Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/634
Article Thumbnail

Intel CPUs: Xeon E5 vs. Core i7

Written on May 15, 2015 by Matt Bach
Share:

Introduction

If you take a close look at the certified configurations we offer here at Puget Systems you will find that while we are a high-end custom PC builder, we actually only have a single system (the Deluge XL) that utilizes the Intel Core i7 High End processors. We don't have any problems with the Core i7 High End CPUs, we simply feel that there is a better option in most situations: the Intel Xeon E5 CPUs.

A number of years ago, we started moving more of our systems over to the Xeon E5 CPUs to the point that all of our Genesis systems (which are designed primarily for content creation) now exclusively use Xeon processors. Every once in a while, we get a customer who asks us why we are using server CPUs instead of the "faster" Core i7 CPUs. This is an understandable question since Intel puts a lot of money into marketing the Core i7 processor lines to consumers, but almost nothing on marketing the Xeon CPUs to that same demographic.

The short of the matter is that - core vs core and frequency vs frequency - Xeon CPUs are just as fast as Core i7 CPUs. There are differences in the product lines that may make either Xeon or Core i7 a better choice for your situation, but if you take a Xeon E5 CPU and a Core i7 CPU that have the same specs the performance in any task will be identical.

In addition to backing up this claim by examining a set of benchmarks, in this article we are also going to go over the differences between Xeon E5 and Core i7 High End CPUs to help you understand when one or the other would be the better choice for your system.

CPU Specifications

To start off, we are going to take a look at the basic specifications of the High End Core i7-5XXX, single socket Xeon E5 v3, and dual socket Xeon E5 v3 CPUs (which can be used as either a single CPU or in a dual CPU configuration). There are other product lines out there (including quad socket Xeon E5, Xeon E7, and low voltage models) but for this article we are going to focus just on the standard CPUs from these three product lines:

Product Line # of Models Cores Base Frequency Max Boost Frequency Smart Cache TDP ~MSRP
High End Core i7-5XXX 3 6-8 3-3.5 GHz 3.5-3.7 GHz 15-20 MB 140W $389-$999
Single Socket Xeon E5 v3 5 4-8 3-3.7 GHz 3.5-3.8 GHz 10-20 MB 140W $294-$1,723
Dual Socket Xeon E5 v3 20 4-18 1.6-3.5 GHz 1.6-3.7 GHz 10-45 MB 85-160W $213-$3,888

[+] Show full CPU list

The chart above is a huge simplification of these product lines, but it is much easier to see the main product line differences when it is compressed like this. Alternatively, we have a chart that lists the individual specifications for all 28 CPUs collapsed below the chart. If you want to see all the full specs for each CPU, you can view them at the Intel Ark pages:

The main thing we want to point out here is simply the shear amount of options Xeon E5 CPUs give you. Compared to the three Core i7 CPUs, there are five single socket Xeon E5 v3 and twenty dual socket Xeon E5 v3 CPUs. Where you are limited to either six or eight cores and a frequency range of just 3-3.5GHz with the Core i7 CPUs, you can choose from between four and eighteen cores and a frequency range of 1.6-3.7GHz with the Xeon E5 v3 CPUs.

If you look at the individual models, you will also see that the Core i7 CPU are not even really unique. The i7-5960X has the same specs as the E5-1660 v3 and the i7-5930K has almost the same specs as the E5-1630 v3 (just a .1GHz lower maximum Turbo Boost). Even price-wise they are not that different. The i7-5930K and E5-1650 v3 have the exact same $593 MSRP and the Core i7-5960X is only about $80 cheaper than the Xeon E5-1660 v3. 

Really, the only unique Core i7 CPUs is the i7-5820K - not because it is better, but because it only has 28 PCI-E lanes versus the 40 PCI-E lanes all the other CPUs have. 

CPU Features

Looking at just the specifications, it would appear that there is not much that differentiates a Core i7 and Xeon CPU. This is because the main differences are not spec-based, but rather the features found in each product line:

Product Line Overclocking Support Max CPUs Max Memory ECC RAM Support VPro VT-x/VT-d TXT
High End Core i7 Yes 1 64GB No No Yes No
Single Socket Xeon E5 v3 No 1 768 GB Yes Yes Yes Yes
Dual Socket Xeon E5 v3 No 2 768 GB Yes Yes Yes Yes

Really, the one advantage the Core i7 CPUs have is that they support CPU overclocking. The Xeon CPUs can use twelve times as much memory as the Core i7 CPUs, support ECC RAM, and include support for vPro and TXT. While vPRO and TXT may not be that important for most users, the ability to utilize large amounts of ECC RAM is extremely useful in certain situations.

Motherboard Compatibility

After explaining to a customer how the only advantage of Core i7 CPUs is that they support overclocking, the next question is often: "but what if I don't want to use a server board?" Many server boards don't have features like onboard audio, USB 3.0, SLI support, or have more than a handful of USB ports. Luckily, you do not need a server board in order to use a Xeon E5 v3 CPU.

While most desktop boards will not have support for vPro and TXT, we have been able to use Xeon E5 v3 CPUs and registered ECC memory on every ASUS and ASRock X99 board we have tried without few, if any, problems. The eight RAM slots on most X99 boards will limit you to a maximum of 256GB (8x32GB) of RAM instead of the 768GB the Xeon CPUs technically support, but that is the only issue we've seen. In fact, we are currently using Xeon E5 v3 CPUs on the ASUS X99 Deluxe motherboard on all of our Genesis I and Genesis I Quiet Edition systems. This lets us have up to 18 cores and 256GB of RAM in a workstation without giving up desktop features like USB 3.0, onboard audio, or WiFi.

The only caveat is that X99 motherboards may not officially list support for Xeon E5 v3 CPUs or registered ECC memory. For example, the Asus X99 Deluxe board we use lists only a handful of Xeon E5 CPUs as being supported, and never mentions anything about registered ECC RAM. We know they work on that board since we have thoroughly tested it, but it is entirely possible another brand or model may use a BIOS that does not allow for Xeon or Registered ECC memory to be used.

Xeon vs. Core i7 Performance

After looking at the specifications and features of Xeon and Core i7 CPUs, it is clear that you can get the same or higher operating frequency and core count from a Xeon CPU. However, even though the Core i7-5XXX and Xeon E5 v3 CPUs use the same architecture we sometimes hear claims that Core i7 CPUs are "optimized for speed" while Xeon CPUs are "optimized for reliability".

