i7 5820K @4.4 slower than i5 3570K@4.4

[myocardia]

The memory controller is on the CPU.. Just like X58 & X79, you are able to exceed stated maximum memory.. X58 max was 24Gb, yet may people are running 48Gb.. It's only dependent on the size of the sticks of ram you're using, and if your CPU will see it. Has little to do with the chipset..

"Gigabyte GA-7TESM Dual LGA1366 is a server board & holds the most memory ever for X58 (288Gb)"

http://www.legitreviews.com/gigabyte-ga-7tesm-server-board-supports-288gb-of-memory-new-record_11487

And now you know why servers actually use server motherboards, not X99s, huh?

 

[Flapdrol1337]

The passmark verdict is in:
Numbers are written first Ivy, then Haswell-E (both @4.4GHz)
system: 2861 - 2949
proc: 8935 - 12655
Mem: 2731 - 2193
Disk: 4529 - 5113

But, the devil is in the detail:
CPU-single threaded: 2434 - 1991 (Millions of operations per second)
So in single threading, my Ivy is a good bit faster then my Haswell-E. This seems to be a known and has been reported on the net (including Anandtech). Unfortunately I have not paid enough attention to it. My simulations run multi-threaded, but only the actual matrix solver. Large parts of it, like setting up the solver and matrix is all single threaded.

It is what it is! I needed the 64Gb memory support.
Anyway thanks for all the input and if there is anybody in a similar situation: Lessons learned!

sjerra

Passmark is completely pointless though.

And haswell should be faster clock/clock, especially with the large cache. Did you monitor clockspeeds during load?

 

[birthdaymonkey]

I agree with the poster who suggested disabling Hyperthreading. When I upgraded from 2500K to 3770K (same mobo), the first comparison benchmark I ran was actually slower on the 3770K despite identical clocks. Turns out that benchmark runs worse with HT on. After I disabled it, the 3770K performed as expected.

 

[crashtech]

The single threaded result for your 5820K indicates it is running at around 3.6GHz, the stock max turbo setting.

 

[PhIlLy ChEeSe]

i7-5820K and 8x16 DDR4 2400 CL16 with SSD.

OK so it is 64GB of memory.
So as you have read the memory controller is on the CPU, at a certain point the voltage for the memory controller needs to be bumped up to be able to deal with more/faster memory(should only be a small bit)in your case unless you over clock them etc.
In your original post it is unclear as to which TEST you are using? Cause you state you tested them both from the on button? that is just seeing how fast one boots over the other? As you stated again in another post you will be future proof and I'm not clear on the Application your using for work?
But in the end your better off the way you went, now if you wanna speed it up we could go on for weeks. I would suggest a Very aggressive CPU cooler even water, as temps are the most important part of your set up. The CPU will heat up faster having more cores and memory controller on the die, add to that 64GB of memory and a bump in voltage..................You get my point.

 

[Dave2150]

Oh really? Forget i7.. Think Xeon
Max Memory Size (dependent on memory type) 768 GB
http://ark.intel.com/products/83352/Intel-Xeon-Processor-E5-2620-v3-15M-Cache-2_40-GHz

Did you not read the OP?

He has a I7 5820K, not a Xeon.

We all know Xeon's support more memory than the I7's.

Read the OP more carefully next time.

 

[soccerballtux]

If you believe memory is the main difference between the two, why not swap the memory in the two systems and run your benches again? There are cases where cas latency is more important than bandwidth, but latency also decreases with frequency.

http://blog.design-point.com/blog/2012/march/hardware-requirements-for-solidworks.aspx

^ This blog makes the claim that Solidworks is not (very?) multithreaded, except when you have multiple drawing views, so when running that particular program, 10-11 of your 12 threads are likely idle. Haswell is slightly faster (by a few percent) per clock than Ivy Bridge in almost all cases, but the clockspeed advantage of the i5 probably makes up for that, so in most cases I'd expect the new and old system to run about the same, all else being equal.

https://forum.solidworks.com/thread/56989

latency is always the same. 10-10-10-32 is the same as 5-5-5-16 DDR2. People complain about the latency but it's still the same real-world seconds to get the data. I don't see what the big deal is.

 

[Burpo]

Did you not read the OP?

He has a I7 5820K, not a Xeon.

We all know Xeon's support more memory than the I7's.

Read the OP more carefully next time.

I can read just fine.. thank you.. I was referiing to the post directly above mine & the blanket statement that was made..

Indeed, not to mention that the X99 chipset supports a maximum DIMM size of 8GB, and a maximum of 8 DIMMS.

