsense guys.

I'm about to pull the trigger on an E5-1660v3 ES which is the equivalent to a 5960X. Hopefully it will be compatible with one of the X99 workstation boards (thinking of getting the cheap MSI X99S SLI Plus).

 

[Tsumi]

Ahh ok. That's definitely some misleading info I'm getting then.

Enthusiast socket (2011) single-socket Xeons are multiplier overclockable. Mainstream socket (1155, 1150, etc) non-K processors (with the exception of the G3528) are not multiplier overclockable.

 

[curioser]

Enthusiast socket (2011) single-socket Xeons are multiplier overclockable. Mainstream socket (1155, 1150, etc) non-K processors (with the exception of the G3528) are not multiplier overclockable.

Thanks. I'll keep that in mind when buying my next cpu

 

[wirk]

Running stable? at what blck are yo urunning it at. for how long did you stress test it. yadda, ydda. It's not applicable to the chipset I'm running right now. But I guess it could help to those running their xeons on an x99 motherboard
Thanks

Someone looking at multi-socket Xeon's isn't going to want to overclock. If you're crunching data or running a VM host do you really want to risk stability?
Single socket Xeon's basically i7's with more cache, ECC support and no GPU. If you don't mind using a 2011 mobo they overclock nicely.

Single socket Xeons overclock precisely as nicely as their consumer cousin i7-5960X. I have my Xeon 1680 with Swiftech X240-X WC and without any special effort I get stable 4.4 GHz.

That's interesting. Most people are saying that even xeons running on x99 don't OC that well. That I can expect 103 blck at best.

Whaaaaaaaa? These are completely idiotic rumors.

I'm about to pull the trigger on an E5-1660v3 ES which is the equivalent to a 5960X. Hopefully it will be compatible with one of the X99 workstation boards (thinking of getting the cheap MSI X99S SLI Plus).

Xeons are DEFINITELY compatible with SOME X99 boards. Main issue here is that not all X99 mobos support ECC RAM, apparently only high-end ones. But Xeon does not need ECC RAM for running so it might be that Xeon runs in every X99 mobo. BUT I advise to check for the Xeon support to avoid potential problems. Repeat again: HIGH-END X99 mobos SURELY support Xeons including 128 GB ECC RAM.

Enthusiast socket (2011) single-socket Xeons are multiplier overclockable. Mainstream socket (1155, 1150, etc) non-K processors (with the exception of the G3528) are not multiplier overclockable.

Thanks. I'll keep that in mind when buying my next cpu

Take the i7-5960X processor which is bit castrated version of Xeon, single-socket Xeons are supporting exactly the same OC. On the opposite note there are some rumors claiming the Xeons OC better than the i7-5960X because the consumer version is made from lower grade Xeons. I have not found any facts supporting such claims. It seems it rather depends on the lottery of getting better binned chips and is the reason for the existence of silicon lottery.

I would overclock dual xeons for crunching BOINC or Folding at home. The EVGA SR-2 still is a popular board for overclocking two xeons.

Dual processor OC is quite paranoic idea since processors have to be kept in sync which makes tuning the OC quite an issue and puts question mark on the overall system stability. No wonder Intel locked the dual Xeons.

 

[lutjens]

Dual processor OC is quite paranoic idea since processors have to be kept in sync which makes tuning the OC quite an issue and puts question mark on the overall system stability. No wonder Intel locked the dual Xeons.

Overclocking a duallie does have more factors at play, the biggest of which are stable power delivery to both chips and the fact that not one slab of silicon is present in the system but two, making well thought out system and CPU cooling an absolute necessity.

Increased latency between the two processors can easily make memory speed increases much more difficult, but this is negated by using only CPU multipliers to overclock.

By locking the dual capable Xeons, Intel has hard locked their very best and most interesting chips, even for those who want to use them in a single CPU motherboard (as the E5-1600 V3 series lacks high core count chips that are available to the channel). Intel should unlock their top SKUs only and allow those who purchase them to use that ability. This ability should come with purchasing Intel's flagship products and if the end user doesn't want to use it, they don't have to. The lack of this ability is the sole reason why I personally didn't purchase $20k of Intel processors last September and I'd hypothesize that quite a few others declined to purchase the top Xeon for the same reason.

 

[XxDaRkReAp3rxXdOtCoM]

Isn't she beautiful

 

[aphexcoil]

Are the Xeon E5-16xx's solder or paste?

 

[XxDaRkReAp3rxXdOtCoM]

Yes, they're LGA 2011 so it's solder and not TIM

 

[wirk]

By locking the dual capable Xeons, Intel has hard locked their very best and most interesting chips, even for those who want to use them in a single CPU motherboard (as the E5-1600 V3 series lacks high core count chips that are available to the channel). Intel should unlock their top SKUs only and allow those who purchase them to use that ability. This ability should come with purchasing Intel's flagship products and if the end user doesn't want to use it, they don't have to. The lack of this ability is the sole reason why I personally didn't purchase $20k of Intel processors last September and I'd hypothesize that quite a few others declined to purchase the top Xeon for the same reason.