The reliability aspect is pretty easy for us to address - we have only had a single Core i7-5XXX CPU fail and no Xeon E5 v3 CPUs fail from the hundreds we have sold so both are extremely reliable. The question of whether Core i7 CPU's are faster, however, requires a little bit more work. To debunk this myth, we ran a set of benchmarks on a Core i7-5960X and a Xeon E5-1660 v3. These are actually the only Core i7 and Xeon E5 CPU that are completely identical in terms of specifications. Both of these CPUs have 8 cores, 3.0GHz base frequency, 3.3 GHz all core Turbo Boost, and 3.5 GHz maximum turbo boost.

For our test system, we used the following hardware:

Xeon CPUs are typically used with ECC or Reg ECC RAM which is technically slightly slower than standard memory. In order to see if Reg ECC RAM affects performance, we will be testing the Xeon E5-1660 v3 with both standard RAM and Reg ECC RAM. This will allow us to test if - core for core and GHz for GHz - a Core i7 CPU is faster than a Xeon E5 v3 CPU or not.

 
 


We are not going to delve too deep into these results since they are pretty straight-forward. There is anywhere from a 0-2% variance in performance across our tests, but which CPU is faster changes from benchmark to benchmark. The funny thing is that the Core i7-5960X was actually not the faster CPU. Most of the time, the E5-1660 v3 was actually a hair faster (although the difference is arguably within the margin of error for these kinds of benchmarks).

If you average out all these results, the E5-1660 v3 with standard RAM was overall the fastest, followed by the i7-5960X with about a .25% performance drop. Lastly, the E5-1660 v3 with Reg ECC memory was about .4% slower than the E5-1660 v3 with standard RAM (or about .16% slower than the i7-5960X). According to our benchmarks, this means that the E5-1660 v3 is actually overall faster than the Core i7-5960X. However, the difference is so minuscule that the difference we saw is likely to be caused by normal testing variations or minor differences in manufacturing batches.

Conclusion

To sum up this article, there are four main points we want to make:

  1. Clock-per-clock, Core i7-5XXX and Xeon E5 v3 CPUs have identical performance
  2. Xeon E5 v3 CPUs have a much wider range of models than Core i7-5XXX CPUs allowing for higher (and lower) core counts and a wider range of operating frequencies
  3. Xeon E5 v3 CPUs have much higher RAM capacity through the use of Registered ECC memory (768GB versus 64GB)
  4. You do not need a server board to use a Xeon E5 v3 CPU. Most X99 motherboards work great even with Reg ECC memory (although Reg ECC memory is usually not officially supported)

So why would you ever buy a Core i7-5XXX CPU instead of a Xeon E5 v3? What it comes down to is that Core i7 CPUs are usually slightly cheaper than their Xeon E5 v3 counterparts and they allow for CPU overclocking. If you do not plan on overclocking, we highly recommend you consider using a Xeon instead of a Core i7 CPU. You get a much wider range of options - which allows you to get exactly the core count and frequency that is best for your application - and the capability to have huge amounts of system RAM. Even if you don't ever anticipate needing more than 64GB of RAM, having the option for future upgrades is almost never a bad thing.

Tags: Xeon, Core i7
Regi Ellis

How's does over clocking change the numbers here? Most people opting for the 5969x will most likely overclock, which would give it the advantage over the Xeon. Do you guys plan on doing a second test with a overclock 5690x, E5 8 core, and the 10 core.

Posted on 2015-05-31 02:38:57

Overclocking would improve performance of the Core i7, and isn't an option on the Xeon. In most applications, at least ones that are CPU bound (rather than being limited by RAM, GPU, etc) increasing the clock speed gives a near-linear performance increase. So overclock 20%, you get 20% faster performance.

However, I would note that most of the people who purchase an 8-core CPU in a system from us don't overclock. Yes, some request that - and I'm sure some do it on their own - but some folks prefer stability and don't want to risk pushing a CPU over its rated specs... even if the particular CPU handles it pretty well, like most of the i7 5000-series do.

Posted on 2015-05-31 04:07:33
eurorootz

Since when does a 20% overclock translate to 20% performance increase? Aside from synthetic benchmarks or benchmarks for apps no one cares about or uses I would like to see a 20% performance increase on paper.

Posted on 2015-09-01 04:05:50

A lot of programs benefit almost linearly from clock speed increases. It depends on whether the application is only CPU limited or has other factors affecting performance too. Most games won't see that sort of increase, for example, as they usually depend on a balance of CPU and video card performance. A program that is almost entirely CPU dependent, though, can easily see *near* perfect scaling from clock speed. I say nearly because yes, 20% increase in clock may only give like 18% increase in speed or something... but the idea is there: faster clock speeds equal higher performance in CPU limited situations.

Posted on 2015-09-01 05:50:44
Domaldel

There's also the factor that clock speed can influence the actual IPC of the CPU with different CPUs having different sweet spots where you get the most performance gain pr GHz based on what frequency it's designed to run at.
The difference isn't really big enough to notice for most people, but it's still there, and yes I know you said "near"...
But I just wanted to point out that it's not *just* other system components bottle-necking the system.

(There's also the sweet spot for the frequency pr watt of course, that's the one that usually decides a chips frequency, but that's a slightly different matter)

Posted on 2016-08-26 11:27:50
Mark Haus

Sorry, but how is that? IPC is the measure of Instructions Per Clock, by definition it is the measure of how many instructions get executed for every clock cycle. Shortening the execution cycle on the same ISA just means you perform the same instructions, in the same sequences proportionally faster and thus won't affect the architecture's IPC. That's why that metric exists, as a way to isolate performance from system clock speeds.

Posted on 2016-09-21 18:07:38
Domaldel

The problem is just that the real world isn't that perfect.
Individual pipeline stages in a super scalar architecture can't always handle the speeds equally well.
You tend to actually get a higher IPC at lower clock speeds because everything works as it should, while at higher clock speeds some parts of the CPU sometimes can't quite handle the frequency and can hold back other components.
Also the IPC changes based on what software you're running on the chip.
Different instructions takes different amounts of time to execute.
It's not a fixed value for a CPU but changes with the circumstances.

Posted on 2016-09-22 09:15:46
Léo Lam

The [Dolphin emulator](https://dolphin-emu.org/) for example?

Posted on 2015-12-16 20:11:34
xostrowx1991

I would note that you don't know that and to claim to is misleading at best. In fact many others working in the industry and even surveys/studies etc.. have said that people with "E" series enthusiast-class i7 CPU's overclock at a far higher rate than people using K series 1155/1151 etc.. consumer i7's do; and even go so far as to say that more people overclock than not with these SKU's.

