Xeon or i7 4790k?

[Sniper82]

I been running a i5 for the last 5 years and the upgrade bug has hit me. I think I am wanting to move to a i7 and sell off my i5. I am looking at used. I can get a 1270v3 for $200, 1271v3 for $225 or a i7 4790k for $250. I am new to Xeon CPU's and only real concern is compatibility. My motherboard(z97-g55) supports all of them but I've heard there could be some memory issues with the Xeon. Is that true?

I don't plan on overclocking right now but might possibly in the future. Price wise is the i7 worth it if I never overclock?

 

[Denithor]

There is NO overclocking with Xeon chips.

Also those Xeons have no iGPU if that matters to you.

My vote would be for the i7. Higher stock clockspeed, better resell later.

 

[Sniper82]

yeah I know of the no OC and the no iGPU. I haven't used it on i5 the years I've used one. I got enough GPU's laying around I'd never need to. Good point on better resell.

 

[Denithor]

My only reason for pointing out lack of iGPU is more than once I've wished for that feature when reselling an old system so didn't have to include a dGPU.

 

[Sniper82]

yeah that does make sense. But I never sell a complete setup. I'll probably just go with a the 4790k because I am sure at some point I'll sell it. Sometimes I sell local and most around here probably wouldn't even know what a Xeon is.

 

[Mercennarius]

There is NO overclocking with Xeon chips.

Not true. Depends what board you are using. If your not going dual CPU then there are plenty of options for running an overclocked Xeon.

 

[LTC8K6]

You can almost certainly run the Xeon with all cores at it's turbo multiplier. I have a Z97 board and a 1231-V3 runs all cores at 3.8 under a heavy load with no problems.

It's basically a 4770 without the igp.

A 1271-V3 will likely run at 4.0 with no problem.

 

[freeskier93]

You can almost certainly run the Xeon with all cores at it's turbo multiplier. I have a Z97 board and a 1231-V3 runs all cores at 3.8 under a heavy load with no problems.

It's basically a 4770 without the igp.

A 1271-V3 will likely run at 4.0 with no problem.

Unless there is some magic I don't know about this is a common misconception on how turbo boost works. The 3.8 turbo boost frequency is only for ONE core. The max all cores can run at is 3.6 with the 1231.

EDIT: After a bit of reading it looks like some motherboards can bypass turbo boost and force all 4 cores to 3.8. Guess I will have to look into this with my MB...

 

[LTC8K6]

Yes I posted about it in another thread, with screenshots. Someone on Youtube was running a 4790S, I think, with all cores at the turbo multi. It was happily running them all at 4.0 with no problem.

My Xeon with all cores staying at 3.8 during IBT:

 

[Magic Carpet]

4790K has only one problem:

worse performance per watt metric, compared to lower clocked 4C/8T Haswell & 4C/8T Skylake (i.e. 6700K).

NB. At 3.1 Ghz 4C/8T Haswell consumes ~45 watts. At 4.0 Ghz it doubles it. As a result, it runs a lot hotter, needing a good aftermarket heatsink, whereas a stock Intel heatsink would be absolutely adequate for lower clocked Haswell SKUs.

But, if you are more interested in raw performance and/or integrated graphics, go with 4790K.

PS. Also, not all boards can lock turbo on all cores. And if you contact your m/b, chances are, they won't give you a definite answer. So, 4790K is also safer in this regard (you are guaranteed to have higher clocks).

 

[LTC8K6]

I believe the cooler is TS15A in the above screenshot. The Xeon stayed pretty cool.

EDIT: it was a 212EVO, I found the thread:

http://forums.anandtech.com/showthread.php?t=2447253

The ASUS Z97-A board has no problems. It runs the Xeon at 3.8 under full load. Just went to manual, sync all cores, set to 38. Reboot. And Bob's your uncle.

So far, so good.

Power set to balanced in Win10. Idles nice and cool. All four cores go to 3.8 when called.

 

[freeskier93]

I believe the cooler is TS15A in the above screenshot. The Xeon stayed pretty cool.

EDIT: it was a 212EVO, I found the thread:

http://forums.anandtech.com/showthread.php?t=2447253

Awesome. Works great with my ASUS Z97-E, I have a 212 also. Lots of TDP headroom too.

 

[Headfoot]

4790k for $25 more every time. Better resale value, higher out of the box clocks, overclocks better, unlocked overclocking. Unless there is a Xeon specific feature you need the 4790k is better.

 

[LTC8K6]

4790k for $25 more every time. Better resale value, higher out of the box clocks, overclocks better, unlocked overclocking. Unless there is a Xeon specific feature you need the 4790k is better.

At the time, I got the Xeon for $250. A 4790K was well over $300.

 

[LTC8K6]

Right now at Newegg, the Xeon is $252 and the Haswell is $339.

Plus, it demonstrated that you could run all the chips at their turbo multi.

If you can get a 4790S for a good price, you can run it at 4.0 just fine.

 

[LTC8K6]

4790S is $304 vs $339 for the 4790K, for example.

There's little performance difference between them if you run the S at 4.0

But you have cooling headaches with the 4790K.

It's hard to tell the difference between my 4790K and my 1231V3, in fact.

 

[nenforcer]

I upgraded my mom from a Phenom ][ X4 840T to a 3.5 GHz Xeon E3-1270 V3 on a MSI Z97-GD65 Gaming motherboard for $60 and a MSI Z87-GD65 Gaming board from a local guy who had a tray of them. Very fast and I just took a OEM Geforce GTX 745 from my Dad's computer and put it into hers. The E3-1270 V3 sell for about $180 on eBay.