Intel has good reason not allowing overclocking Xeons apart of single socket ones. People would not be setting overclocking for both processors properly, correctly, or made gross errors. Then they would experience problems blaming Intel and everybody else on it. Multisocket processors anyway are intended for rock stable professional systems, any problems due to overclocking would by undermining Intel reputation. Single socket is simple, no problems with sync.

 

[Tsumi]

Intel has good reason not allowing overclocking Xeons apart of single socket ones. People would not be setting overclocking for both processors properly, correctly, or made gross errors. Then they would experience problems blaming Intel and everybody else on it. Multisocket processors anyway are intended for rock stable professional systems, any problems due to overclocking would by undermining Intel reputation. Single socket is simple, no problems with sync.

That's just an excuse though. The primary motivator for locking them down is money. An overclocker takes responsibility for overclocking stability. The masses don't know how to overclock.

 

[XxDaRkReAp3rxXdOtCoM]

That's just an excuse though. The primary motivator for locking them down is money. An overclocker takes responsibility for overclocking stability. The masses don't know how to overclock.

I'd have to agree. Otherwise you would get people buying the 2600 series over the 1600 and OCing them to equal levels:

$1080 tray price: http://ark.intel.com/products/82766/Intel-Xeon-Processor-E5-1660-v3-20M-Cache-3_00-GHz
$939 tray price: http://ark.intel.com/products/83359/Intel-Xeon-Processor-E5-2640-v3-20M-Cache-2_60-GHz

So if you want a faster 2600 you buy the 2667 for $2057 since you can't OC it:

http://ark.intel.com/products/83361/Intel-Xeon-Processor-E5-2667-v3-20M-Cache-3_20-GHz

 

[Tsumi]

I'd have to agree. Otherwise you would get people buying the 2600 series over the 1600 and OCing them to equal levels:

$1080 tray price: http://ark.intel.com/products/82766/Intel-Xeon-Processor-E5-1660-v3-20M-Cache-3_00-GHz
$939 tray price: http://ark.intel.com/products/83359/Intel-Xeon-Processor-E5-2640-v3-20M-Cache-2_60-GHz

So if you want a faster 2600 you buy the 2667 for $2057 since you can't OC it:

http://ark.intel.com/products/83361/Intel-Xeon-Processor-E5-2667-v3-20M-Cache-3_20-GHz

And if you're already at the top tier, you have to buy the next generation top tier + motherboard to get more performance.

 

[XxDaRkReAp3rxXdOtCoM]

Having them locked makes perfect sense. I can't imagine something like a 155w 15c/30t CPU being unlocked and being able to OC it much, heat output would be ridiculous. Especially in a server environment. Dell sells blades that you can configure with four 130w CPUs and they're small as hell https://www.youtube.com/watch?v=JMl-2cj6lT4 I'm going deaf just thinking about fan decibels.

 

[Tsumi]

Having them locked makes perfect sense. I can't imagine something like a 155w 15c/30t CPU being unlocked and being able to OC it much, heat output would be ridiculous. Especially in a server environment. Dell sells blades that you can configure with four 130w CPUs and they're small as hell https://www.youtube.com/watch?v=JMl-2cj6lT4 I'm going deaf just thinking about fan decibels.

And over the top watercoolung systems can cool 300+ watt CPUs and GPUs without being heard.

It's not about any actual limitations, heat or otherwise. It's money.

 

[lutjens]

Intel has good reason not allowing overclocking Xeons apart of single socket ones. People would not be setting overclocking for both processors properly, correctly, or made gross errors. Then they would experience problems blaming Intel and everybody else on it. Multisocket processors anyway are intended for rock stable professional systems, any problems due to overclocking would by undermining Intel reputation. Single socket is simple, no problems with sync.

There are a number of issues with your line of reasoning:

1) As others have mentioned, issues with overclocking are the responsibility of the overclocker. What you are saying is the equivalent of a vehicle owner modifying their car to produce more horsepower and blaming the manufacturer for reliability issues they have with it. If people encounter errors while overclocking, they have the very simple option to lower the damn clock speed until the errors go away!!!
2) Many people who'd buy unlocked top SKU chips (the only ones that should be unlocked) would be buying them as part of a pre-validated package from a boutique builder, so their personal knowledge of overclocking isn't relevant. Those who are buying the chips separately are certainly NOT going to act irresponsibly with them, considering their very high price. Someone is HIGHLY unlikely to slam 1.6V through a $4,000 CPU like they might through a $400 i7.
3) You say Intel's reputation will be undermined. Like it currently is when people have issues with their i7 overclock? No. Intel says "return your clock speed to default and try again". Again, only when someone has issues when running at default would such a statement be justified. Issues you encounter when you overclock or as a result of an overclock are NOT the manufacturer's problem and the manufacturer cannot be blamed for such errors.
4)"Sync" is not an issue. A processor that ceases to be reliable as the result of an overclock causes the issues. Again if issues occur during an overclock...lower the CPU's clock speed!! This is not rocket science...