And to further the mis-information from this article, you are acting like people are afraid of overclocking, acting like a mild overclock is dangerous or something? Intel themselves have said that Haswell CPU's are quite safe within any reasonable voltage level, and the widely considered safe zone goes to around 1.4v for core voltage and 2v for input voltage; I've run two 5820K's and a 5960X and not a hint of any kind of stability issues. Just pump input voltage to 1.95v, raise core voltage to 1.3v and you have an instant STABLE 4.5ghz on the vast majority of 5820K's (and my 5960X hits that same 4.5ghz at 1.3v, others may vary due to the "lottery" of course. However the 5820K being a 6 core is almost ALWAYS going to hit 4.5ghz stably without a ton of voltage.) And the thing that most outs your statements as untrue is the fact that an unstable overclock (as long as you aren't a complete fool trying to go straight to 5ghz with 1.6v core voltage etc..) will simply result in a blue screen crash which you just recover from by hitting F2 on reboot and resetting to defaults or trying a lower overclock. There's literally almost NO danger to overclocking anymore, sans higher voltage having the possibility of slightly shortening the overall lifespan. And if that's a concern you can simply go for a 4ghz or 4.2ghz etc.. overclock which requires little to no voltage increase if any. We already know that Haswell architecture can 100% perfectly and stably handle a 4ghz constant frequency otherwise Intel wouldn't have released the i7 4790K which runs at a base 4ghz and 4.4ghz boost; you know that if there was even the SLIGHTEST hint of lifespan problems at that range they would've tanked that 4.4ghz boost concept and dropped down to the 3.5ghz stock / 3.8ghz boost Ivy-Bridge levels.

So all in all, this comes off as either you guys just don't have nearly as much experience with Consumer/Enthusiast CPU's as thought, or having A) An irrational fear of overclocking, or B) An irrational bias towards Xeons. The truth of the matter is that the i7 5960X is just a 10 core Xeon that was marketed as an 8 core i7 for some reason (the most likely scenario is low core yield; i.e. only 8 or 9 of the 10 cores functioned right so they disabled the remaining 2 cores and left 8 good ones so they could sell it as an 8 core i7 rather than losing money by scrapping the whole silicon. This 100% confirms your statement that there's really NO difference in "speed" between them, clock for clock, but once overclocked an i7 5960X SMOKES an E5 whether we're talking productivity (i.e. video editing, 3d modeling etc..), Server work, high-tier gaming, or Physics engines and Supercomputers etc..

Also, some of the other Xeon advantages are incorrect as well. For example, there are nearly ZERO reasons to need 768GB of RAM. And the i7 5820K/5930K/5960X can ACTUALLY use 128GB of RAM, and are not limited to 64GB like was thought on the initial release. This has been known for some time so i'm kinda surprised you didn't update the article; assuming it wasn't already changed before the article released. And again, 128GB of RAM is more than anyone really needs unless you are such a perfectionist that having two Samsung 950 Pro NVMe 2,000Mbps+ sequential speed drives in RAID 0 is not fast enough for you and you HAVE to have a 128MB RAMDISK cache and still need another 32-128GB of extra RAM or something. (so again, nearly nobody....)

Posted on 2016-05-05 16:29:19

You are correct that raising the clock speed will increase performance in almost any application (unless the performance is bottlenecked by another component in the system). And you are also correct that modest overclocks don't necessarily risk stability all that much. And in fact, we have found that some of the Xeon E5 series processors - the ones which have an exact Core i7 equivalent, like the E5-1650 v3 = i7 5930K - can be overclocked like their i7 brethren.

However, there are still downsides:

- As you mentioned, potentially decreased lifespan. This may not matter for someone who is planning to replace the computer after only 1-3 years, but not everyone is on such a fast upgrade cycle.

- The potential for instability. Even a single crash of a program or the OS can be disastrous in some fields, like with folks running servers or computations that last days or even weeks. Instability can also mean incorrect results from calculations, even if a program doesn't crash outright. No matter how low the increased risk of instability is, if there is *any* increase then I would not advise it for a wide range of our users. For others, though, the added performance may well outweigh any risks in this area... it really depends on your situation.

- Added heat. Overclocking increases the heat output of a processor, which means that more powerful cooling is needed... and that, in turn, means an increase in noise level. For some this isn't a big deal, of course, but for many of our customers a very quiet or nearly silent computer is part of their goal. We can achieve that with stock speeds on processors, but the added cooling requirements of overclocking necessitate some amount of additional noise.

Now with those things said, we do offer overclocking when folks want it! And I'm sure some of our customers overclock on their own, though I would (and do) strongly recommend they let us handle it so that the overclocking is covered under our warranty. But I looked over our records, and in the past year only between 2-3% of systems we sold included overclocking. If you factored out the systems with processors incapable of being overclocked the percentage would be higher, but it is still quite a small minority of our overall sales. That may be different for other system builders, especially those more focused on the gaming market or liquid-cooling.

Posted on 2016-05-05 16:53:54

Where are you pulling these figures from? youropinion.com? You 'obviously' have zero experience dealing with server/workstation class computing or you wouldn't have made such a fool of yourself. You'll learn, one day.

Posted on 2016-05-19 16:25:27
xostrowx1991

Your arrogance doesn't become you, you know. i've been building workstation pc's and servers for years and have seen every generation of Xeon all the way back to 90nm back when i was fixing up the crappy nodes at my university in my teen years. Fact of the matter is everything i said is 100% true. Of course puget's personal fanboy squad won't let petty things like facts get in the way though.

Pretty much every feasible situation other than PURE proffesional grade server workloads will see near identical if not better results using i7 CPUs compared to Haswell-EP V3 Xeons. On haswell and broadwell it's generally accurate that ~400-500mhz extra clock speed is roughly equal to one additional core. This is evident by how the i7 5930K and i7 6850K both easily hit the same performance in 100% load render testing such as Cinebench or even things like Linpack etc.. with, for example, an i7 6800K hitting 1,350 points at ~4.4ghz which is identical to a 6900K and 5960X 8 core chips at it's stock 3.5ghz/4.7ghz respective turbo speed (which is a clock speed difference of 700mhz, even that small increase of 0.7ghz gives a 2 core effective increase). The same is shown comparing the 5960X and 6900K to the 10 core 6950X; the 5960X/6900K at their common max overclock levels of 4.4ghz and 4.6ghz, get roughly ~1,750 points in Cinebench 15 which is exactly what a 6950X at 3.5ghz gets.

