Keep R3E + Xeon [email protected] or i5 4670k+Z87 Sabertooth?

CruisD64

2[H]4U
Joined
Mar 6, 2007
Messages
2,461
So I happened to get a screaming deal on an i5 4670k and a Z87 Sabertooth motherboard yesterday. Screaming to the point where I ran to the bank immediately to make sure I'd get it knowing it would AT LEAST go into my backup PC, possibly my primary.

My question is, do I keep my sig rig as my primary or upgrade to this newer setup. I would plan on overclocking it to about 4.5 if I could, but my gut is telling me what I already have would suit my needs better. I primarily use my PC for gaming and media, but I do also spin up the occasional VM (sometimes 4 or 5) for testing new things for work.

It's also worth noting that I'm strongly considering selling my backup PC. My thoughts are the Z87/4670K combo would fetch a higher price than with the R3E/x5660 (Yes, I know what X58 Motherboards are fetching these days, but I'd rather not part out)

The benchmarks I've looked at suggest that my primary rig would perform nearly as well as the 4670k in gaming currently, but will the gap spread as graphics cards improve and I consider upgrading mine?

Anyway, I think I already know the answer, but I'd like to hear your thoughts.

Thanks!
 
Put it together and run it - then decide, you will primarily be amazed with how much cooler the room is.
 
4670k cranked up to 4.5 would blow that 5660 out of the water, 2 extra cores or not.
 
Blow it out of the water? How much you think? I could see it being like 15% difference max...

I should have clarified: in gaming. I can easily see far more than 15% difference if you got the GPU to back it up.
 
This page and the next one in the article have a lot of relative information http://www.anandtech.com/show/7003/the-haswell-review-intel-core-i74770k-i54560k-tested/6

Rough math adjusting for clock speed says you are looking at a 10-30% difference in single threaded stuff. I know its 2014 and a lot of games use 3 and 4 cores, but most are still strongly loading one primary core.

I have a 980x x58 system, if it was all I had I don't think I would run out and upgrade, but if you have the Haswell system I would give it a shot. They run much, much cooler than x58.
 
My Westmere at 4.6ghz beats my stock 5820k. At around 4.2ghz it'll have roughly stock 5820k performance.

Now you can reposition yourself and decide if you're alright with stock Haswell-e 5820k performance or would rather have an overclocked haswell i5. The i5 obviously have an higher ipc but it'll never beat hexacores under normal air/water overclocking in multi threaded apps.

Just some food for thought. But in my opinion going from 6c/12t to 4c/4t is very backwards.

4fce39cb_fXVLSz3.jpeg
 
This page and the next one in the article have a lot of relative information http://www.anandtech.com/show/7003/the-haswell-review-intel-core-i74770k-i54560k-tested/6

Rough math adjusting for clock speed says you are looking at a 10-30% difference in single threaded stuff. I know its 2014 and a lot of games use 3 and 4 cores, but most are still strongly loading one primary core.

I have a 980x x58 system, if it was all I had I don't think I would run out and upgrade, but if you have the Haswell system I would give it a shot. They run much, much cooler than x58.

I didn't take that away at all. Only one of the benchmarks compared a hex core Xeon (in Civilization V which is heavily CPU bound). The next bench down is consistent with the fact that the majority of games are not CPU bound and didn't compare an X58 system at all.

All said and done, for gaming I wouldn't expect to see any difference in most games. The quoted page has very little in the way of threaded benchmarks and if you run a few VMs for testing the Xeon may end up being a little faster (12 threads vs 4) but I would suggest you build it and test it out, see how you like it, then decide.

Temperature wise, the Xeon runs fairly cool and actually doesn't output a whole lot more heat than the newer CPUs if you enable all the C States etc. However, the X58 chipset can output a lot of heat, I know the R3E runs VERY hot if you disable the power saving features on the CPU and start pushing extra voltage for overclocking (my video card ran 10C higher from the increased case temps) it's not awful with all the power saving enabled though and my system is more or less silent.

The subsystem on the newer system is significantly faster, i.e. native SATA 6GB/s, native USB 3.0, PCIe v3.0, the only advantage the X58 has over the Z87 is memory bandwidth which may not mean much at all. For this reason alone I would lean towards the Z87, the difference in file transfer speeds on my Z87 and Z97 systems are substantially faster than my old X58.
 
My Westmere at 4.6ghz beats my stock 5820k. At around 4.2ghz it'll have roughly stock 5820k performance.

Now you can reposition yourself and decide if you're alright with stock Haswell-e 5820k performance or would rather have an overclocked haswell i5. The i5 obviously have an higher ipc but it'll never beat hexacores under normal air/water overclocking in multi threaded apps.

Just some food for thought. But in my opinion going from 6c/12t to 4c/4t is very backwards.

That is my thinking as well. My plan originally was to ride out my current setup until I'm ready for a fresh build. I'll probably put the Haswell in, play with it for a few days, then sell the system off. It's going a be quite a decent gaming machine for someone.
 