 

[20TenGTS]

The passmark verdict is in:
Numbers are written first Ivy, then Haswell-E (both @4.4GHz)
system: 2861 - 2949
proc: 8935 - 12655
Mem: 2731 - 2193
Disk: 4529 - 5113

But, the devil is in the detail:
CPU-single threaded: 2434 - 1991 (Millions of operations per second)
So in single threading, my Ivy is a good bit faster then my Haswell-E. This seems to be a known and has been reported on the net (including Anandtech). Unfortunately I have not paid enough attention to it. My simulations run multi-threaded, but only the actual matrix solver. Large parts of it, like setting up the solver and matrix is all single threaded.

It is what it is! I needed the 64Gb memory support.
Anyway thanks for all the input and if there is anybody in a similar situation: Lessons learned!

sjerra

The problem with your system is that it is not running at full 4.4ghz when running single thread. I have a 5820k and I had the same issue. At stock speed, I score right at 2000 for passmark single thread. At 4.2ghz, it score the same. Yet, multithread test shows a big different. Checking my clock speed, it shows that running single thread, it is not running at my max 4.2ghz.

The only way to get it to run at max speed 4.2ghz on single thread application is to turn the power option mode to High performance.
Once I'm on high performance, I get 2525 for passmark single thread test.
For Cpu processor I get 15445.

I don't know why the cpu won't boost to max speed on single thread on balance mode. Maybe it thinks single thread are not stressing the cpu enough. Even stressing with cinibench 15, it takes like 5 minute to finish the single thread benchmark. After I turn on high performance power option, it takes 4 minute.

 

[crashtech]

Having CPU-Z running on the desktop during the tests will quickly reveal what's going on. But it was clear from his score that something is not right.

 

[Magic Carpet]

@20TenGTS

Interesting observation. Anybody else could confirm that?

 

[biostud]

Indeed, not to mention that the X99 chipset supports a maximum DIMM size of 8GB, and a maximum of 8 DIMMS.

X99 supports 128GB memory with Xeon CPU's.

 

[biostud]

Indeed, not to mention that the X99 chipset supports a maximum DIMM size of 8GB, and a maximum of 8 DIMMS.

X99 supports 128GB memory with Xeon CPU's.

 

[sjerra]

PhIlLy Cheese,
The test was conducted is followed: I'm running an Ansys Maxwell 3D simulation and I started it on my old and new rig at the same time. I know it might sound unprofessional, but it was the only real-world option I had. On step better for me was to time it, which, agreed, I could still do. Anyway. As they were progressing through the job, I could see my new rig falling behind.

I'm not big at overclocking. I find the bios options overwhelming and rig stability is important. Having said that, I don't by a K for nothing! But from history, I learned to stay away from the BCLK and overclocking memory. I used to get hidden data corruption in the past. So normally only the multiplier and the core voltage. But as said, all the other options stay on auto. (here might actually lay the problem) And I do run a corsair H110. Watercooling has been recommended and I found 1.5kgs of cooler hanging of my MB very unpractical/awkward.
Also, hyperthreading has been disabled


20TenGTS,
Your insight is very interesting. I thought to live with the little under-performance, but now I run into another snag. Something very weird. In Ansys, when drawing a polyline (so start with a point and continue onwards putting more points on the screen that get interconnected as lines) I get a ridiculous lag. It's unbearable. He keeps refreshing the screen and after twenty or so clicks it takes about half a minute to finish the redraw and produce the complete segmented line. Mind you, I have a Firepro (that works fine with solidworks)
My Ivy does not have this lag problem, even now that I have taken out the firepro from this machine and am running on the internal graphics card there is no lag whatsoever when drawing such type of line.
To test what could be happening, I reduced the multiplier on my IVY to the lowest (1600) and law and behold, the same behaviour. Terrible lag when drawing a polyline. So this seems a processor speed issue.

When monitoring the cores on my 5820K I see the multiplier go up to x36 (I went back to bios defaults) but taking your experience into account and jumping to quick conclusions from my IVY tests it seems that for this specific task my 5820K doesn't even bother to come out of its low multiplier mode ( x12) ...
Is this even possible? I'm having a nightmare time troubleshooting this cause I don't have a reference and can not consistently monitor all the variables involved for an extended period of time. (Single threaded processes keep jumping core and in AiSuite I must toggle between cores to check individual multiplier states so half the time I'm too late or don't notice any jump in multiplier at all)
In Aisuite my power is set to performance.
20TenGTS, did you set your power options in the bios or in the AIsuite software?