This means that an i7 5960X at the top of its class bin category (it isn't that uncommon to see 5960X at ~4.75ghz or 4.8ghz in the hands of enthusiasts or OEM dealers etc..) is perfectly capable of keeping up with quite a few workstation class Xeons. Haswell-EP E5 Xeons are limited to 18 cores and pretty much every one with over 14 cores has a huge disadvantage in clock speed; most not even reaching 3ghz with all core turbo. Take for instance the 3.2ghz E5 2667 V3, an 8 core 2S capable chip; even in dual CPU config you have 16 cores at 3.2ghz (3.4 all core turbo), compare that to a 5960X at 4.8ghz and you see that the 5960X has a "4 core advantage" over it due to a 1.6ghz clock speed increase; meaning the 5960X would perform as fast as a 12 core 3.2ghz chip and this is only an 8 core. Even with two of them you only have 16 cores, which is only four more than the 5960X's "effective" performance is....and costs FOUR times more than the 5960X!! (not including the extra cost of the dual socket C612 chipset board)

Look, don't get me wrong here. I think Xeons have their place, and CAN be a better choice in some cases; but specifically the haswell Xeons are just not worth it except in EXTREMELY intensive situations like medical imaging and nuclear isotope work etc.. for things like video rendering in Premiere, 3d modeling, encoding/transcoding, etc..etc.. you're often better off with a haswell i7 than a haswell Xeon; ESPECIALLY if you use programs and workloads that benefit more from clock speed like Virtualization, Virtual Reality development and game development in general etc.. Interestingly, BROADWELL-EP actually changes things up a bit, disproving some of what i've said. There's chips like the E5 2687W V4 etc.. that give 12, 14 etc.. cores per chip and give it at ~3.5ghz in many cases; that combined with Broadwell-E i7's having a general overclock limit of ~4.3-4.4ghz on average means that Xeons can DEFINITELY be a great choice for the discerning workstation user. Two 2687W V4's givs you 24 cores 48 threads at 3.2ghz all core turbo that can be had for only ~$500-600 more than the cost of two 6950X 10 core BW-E chips; and in this case, the 6950X only gives you an "effective" power of ~13 cores via it's overclock, which unlike the Haswell comparison we are getting 11 extra "effective cores" with the dual xeons on the broadwell instead of only 6 more effective cores.

Posted on 2016-09-14 05:08:07
disqus_pgsT8viUhk

what would be the best processor to get for using software that solves poker?* currently I'm using a middle-of-the-road i7 and I was thinking of getting the 6950x, but software developer is telling me that would still take a bit of time. He says more cores will solve flops faster.

which leads me to xeons cpus. Not sure if that would be the best? I was always under the impression they were not used for computational programs but for media editing or servers.

*the program uses an algorithm designed to solve imperfect information games.

Posted on 2016-12-29 00:32:02
Kotlos Kotlos

There are reports that the E5-1660 v3 & e5-1680 v3 are unlocked, can you verify this?

Posted on 2015-06-12 19:21:52
Chucrute

All of the E5-1650, 1660 and 1680 are unlocked, V1, v2 and V3. K|INGP|N set the E5-1660 V3 to 5.8GHz. http://hwbot.org/submission...

Posted on 2016-02-04 07:01:03
JeffreyQC

So...we can overclock a E5-1650 V3 like a i7 5930K right?? Same multiplier limit? I need this info plz,i am ordering a E5-1650 V3, i want to be sure is like the 5930K, but more robust. (better binned) :) Thx

Posted on 2016-03-25 16:35:42

I don't know that it is better binned, but the Xeon E5-16XX processors are unlocked like their Core i7 counterparts.

Posted on 2016-03-25 16:39:35
RonWade

Is Xeon amd64 compatible? Does it matter?

Posted on 2015-06-22 01:31:05

Intel has a very similar instruction set call "Intel 64". For normal users they are effectively the same / compatible, but there are some small differences which could impact niche groups. Wikipedia covers it well in this article: https://en.wikipedia.org/wi...

Posted on 2015-06-22 06:00:03
Bavani Sankar

Lol, the Itanium flop! Intel tried to being 64bit tech but AMD did that better!

Posted on 2015-08-01 03:40:01
wirk

This article has nothing to do with reality and should be corrected. Enough to say I am typing this on machine with Xeon E5-1680 v3 OVERCLOCKED to 4 GHz and with 128 GB ECC RAM in Asus X99-E WS motherboard. Xeon single socket processors DO support overclocking. Some X99 motherboards DO support 128 GB ECC RAM.

Posted on 2015-07-13 16:36:04
Nigel

Dear wirk,
I'm intrigued by your proposal. I'm on the trail of a new workstation for 3D animation and video editing and would love to see the specs of the machine you mention above. I've built a few systems before and overclocked, with help from an enthusiastic friend but he's moved interstate and I'm a bit hesitant to dive in alone because I remember the myriad settings in the bios and feared bricking my machine.

In my new machine I'm looking for upwards of 16 GB RAM and a processor targeting single thread tasks because, apart from software rendering, most of the work I do in 3D is single threaded. I'm also wanting Thunderbolt 3 capablility which is, I believe due out next month.
Best wishes
3Dack

Posted on 2015-07-23 02:37:45
wirk

Hi,
Since you are in single threading you need processor with less cores (4 is OK and 8 would make no sense) but which is overclocking very well. Here there is very, very exciting news: Intel Skylake processors coming in August 2015 are really good in this respect judging from first tests. Look at them, they are better than anything else. BTW, if you are in 3D and video the best point to start is 32" 4K monitor, I have one and it is really satisfying.

Posted on 2015-07-23 11:19:45
Nigel

A quick Google on Skylake processor range turned up a list of ten variants. No xeons on the list but two 'unlocked' chips, the Core i7-6700K with 4 cores hyperthreaded I think and a Core i5-6600K with for cores and no threading. I believe some of the new Skylake motherboards will also feature Thunderbird 3 which for 4K video is my other wish.

If you have any tips on where to learn to safely overclock the new processors I'd be grateful.

Thanks heaps,

Posted on 2015-07-23 23:34:18
wirk

Go with the best of the lineup, i7-6700K. For details on overclocking one has to wait when the processors and mobos are released. Thunderbolt 3 is supported and integrated graphics will be fine for 4K, it remains to be seen what is its performance for 3D but it can be OK for basic work.