 

[freeskier93]

Right now at Newegg, the Xeon is $252 and the Haswell is $339.

Plus, it demonstrated that you could run all the chips at their turbo multi.

If you can get a 4790S for a good price, you can run it at 4.0 just fine.

If one happens to live near a Microcenter you can get the 1231v3 for $210.

 

[lyssword]

Unless there is some magic I don't know about this is a common misconception on how turbo boost works. The 3.8 turbo boost frequency is only for ONE core. The max all cores can run at is 3.6 with the 1231.

EDIT: After a bit of reading it looks like some motherboards can bypass turbo boost and force all 4 cores to 3.8. Guess I will have to look into this with my MB...

Actually you guys are both right, with latest bios all of the motherboards will lock it at 3.6 turbo, thats why you have to "retro" the bios to ealier bios which will let you max out all the cores at 3.8 or whatever.

 

[h9826790]

IMO, use XEON because of prefer to use ECC memory. From statistics, there is about 1bit error per GB memory per month (due to cosmic radiation). That means, nowadays, on a 32G memory system, there is about 1bit error per day. If that bit is on the idle memory, then doesn't matter. If that happen in a game, may be crashed (or the whole system crash). The worst thing is that error bit hided somewhere and affect the final product of some work. And ECC memory can avoid this effectively.

Also, when there is some strange behaviour on the computer, ECC RAM can quickly allow the user determine if the problem is coming from memory (in real world, AFAIK, there is no undetected error with ECC memory).

Apart from this, XEON has almost no benefit over the i7. My computer runs 24/7 with 48G of RAM, so I prefer ECC memory. However, in general, for non-mission critical computer, non ECC RAM is good enough. Also, ECC RAM is more expensive and about 2% slower than normal memory.

XEON is not that picky in memory, normal PC RAM works, not necessary go for the server memory. I guess the reported memory issue is that some one mix the RDIMM with UDIMM. RDIMM is very common in the server world, and mix RDIMM with UDIMM will cause unable to boot. It's quite common some people buy extra 1 or 2 larger capacity stick, and mix use with the current smaller sticks. However, the larger capacity server memory stick mostly are RDIMM, and most PC memory are UDIMM. I won't say it's a common mistake, but easily happen with little bit careless, and the user may blame the XEON.

On the other hand, mix ECC with non ECC RAM is not an issue at all, ECC will be disabled on all memory, and the system should work as usual.

Without iGPU not just mean that you didn't have the spare GPU, but you can't use QuickSync. If you ever use your computer for video encoding, QuickSync can save you lots of time (under some specific condition, e.g. only work on single pass encoding). Also, the coming DirectX 12 game may able to use the iGPU to assist the dGPU. Even though DirectX 12 is not common now, but if you plan to stay with this CPU for few years, you may better go for the i7 (if you use it for gaming).

Last but not least, if your mobo do not support ECC RAM, then the only reason to go for XEON may be just because it's cheaper. And it's cheaper because it has no benefit over the i7 for most of the people.

 

[Denithor]

IMO, use XEON because of prefer to use ECC memory. From statistics, there is about 1bit error per GB memory per month (due to cosmetic radiation).

Cosmetic radiation?



Autocarrot strikes again I presume.

Besides which, for most non-critical (non-server) applications, that 1 bit error per month likely occurs during a game or such having no significant impact.

 

[h9826790]

Cosmetic radiation?



Autocarrot strikes again I presume.

Besides which, for most non-critical (non-server) applications, that 1 bit error per month likely occurs during a game or such having no significant impact.

Haha, I didn't realise that I use the wrong word. Terms corrected, thanks for pointing that out, and sorry for my poor English.

Anyway, I am not kidding. A spaceship in 1997 proved that it receive 280 bit error per day constantly. And some solar activity will double that figure. However, The error rate on the Earth surface is still a mystery, but sure exist. Geographic location, solar activity, transistor's size, etc can cause the error rate million times difference. In the very old days (1990, from memory), IBM believe the error rate is about 1bit per 256MB per month. And Google did a very large scale research in 2009, which shows the peak error rate is 5 bit per 8GB per hour! And my 1bit per 1GB per month is an old figure which from google as well (another small scale research before that 2009 research).

When the transistor get smaller and smaller, this effect getting bigger and bigger. I believe this is one of the reason why the newer research always produce a figure that at least thousands times higher than the old figure.

If the peak is 5 bit per 8GB per hour. Then if the average is only 20% of that figure. Most of the modern computer on the world still experience a single bit error every hour. Of course, from real world experience, we know that our computer won't crash every hour even though we have 128GB of non ECC RAM. But that's just a suggestion why ECC RAM is better than non ECC RAM, and why so important in any mission critical computer.

Since OP ask about Xeon vs i7, that's why I bring up this issue.

 

[Magic Carpet]

XEON is not that picky in memory

Apparently so, but I haven't really dived into the details... Funny, Intel still lists TSX as supported by the Xeon. Guess, this info isn't really up-to-date.

 

[h9826790]

Apparently so, but I haven't really dived into the details... Funny, Intel still lists TSX as supported by the Xeon. Guess, this info isn't really up-to-date.

Yeah, they never update that page. I guess they just type in all the info at the beginning base on the CPU's "expected behaviour", and keep it forever, no update, no correction.