I didn't take that away at all. Only one of the benchmarks compared a hex core Xeon (in Civilization V which is heavily CPU bound). The next bench down is consistent with the fact that the majority of games are not CPU bound and didn't compare an X58 system at all.

All said and done, for gaming I wouldn't expect to see any difference in most games. The quoted page has very little in the way of threaded benchmarks and if you run a few VMs for testing the Xeon may end up being a little faster (12 threads vs 4) but I would suggest you build it and test it out, see how you like it, then decide.

I guess I should've been more specific but a lot of those tests include a Nehalem quad core, and that's what I was referencing. I know there is a 1% IPC gain or something to westmere. Assuming that the vast majority of games do not use the hex core it should be relevant info.

I have a 980x/RII Gene system and a 3770k system sitting side by side both with all C States enabled and both with Hyper 212s. In basic tasks I can hear the fan always working on the 980x and feel the heat coming out. Unless the 32nm xeons run far cooler than the 32nm consumer chips?
 
The next bench down is consistent with the fact that the majority of games are not CPU bound and didn't compare an X58 system at all.

Ding ding ding. If you're coming off a highly overclocked 6-core Westmere, don't expect anything tangible while gaming. A few loud voices on this forum will continually recommend the newest Intel processor for gaming, as if there are massive gains to be had. In reality, intel has failed us enthusiasts since Sandy Bridge and we should punish them by declining to purchase their glacially-improving products. Ivy and Haswell have brought almost nothing new to the table, while pushing down the overclock ceiling and barely breaking even with IPC gains. Add the TIM vs Solder fiasco, deliberate stagnation at 4-cores, etc, and it's clear: Intel is pinching pennies and taking us for granted.

Fact: A Xeon 5660 at an easy overclock of 4.2 is not a bottleneck is nearly all conceivable gaming scenarios. The only possible situation where this MIGHT be tested is:

At 1080p and below

AND

In a highly cpu-intensive game

AND

When any possible GPU bottleneck has been completely removed by SLI/Crossfire of high-end cards



In a world of 1440p, eyefinity, and even 4K quickly replacing 1080p, things are going to be even less CPU-bound (if that were possible).

People are blowing hot air when they tell you you'll see massive improvements in gaming, outside of the above absurd scenario. The improvements going from overclocked Westmere to overclocked Haswell are going to be extremely negligible for gaming, if you can notice them at all. And any gains you think you've locked in with the new chip will disappear when you upgrade to a higher resolution and as more graphically-demanding games inevitably come out.
 
Last edited:
I guess I should've been more specific but a lot of those tests include a Nehalem quad core, and that's what I was referencing. I know there is a 1% IPC gain or something to westmere. Assuming that the vast majority of games do not use the hex core it should be relevant info.

I'm confused. You are referring to the same link that was posted here right? That bullshit Anandtech list? Look again, just two games were benchmarked and only one contained X58 (Civ V which, ironically, is highly threaded so going to hex core would benefit a great deal, yeah, they list dual and single X5680 but on what?). Explain how it's relevant?

It might be relevant for the other dozen or so synthetic benchmarks.

Go have a look at the X58 Xeon enthusiast thread, there is a link in there to a very comprehensive list of gaming benchmarks with an X58 + Xeon.

Unless the 32nm xeons run far cooler than the 32nm consumer chips?

The 980x is 130W vs X5670's 95W but after overclocking it probably isn't any different (the X5680 is the same spec and wattage as he 980x).

Ultimately I don't think the CPU wattage is worth a damn when it comes to system temps because we're [H]ard and we like to overclock the shit out of everything. 84W (Haswell) vs 95W (X5670) isn't going to break the thermal bank. I think the major heat reductions come from across the board i.e. chipset, integrated extras (sound, NIC, bluetooth etc) and as everything moves to a smaller scale across the board you start to see large changes in internal case temps, lower case temps = lower cpu temps and fan speeds.

There is no question though, we are pushing the limits of clock speeds here and no matter what kind of temperature difference you see, none of it equates to a better overclock, we're pretty much stuck at sub 5Ghz speeds (unless you are super rich or super lucky).
 
Last edited:
You know what? I'm calling BS on that article.

At first it appears as though they are doing a direct comparison between 6 different systems but later down the list it looks like much more, then a different list again?

It's seems like they have simply pulled numbers from older comparison articles to lump in to the list, they also don't provide any information about the setup beyond the CPU/GPU for the only two gaming benchmarks, no driver versions, memory speeds, clock speeds, type of disk used, no min or max fps, only average. It's piss poor.

Lets say for example, they pull numbers from a Civ V benchmark run on the X5690 and 7970 a year ago. That 10fps difference in the graph could simply be from video driver updates. Compare to any [H] article where they tell you the comparison systems up front including display driver version.
 
Back
Top