Posted on 2015-07-24 06:06:04
Nigel

Thanks again,

Holding breath for the release of the boards.

Nigel

Posted on 2015-07-24 10:07:25
xostrowx1991

Except he said he also does Video Editing, which is a core/thread hungry task! He would be much better compromise between the 4 core 8 thread like the Skylake 6700K and the 8 core 16 thread i7 5960X etc.. and just getting something in the middle like the i7 5820K which is a 6 core 12 thread CPU with full overclocking so its single thread performance once overclocked is right up there with the 6700K (basically identical) but it excels much better in multi-thread performance due to having 4 more threads.

Plus, the i7 6700K is about $350, and the i7 5820K is only $375, so virtually the same price. And the 5820K also supports the same DDR4 RAM etc.. so you aren't on an "old platform" or anything. Not to mention you also get other advantages with the 5820K that you don't with the 6700K: for instance you have much better PCI SSD support due to more and/or better M.2 support etc.. you have 28 PCIe lanes compared to 16 on the 6700K so you can do proper SLI support for 3 graphic cards whereas you can only use 2 at most on the 6700K and if you did use 2 you can't use ANY other PCI cards like a wi-fi card or PCI SSD etc.. whereas in the 5820K you can have 2 graphic cards in SLI and STILL have 12 more PCI lanes left for other cards; hell you can even have THREE graphic cards and still have 4 more lanes left-over! And if you get an X99 motherboard with a PLX chip in it for your 5820K it will give you a total of 32 lanes, allowing FOUR way SLI graphic cards to be used!

Posted on 2016-05-05 16:56:34
Przemysław Zachara

Could you clarify what model exactly yoy have? Es,qs ,oem,box? Cpu-z screen would be helpfull here too.

Posted on 2016-01-01 14:29:59
wirk

You can see the processor data in Cpu-z and the box in which the processor was packed. Wystarczy?

Posted on 2016-01-01 15:45:32
Przemysław Zachara

Ok, great. Hwinfo would br great here too for BOX vs ES/QS compare.

Posted on 2016-01-01 16:39:49
wirk

I have box version bought through normal channel, not ES/QS. I doubt if there is any difference between them. There are e.g. discussions about differences about Xeon E5 and i7-5960X, saying that Xeons are 'better' but to me these are urban legends. Xeons are supporting ECC RAM memory while i7-5960X not and that's it. I am running Xeon E5-1680 v3 with 128GB ECC RAM in Asus X99-E WS motherboard.

Posted on 2016-01-03 11:28:15
Przemysław Zachara

Situation is like this
- ES / QS can have different multiper, 32; 34;or 35 settings, while they have same base max MHz frequency at 3 or 3.2 GHz.
- they can have idfferent turbo settings for core/cores
- they can have different stepping and revision too.
Thats why i asked about hwinfo screen... :)

Posted on 2016-01-03 15:54:22
wirk

In any case, my box Xeon overclocks well, 4.3 GHz all cores, 4.5 GHz single core and that possibly could be increased by 0.1-0.2 GHz with very sophisticated watercooling. My present watercooler is full cover block for
the X99-E WS motherboard from liquidextasy.de. It is one big block which cools the chipset, voltage regulators and processor. But I have not used very special water pump and radiator, just a standard one.

Posted on 2016-01-03 19:11:59
goblin072 .

Trouble with that set up is you get no ECC reporting. You have no idea if module in slot 3 is having higher than normal single bit corrections. You just hope its working.

Posted on 2016-02-19 21:10:09
xostrowx1991

Actually no. It's only the E5-1650,1660, and 1680 Xeon models that are unlocked, and they are a bit more limited than i7 CPU's so you won't get as high of an overclock, possibly as evidenced by your 4ghz compared to most i7 users having 4.5ghz or more. And you are a bit more limited in motherboard choices with a Xeon. The one thing you are correct about is the 128GB RAM support, but it's got nothing to do with ECC; X99 boards and i7 5XXX CPU's can support 128GB of NORMAL DDR4 RAM. ECC will have you limited in board choice of course like i mentioned, but there are boards that will take it. It's just that ECC has nothing to do with whether you can use 128GB or not.

Posted on 2016-05-05 16:48:29

It may not be advertised to work, but every X99 board we have tested works with DDR4 Registered ECC memory - when using either Core i7 or Xeon processors. Intel certainly doesn't market either X99 or Core i7s as being compatible with ECC, most motherboards are not listed by manufacturers as supporting it, and the BIOS on X99 boards doesn't usually support ECC error logging like some server boards do - but the fundamental functionality of compensating for single-bit memory errors works great.

Posted on 2016-05-05 16:58:00
Guest

Any idea how to get Premiere to run faster than molasses o a dual E5-2630 with 128g ram and Titan X? It only peaksat 3% CPU use and drops most or the frames.

Posted on 2015-08-08 02:37:54

That sounds really strange - like something is not configured correctly somewhere, maybe in Premiere or maybe at the hardware / driver level. If this is a system you purchased from us, I would recommend calling our support department next week (during business hours)... or if not, maybe check with whoever did build your system.

I suppose it could also be related to a bottleneck in the storage area. What sort of drive are you editing from?

Posted on 2015-08-08 20:54:17
Guest

The new Samsung 2TB SSDs. OS is on a Crucial 240GB SSD. Scratch drives are Samsung SSD. All drives test at over 500MB/sec SSDbench.
I built the system.. consists of Supermicro X10DRi, two E5-2630 v3 CPUs with Noctua coolers, 128GB Samsung ECC RAM in four banks per CPU (16GB DIMMs), Titan X 12GB GPU and EVGA 1000 watt PSU.
I've gotten Premiere memory usage down from 82.1GB to 19.7GB after updating to the latest nVidia driver. Playback is a touch better. There is a stutter at the start of XAVC, almost like things need to spool up, and then it doesn't drop any additional frames. But multicamera edit of four 4K XAVC streams, is unusable. Switching to 1/8 resolution does not make any perceivable improvement in playback. At full resolution it's the same.. BUT... if I scrub the timeline with the mouse, I get fluid smooth playback at full resolution on all camera angles and on the 4K monitor. If I hit play, I get a series of 1 still frame per second approximately in a freeze-play-freeze-play sort of fashion. Scrubbing is totally smooth and fluid. Puzzling!

Posted on 2015-08-08 22:05:16

Wow, four 4K streams at the same time? That might be part of the issue - that sounds like a lot for the system to deal with. However, I have only dabbled in video work myself, and not with multi-cam stuff. You'd probably be better off checking with a forum that has lots of heavy Premiere users to see if there might be something with the software or workflow that isn't ideal. Have you tried someplace like CreativeCow?