Magic Carpet, I'm going to test this tonight, but my feeling is I'm experiencing at least similar issues....


Thanks for all your input,
sjerra

 

[myocardia]

Neither. 20TenGTS is talking about the Windows power plan options. Yours is now set to Balanced, like everyone else on Earth who hasn't changed it. This page shows the steps to change it, in Windows 8: http://www.thewindowsclub.com/configure-power-plan-windows-8

edit: Anyone else having problems quoting posts?

 

[crashtech]

edit: Anyone else having problems quoting posts?

Not so far, though I have made some double posts recently, and I had never done that before the most recent downtime.

 

[sjerra]

20TenGTS,
I can confirm your theory. In balanced mode, it looks like for a single threaded apps, the processor does not go past the rated multiplier even when overclocking. In high performance mode I get a 2649 score.
Is this a bug? Is this a reality or an issue with passmark and windows?
I'm not sure if I should leave my cpu at full freq & voltage nonstop with high performance mode.

The lag in my graphics is still an issue though. even in high performance mode it is still there.

sjerra

 

[crashtech]

If this problem proves difficult to solve, just put a shortcut to the power applet on the desktop, and switch to High Performance when you need the extra power.

Do you have another GPU you could try? It seems you mentioned the lag went away (oddly enough) with integrated graphics, which might indicate some odd interaction between the new board and old GPU, or a driver problem.

 

[Smoblikat]

If you believe memory is the main difference between the two, why not swap the memory in the two systems and run your benches again? There are cases where cas latency is more important than bandwidth, but latency also decreases with frequency.

http://blog.design-point.com/blog/2012/march/hardware-requirements-for-solidworks.aspx

^ This blog makes the claim that Solidworks is not (very?) multithreaded, except when you have multiple drawing views, so when running that particular program, 10-11 of your 12 threads are likely idle. Haswell is slightly faster (by a few percent) per clock than Ivy Bridge in almost all cases, but the clockspeed advantage of the i5 probably makes up for that, so in most cases I'd expect the new and old system to run about the same, all else being equal.

https://forum.solidworks.com/thread/56989

Uh, he cant use DDR4 in a DDR3 board, and the other way around doesnt work either......

 

[sjerra]

Now for some tangible verification:
I ran a smaller version of my simulation two times with my 5820K overclocked to 4.4.
With high performance settings it took 48 minutes
With balanced settings it took excactly 1 hour
That's 20% more in balanced.
So somehow, the proc can't be bothered (for lack of better terms) to speed up to 4.4 when overclocked and in balanced mode.

Crashtech, I'm not yet giving up on finding the problem. I do have another graphics card -crap htpc type- lying around so that will be going into the system to check. I am suspicious though that there is some odd interaction going on. Otherwise this problem would have been widely discussed on the internet.
I'll also be removing some of the ram and see.

sjerra

 

[crashtech]

Crashtech, I'm not yet giving up on finding the problem. I do have another graphics card -crap htpc type- lying around so that will be going into the system to check. I am suspicious though that there is some odd interaction going on. Otherwise this problem would have been widely discussed on the internet.
I'll also be removing some of the ram and see.

sjerra

Well, that would presume enough users are running a Firepro W5000, an X99 Deluxe, are sharp enough operators to notice there is a problem, AND are posting about it on the Internet. I'm not a statistics whiz, but it seems possible that there could be an unreported issue of some sort, probably centered around drivers. You might want to try other PCIe slots just for fun as well.

 

[sjerra]

So,
The latest tests revealed that there are two unrelated issues here.
1. The lag in my cad application is graphics related (could be driver, I have to see) because swapping out the firepro for something else solved it.

2. The fact that my proc does not go in turbo mode (it stays in x33) for single threaded applications in stock or when overclocked (when in windows balanced power). It only goes into turbo with multithreaded applications

If I can get the second issue solved I'll be a happy man. I really don't fancy going in and out of the power options panel every time I want to run a simulation.

sjerra

 

[Magic Carpet]

@sjerra

Can't you just use MultiCore Enhancement? Set all the cores to 36 or something and call it a day until Asus/Microsoft fixes it.

X99-DELUXE BIOS 1601

 

[crashtech]

I think x36 represents a performance regression for the OP.

 

[Magic Carpet]

I think x36 represents a performance regression for the OP.

Well, whatever works best for that setup. Should easily do forty, since piss-poor tim is no longer an isssue. 3570K will be left in the dust