I'd have to agree. Otherwise you would get people buying the 2600 series over the 1600 and OCing them to equal levels:

$1080 tray price: http://ark.intel.com/products/82766/Intel-Xeon-Processor-E5-1660-v3-20M-Cache-3_00-GHz
$939 tray price: http://ark.intel.com/products/83359/Intel-Xeon-Processor-E5-2640-v3-20M-Cache-2_60-GHz

So if you want a faster 2600 you buy the 2667 for $2057 since you can't OC it:

http://ark.intel.com/products/83361/Intel-Xeon-Processor-E5-2667-v3-20M-Cache-3_20-GHz

What would benefit Intel is to sell more 2600 series chips, as they have a higher profit margin than the 1600 series. Only the very top, flagship SKU should be unlocked, the E5-2699 V3. This ensures that you can't take a lesser chip in the SKU stack and overclock it to match or exceed the performance of the flagship (protecting the flagship's profitability). But purchasers of the flagship CPU shouldn't be restricted from overclocking their CPUs if they choose to.

Also consider the following:

1) OEM servers using the flagship CPU for critical tasks are NOT going to overclock. They lack the ability to do so. The fact that the CPU would be unlocked would be completely ignored by the system.
2) Intel's dual CPU chipset (the C226) doesn't support BCLK strap overclocking and may not even support multiplier based overclocking (multiplier support is unknown as no chips exist), rendering an unlocked flagship usable in only single CPU based systems.

 

[lutjens]

And if you're already at the top tier, you have to buy the next generation top tier + motherboard to get more performance.

People with the money to buy the top SKU of E5-2600 Xeon (if it were unlocked) will very likely simply buy the next generation and overclock it as well. These folks want the best and are willing to pay huge dollars to get it. Unfortunately at present, Intel seems to not want this money and totally ignore this segment. Intel is offering these folks the equivalent of a Corvette ZR-1 that's speed governed to 60 mph because they're concerned people won't drive them responsibly and may have an accident driving one.

 

[XxDaRkReAp3rxXdOtCoM]

People with the money to buy the top SKU of E5-2600 Xeon (if it were unlocked) will very likely simply buy the next generation and overclock it as well. These folks want the best and are willing to pay huge dollars to get it. Unfortunately at present, Intel seems to not want this money and totally ignore this segment. Intel is offering these folks the equivalent of a Corvette ZR-1 that's speed governed to 60 mph because they're concerned people won't drive them responsibly and may have an accident driving one.

That's because these are enterprise CPUs really meant for servers. It just so happens that some of them are available via retail channels to use for workstations. The REAL Xeons (E7s) have no retail availability:

http://ark.intel.com/products/family/78584/Intel-Xeon-Processor-E7-v2-Family

 

[XxDaRkReAp3rxXdOtCoM]

And over the top watercoolung systems can cool 300+ watt CPUs and GPUs without being heard.

It's not about any actual limitations, heat or otherwise. It's money.

Watercooling CPUs in datacenters is rare, most of them are all air. There are boards that support four 155w E7s which is 600w+ already, if you could add overclocking to it, that's a hell of a lot of heat that would pretty much need something like DCLC http://www.coolitsystems.com/index.php/data-center.html That's not even the highest density possible. Take a look at Supermicro's TwinBlade for example:

http://www.supermicro.com/products/SuperBlade/module/SBI-7228R-T2F2.cfm

This is a single blade and check out what it says for supported CPUs:

Intel® Xeon® processor E5-2600 v3 family *

* E5-2697 v3 at 30°C / E5-2699 v3 at 25°C

"Supporting up to 120 DP servers with 240 Intel® Xeon® E5-2600 v3 family processors per 42U rack, the TwinBlade achieves an unmatched 0.35U per DP node." and there's already limits to what TDP processors you should be using depending on ambient temps.

That's about 1U for three nodes, and each node is dual socket capable. Potentially six 150w CPUs in about 1U which about yea tall:

http://ri-vier.eu/images/1U-Rackmount/rpc-170.jpg

Full 42U rack = ~36000w max @ 25c recommended. Now if you could OC these for the sake of argument, you're looking at another 240*n, say 100w extra. 60000w total.

 

[Tsumi]

Overclockers that are watercooling wouldn't be putting those on racks.