Posted on 2015-08-09 04:59:13
Guest

It's an XAVC issue, for certain. Prores 422 plays smooth, even 4 streams in multicam. The decoding of XAVC seems to be the bottleneck.
I used to be on Creative Cow, but the mods banned me when I exposed the audio flaw in the Sony HVR-V1U. I suppose they were worried about upsetting their sponsors.

Posted on 2015-08-09 11:29:20
Guest

Today, I got a clue as to what might be causing this hesitation. It's not XAVC. It's happening with xdcam and prores too, but only on cc2015.
My old core2quad, after disabling Lumetri scopes, can play 4k pretty smooth with no startup delay. On cc2015.
My dual Xeon drops about 100 frames at start of play.
Cs6 does not have any delay, though no LUTs or XAVC.
Today, I was copying some files to a new SSD drive in the Xeon system and none of the video storage drives were involved in the i\o operation. However, cc2015 could not play any video. It was all locked up while the file copying was going on on different SATA channel. I had to stop\start six times before it would begin to play at all and with many dropped frames.
My old c2q system from 2007 can multitask while playing prores UHD footage in cc2015 without dropping a single frame.
This has me thinking there's something going on with the low level SATA drivers and south bridge controller. But I don't know what it could be.

Posted on 2015-09-23 23:35:44
John

I see Intel has some sort of cache "hotfix" for their XEON CPU family - can someone tell us the cache problems in XEON?

Posted on 2016-01-06 00:47:58
TheDude

Hey guys! I've heard that it is possible to turn on Turbo Mode for one core on XEON CPUs to run permanently on all cores in previous generation X79 chipset. So it is an awesome overclock. And here I have question is it possible to do this in X99?

Posted on 2015-11-12 11:24:50
Tamás Katona

Asus x99 deluxe support Xeon Phi?

Posted on 2015-12-11 19:35:39

Yes, it should support Xeon Phi. We typically use the X99-E WS in our Peak systems (https://www.pugetsystems.co... that include Phi support, but I believe the X99 Deluxe has the same support needed. You might want to read https://www.pugetsystems.co... which has a section about what motherboards need for Phi.

Posted on 2015-12-11 19:43:20
Tamás Katona

Thank you. I read it and checked the bios. "Above 4g decoding" is available on Asus x99 pro motherboard too.

So in that case can asus x99 pro be enough? It is very cheap x99 motherboard.

Posted on 2015-12-11 19:50:47

As long as it has Above 4G Decoding, it should work fine with Phi. That is really the one requirement for Phi beside needing an x16 PCI-E slot to install the card into. A cheaper board tends to have it's own problems (many Phi users want IPMI or dual NIC), but if the board has all the features you need it shouldn't be a problem. We haven't actually used the X99 Pro with Phi, so there is the chance there is some unknown issue, but everything on paper says it will work.

Posted on 2015-12-11 20:13:19
Tamás Katona

Have you ever tried Xeon Phi with Core i7?

Posted on 2015-12-12 17:24:26
Donald Kinghorn

Hi Tamás,

What Matt has said is correct. The main thing is haveing a motherboard with large BAR's i.e. Above 4G decoding. I have run a Phi on a system with a i7 4770 on an ASUS board with a custom BIOS.

Posted on 2015-12-15 00:01:15
strangemonkey

I just got a HP Z420 with an E5-1650v2 from ebay and found your page - it replaces a 7 year old Gateway lx6810-01 / Q8200 2.33Ghz Core 2 Quad - which had heating issues, after cleaning it up / dusting it out, and reassembling recently, it never started again - I think I messed up the CPU holder pins during cleanup. Was a decent enough machine for 7 years! The HP Z420 feels 500x faster and has 16GB DDR ECC vs 8G on the Gateway, it appears overall to be much higher quality, and in a different class of machine entirely - chosen in part for its on-board 1394 controller - as my film scanner requires legacy FireWire 400. Probably overkill for me doing 2D graphics / photography, but it's definitely been a joy to use so far.

I'm wondering what Core or Desktop Intel chip the E5-1650v2 most closely compares to? Is it the Core i7 this article mentions? What is the equivalent desktop chip, e.g. i3, i5, i7 ... What are the main differences, I think power consumption is a big one?

Thanks,
SM

Posted on 2015-12-31 17:02:10

The E5-1650 V2 ( https://www.pugetsystems.co... ) is basically a copy of the Intel Core i7 4930K ( https://www.pugetsystems.co... ) only with a .1GHz higher base clock. They have the same maximum Turbo Boost, however, so in most situations the performance would be identical. They have the same numer of cores, cache, wattage, and PCI-E Lanes.

The main difference is that the E5-1650 V2 supports ECC (and Reg. ECC) memory, while the i7 4930K does not. In addition, the E5 supports VPro and TXT which is mostly used for server environments.

Posted on 2015-12-31 18:33:40
Chucrute

Better to say that the 4930K is a copy of the Xeon since all the Intel Core series are designed after Xeons . Also you should correct your article as ALL the Xeons E5-1650, E5-1660 and E5-1680 support overclock, no matter the version.

Posted on 2016-02-04 06:56:00
Gordon James Taylor

5820k ;-)

Posted on 2016-09-25 21:30:11
Juan Gutierrez

For work with CAD software...what can I expect from dual xeon e5 2620 v3's and a quadro k620, should I go for an i7 5xxx instead and or a gefore card, or maybe something completely different what would you recommend? I will be taking a course in 3D design next semester, using Autodesk Inventor to be specific, and I would like to know what is an optimal build for a vast amount of projects, I want to be able to take projects home and continue to work on them but also be able to create seperate and maybe even larger projects in my spare time

Posted on 2016-01-01 09:38:29

For most CAD programs (including AutoCAD, Inventor, and Solidworks) dual Xeons are not going to be very good. CAD programs are almost entirely single threaded, so having the latest generation CPU with the highest frequency is much more important than having a higher core count. In fact, we've found that having two CPUs will sometimes give lower performance than a single CPU due to the additional overhead associated with having two CPUs. The only exception we have found is rendering which can really take advantage of as many cores as you can throw at it.

We haven't tested Inventor specifically, but we do have recommended systems for AutoCAD (https://www.pugetsystems.co... and Solidworks (https://www.pugetsystems.co.... Both are very similar, so I would look over the "Hardware Recommendations" tab for both of those. Basically, you want the highest frequency quad core CPU you can afford (Core i7 6700K right now is the best), 16GB of RAM, and a SSD drive. For the video card, since you are using Inventor I would recommend doing a GeForce GTX 950/960/970 depending on your budget. AutoDesk products seem to run great with GeForce cards - although if you plan on using Solidworks in the future Solidworks does see pretty big performance drops with GeForce cards compared to Quadro.

Posted on 2016-01-04 20:06:59
Juan Gutierrez

Matt, so for my use it looks like a core series chip is a better option and for autocad a geforce card is also the better option, thanks so much for the insight, I feel a lot better now knowing I won't be overspending on something like dual processor set up, cheers, happy new year

Posted on 2016-01-05 14:00:19

Yep, that is pretty much the best way to go for AutoCAD. Glad I could help out!

Posted on 2016-01-05 17:39:35
John

Any plans to run Premiere exporting 4K video to H.264 benchmark?

Posted on 2016-01-06 00:43:41

Do you mean like this?

https://www.pugetsystems.co...

That shows that, at least in our test sample, such exporting scaled well up to around 5-7 cores. So a six core CPU at high clock speed should be optimal; any more cores, and the clock speed would usually drop enough to result in worse performance. That is true regardless of Core i7 vs Xeon branding, as the two perform nearly identically (with a couple %) in other tests.

Posted on 2016-01-06 00:48:38
John

I saw that report, how about with GPU encoding support in Adobe 2015, do we need as much CPU cores?
Some of the thing I like to see Xeon and AMD R9 295x2 vs Nvidia GTX980 or faster. Some of the plug ins are GPU and CPU intensive.

Posted on 2016-01-06 01:03:57

We can't test every component out there, but we have done some comparisons between various GPUs in Premiere:

https://www.pugetsystems.co...

https://www.pugetsystems.co...

Posted on 2016-01-06 01:08:03
John

Thanks - but are those 1080p video files or 4K files - We're seeing system slowness with 4K video throwing at GPU to do the processing.

Posted on 2016-01-06 01:17:40

I'm not sure on the first article - it is a few years old, and likely just 1080P because of that. The newer article (the second link) shows both 1080P and 4K results, and interestingly the 4K results show the choice of video card having much more impact on performance than at 1080P. Still, there are so many variables (original and target resolution and format, just to name four) that it is impossible to test every potential situation. The main take-away I have is that the higher resolution your work in Premiere, the more a fast video card will help... and potentially dual video cards as well.

Posted on 2016-01-06 01:23:43
John

A lot of things become antiquate overnight with 4K video !! including software, and work flow.
Can't wait for 8K or 6K.

Posted on 2016-01-06 01:31:57
Carlos Díez

Hi,

I don't know if this question is simple or not, but i've not found the answer so far:

Is is possible to replace an i7-6700K with a Xeon E5-2620 and vice versa?

Posted on 2016-01-11 20:14:27

No, those processors fit two very different sockets - and therefore need different motherboards. The 6700K is for socket 1151, while the Xeon E5-2620 V3 uses socket 2011-3.

Posted on 2016-01-11 20:17:29
Carlos Díez

Thank you!

Posted on 2016-01-11 21:24:38
Chucrute

All of the E5-1650 Vx, E5-1660 Vx and E5-1680 Vx do support overclocking. What about correcting the article ?

Posted on 2016-02-04 06:53:58
Jakewa2

I'm totally confused between cores vs frequency, when comparing the E5v4-2620 to E3v5-1275. Will the E5 really give me better performance?

Posted on 2016-04-05 18:05:03

That depends on your application and scalability. The E5 v4 2620 is an 8-core processor at 2.1GHz base speed (3GHz max, when fewer cores are running). The E3 v5 1275 is a 4-core processor at 3.6GHz base speed (4GHz max, when fewer cores are running). That means that for applications using 4 or fewer cores, the 1275 is definitely faster - thanks to a higher clock speed. In situations where more cores can be used effectively, the E5 could be better... but it depends on how good that scaling is.

A couple of other points:

- The E5 2000 series processors can be used in pairs, provided the right motherboard.

- The E5 also supports more memory, so if you need >64GB then it would have a clear advantage over the E3 series.

Posted on 2016-04-05 18:12:42
Jesse

Okay I read on this link of puget systems that the x99 deluxe motherboard only supports up to 128 gigs of ram https://www.pugetsystems.co...

& Pc part picker says it only supports 64gigs https://ca.pcpartpicker.com...

But this article says it supports 256? What is the right answer, could you please clear this up.

Also how would these asus boards listed below fit into this. They are all x99 boards would they still only support 64gigs like it says on pc part picker or would it be 128, or 256. I would be using a Intel Xeon E5-1620 V3 3.5GHz Quad-Core Processor or similar xeon processor.

Asus X99-A/USB 3.1
or Asus SABERTOOTH X99
or Asus X99-PRO/USB 3.1

Posted on 2016-05-20 04:35:07

The original 64GB listing is because when the X99 chipset came out the largest non-ECC memory modules were 8GB. With 8 slots maximum, that meant 8 x 8GB = 64GB maximum. 16GB modules are out now, so places that have updated their info show 8 x 16GB = 128GB maximum.

However, ECC Registered DDR4 memory is available in even larger sizes! And these large ECC RDIMMs work well in the X99 motherboards. It is simply a feature / functionality that Asus (and Intel) don't really advertise. Check out this screenshot from a build with the Xeon E5-1650 v3 and 8 x 32GB memory modules: https://www.pugetsystems.co...

Posted on 2016-05-20 04:56:49
Jesse

Thank you for clearing this up! Does the ram need to be registered to get 128gigs to work?

I know puget systems uses the asus deluxe board. Do you know if the x99 sabertooth or asus pro board would work with registered ecc or unregistered ecc to get 128gigs or more gigs of ram? Price wise these seem like better options. Would love your opinion on these boards. You guys are the best for reliable information.

Posted on 2016-05-24 07:14:38

Now that normal (non ECC / Registered) DDR4 memory is available in 16GB sizes you should be able to do 128GB without worrying about that. To go above 128GB, you need modules which are larger than 16GB - and so far those are only available in ECC Registered modules.

As for other motherboards, the normal RAM up to 128GB should work on any X99 board with 8 RAM slots so far as I am aware. ECC Registered memory has worked on all the boards we have tried, but the Sabertooth and Pro variants are not models we carry... so I can't say for sure.

Posted on 2016-05-24 15:50:59
Ali Rafaat

hello ...thx for this article ,,, but would someone tell me which is better for a programm like 3dsmax ? or for rendering ?

when you have 64 0r 32 ram which is better to use ?

Posted on 2016-04-08 13:04:54
Guillaume Goa

0.5% on meaningless benchmarks and considering the price of the x99 motherboards, that is useless, it is a complete marketing bullshit. Go for an i5/i7 or
wait for the AMD Zen :)

For domestic use, people do not need ECC ram.

Posted on 2016-04-15 20:15:03
Guillaume Goa

0.5% on meaningless benchmarks, considering the prices of the X99 motherboards, xeon is a bad choice for domestic use. A complete marketing bullshit.
Go for an i5/i7 or wait for the AMD Zen :)

People do not need ECC Ram for domestic use.

Posted on 2016-04-16 08:35:59
WesternGentsUnited

Can you use non ECC memory with the Intel Xeon E5 2630 v4 on a normal motherboard like above and does everything above apply for the new v4 E5?

Thanks Champs

Posted on 2016-06-24 05:37:32

You should be able to use a 2630 v4 on a normal x99 motherboard, but I would double check the list of compatible CPUs for the specific model you are considering. I haven't seen any x99 not work with the new E5 v4 CPUs (although it often requires a bios update) but it is anyways good to check. And yes, the Xeons will work fine with standard ddr4 RAM.

Most of the general information in this article is still relevant, although the specifics like the number of cores and frequency for the different product lines is obviously no longer accurate. The big thing is that Core i7 and Xeon E5 should still have the same performance clock-per-clock.

The only odd thing with the new models is that the specs don't line up as neatky between the Core i7 and Xeon E5 single CPU (1600 series) processors. For example, there is a 10 core i7, but only an 8 core Xeon E5-1600. You can get 10 cores with a dual Xeon CPUs, but another change is that the dual CPU models don't support the new turbo boost 3.0 - so single theaded performance isn't quite as good.

Posted on 2016-06-24 05:57:06
Minh Hoang

Hi, I have a E5-2667 V3 and ASUS X99 PRO mobo. My question is can I use non ECC memory at higher clock than 2133Mhz with this build? For example can I install and use RAMs at 2400Mhz or higher clock? How will it affect the components? Thanks! :)

Posted on 2016-06-27 16:06:59
Chris Fromgogle

I want to point out two big things for me that I am not seeing much discussion of:

1. Xeon has better PCIe lane count. I run hardware RAID SSD's (LSI 9271-8i) along with video (GTX 1070) and 10Gb Ethernet and I'd like to run some U.2 or M.2 drives, so more lanes is a huge selling point for me. PLX is a workaround so you can share the lanes but Xeon just has more lanes so no need to share.

2. Memory bandwidth is a lot better on the E5 than the i7. Amount of RAM available is bigger too if you need more than 64GB. I presume the tests run above had the RAM dispersed so all 4 channels were in use on the E5 and not running on just 2 channels.

The v4 E5's are out now too.

Check some of the stats in this comparative link: http://ark.intel.com/compar...

If you do any virtualization or other potentially heavy memory intensive tasks, the E5 should win even over overclocking i7. Probably not a noticeable difference between E5 and i7 in most games, but in development or in multiple OS virtualization you should...I imagine you'd notice a big difference in larger photo editing as well.

Posted on 2016-09-28 22:28:39

This article is about a generation old now, so there are a few things that are different in the current models. However, it looks like you might be comparing across multiple product families - your Ark link includes some standard Core i7 CPUs which are equivalent to the Xeon E3 family, not the Xeon E5. If you look at just the "High End Desktop" processors, the specs for Core i7 are much closer to the Xeon E5 CPUs. This is the product comparison that I would use to compare Xeon E5 to Core i7 today: http://ark.intel.com/compar...

For your two points, with the exception of the Intel Core i7 6800K, all the Xeon E5 and Core i7 CPUs support 40 PCI-E lanes. However, one thing to keep in mind is that X99 boards (which are primarily designed for Core i7, but work fine with Xeon E5 as well) tend to have more add-ons like extra USB controllers, Wifi, etc. built in which might use up some of those lanes. However, you can use COre i7 perfectly fine on server boards as well, so from a purely CPU perspective the PCI-E lanes are the same across the two product lines. As for the memory bandwidth, they should all support up to DDR4-2400 across 4 channels so unless I am missing something the bandwidth should be the same regardless of CPU. The Core i7 CPUs do only officially support a total RAM capacity of 128GB of RAM, but you can actually do up to 512GB and it works great. We typically don't sell more than 128GB of RAM max with those CPUs because users who need more RAM tend to need dual Xeon setups, but I do happen to be doing some testing with a Core i7 6950X (using an Asus X99 Deluxe II) with 256GB of RAM that I could grab a screenshot from: https://www.pugetsystems.co...

With the current CPUs, really the only advantage Xeon E5 V4 CPUs is that they support vPro and can use load reduced RAM if you need even more than 512GB of RAM. On the other hand, the Core i7 CPUs support Turbo Boost 3.0 (which is actually pretty significant, especially for lightly threaded tasks), are available in up to 10 cores (single socket Xeon E5 V4 only goes up to 8), are a hair cheaper, and support overclocking.

Posted on 2016-09-29 00:25:44
eugenia

I have no expertise in pcs but I have to ask and I would appreciate if you could answer. I am looking to buy a good pc for my studies. I am an architecture student and I use software like AutoCad, ArchiCad, Photoshop and Sketch-up. Which of the two CPus should I go for?

Posted on 2016-11-01 22:05:41
Hloan Gaming

xeon e5

Posted on 2018-01-04 22:11:04
BC Shelby

...nice eveluation. I didn't realise that the i7 family tops out at 64 GB. Looking at building a 3D graphics workstation with dual 8 core Xeon 2630s and 128GB (4 x 32 GB) eventually expanding to 256 GB (8 x 32 GB) to handle large rendering loads. GPU rendering while "fast" is still very limited memory wise unless you go for a couple Quadro P6000s.

Posted on 2016-11-03 18:30:04
rarmo

Interesting. I would like to see the results from open world or big sandbox games such as GTA 5, XPlane11 and Far Cry (Crysis' engine) - Certainly, XPlane may use individual cores for each aircraft in the environment. Crysis... well, you need a supercomputer to run that game, methinks lol

Posted on 2016-11-16 03:30:54
Hloan Gaming

may i ask what xeon version compair with i7 6800 or 6850k

Posted on 2018-01-04 22:12:55