Being pushed toward a new PC
 

Over the last several months, I have seen variations in both P95 and mfaktc performance. I won't go into the details, ATM. However, I recently got into a situation in which I realized that the Windows reports of the amount of RAM had changed. Instead of saying 32 GB, it was saying 32 GB, 15.9 GB available. I did some digging, and found references to the other 16 GB being devoted to hardware mapping. This did not make sense, as I was not aware of any hardware changes, and such a division of RAM had not appeared previously.

In the course of trying to understand the situation, I took to reseating, the memory sticks. This led to testing pairs, and then individual DIMMs. To cut to the chase, I seem to have lost the use of the third and fourth memory slots. Any of the sticks will work in the first two slots. I see 16 GB of RAM, but it is only single channel. If a DIMM or DIMMs are in the other two slots, and the first two are empty, the machine beeps "No RAM" at boot. I only ever see 16 GB, single channel, no matter if slots 3 and 4 are populated.

I have now gone through checking the slots for debris, repeated contact cleaning, and installing different DIMMs in different slots. Nothing changes. Oddly, in the Asus BIOS, the GPU-DIMM POST still sees all four sticks, but the Main tab only sees 16 GB. memtest86 sees the same.

I don't mind that much, under my current pursuits, that I can't access half of the RAM. Loss of dual channel operation is devastating.

I'm not quite sure was my next step is. A co-worker recently offered me a Dell, still under warranty, for a really good price. I will have to get more details on that. In an ideal world, I'd be looking at an X99 board and perhaps an i7-5820k, but I'm not sure if that is currently possible.

[QUOTE=kladner;446087]Over the last several months, I have seen variations in both P95 and mfaktc performance. I won't go into the details, ATM. However, I recently got into a situation in which I realized that the Windows reports of the amount of RAM had changed. Instead of saying 32 GB, it was saying 32 GB, 15.9 GB available. I did some digging, and found references to the other 16 GB being devoted to hardware mapping. This did not make sense, as I was not aware of any hardware changes, and such a division of RAM had not appeared previously.

In the course of trying to understand the situation, I took to reseating, the memory sticks. This led to testing pairs, and then individual DIMMs. To cut to the chase, I seem to have lost the use of the third and fourth memory slots. Any of the sticks will work in the first two slots. I see 16 GB of RAM, but it is only single channel. If a DIMM or DIMMs are in the other two slots, and the first two are empty, the machine beeps "No RAM" at boot. I only ever see 16 GB, single channel, no matter if slots 3 and 4 are populated.

I have now gone through checking the slots for debris, repeated contact cleaning, and installing different DIMMs in different slots. Nothing changes. Oddly, in the Asus BIOS, the GPU-DIMM POST still sees all four sticks, but the Main tab only sees 16 GB. memtest86 sees the same.

I don't mind that much, under my current pursuits, that I can't access half of the RAM. Loss of dual channel operation is devastating.

I'm not quite sure was my next step is. A co-worker recently offered me a Dell, still under warranty, for a really good price. I will have to get more details on that. In an ideal world, I'd be looking at an X99 board and perhaps an i7-5820k, but I'm not sure if that is currently possible.[/QUOTE]

Perhaps reseat the CPU? I have lost a Memory channel on a system where it turned out the CPU wasn't making good contact on a pin and/or a socket pin was bent. Heat cycles can affect this even if you haven't messed with the CPU itself.

[QUOTE=airsquirrels;446089]Perhaps reseat the CPU? I have lost a Memory channel on a system where it turned out the CPU wasn't making good contact on a pin and/or a socket pin was bent. Heat cycles can affect this even if you haven't messed with the CPU itself.[/QUOTE]
I will certainly have a go at it.

However, there is a more complicated story going on, which has involved more than one CPU change. While I was wondering about variable throughput, my Corsair H100i (original model) started going bad. It was so bad by the time I set up the RMA, that I found myself better off with the stock AMD cooler.

When the H100i replacement, a V2 model, arrived, and I was removing the stock cooler to put it in, the heatsink compound yanked the 8350 chip out of the socket. It bent pins on one edge, and trying to straighten them just broke one. Long story short, I ended up with an FX-9590. No problem, I thought at first, as the V2 cooler with the 9590 (4.7 GHz) kept things cooler than the V1 cooler had kept the 8350 (4 GHz).

However, this was the point that I realized that I was getting worse performance from the faster chip. So, there is a definite possibility that something is borked in the CPU socket. I will have to do some more cleaning, digging around and reseating of things.

Thanks for the tip!

[QUOTE=kladner;446097]...When the H100i replacement, a V2 model, arrived, and I was removing the stock cooler to put it in, the heat-sink compound yanked the 8350 chip out of the socket.[/QUOTE]

I have found that slightly twisting a heat-sink side-to-side will break the seal created by the compound. It doesn't take much. You'll feel it when it lets go.

[QUOTE=storm5510;446256]I have found that slightly twisting a heat-sink side-to-side will break the seal created by the compound. It doesn't take much. You'll feel it when it lets go.[/QUOTE]
Yeah. I was impatient. That technique is a very good idea.

[QUOTE=kladner;446260]That technique is a very good idea.[/QUOTE]

No, it isn't.

LaurV et al, care to interject?

The situation with my system has deteriorated. I cannot run Prime95: it freezes or blue screens. Indications are toward hardware problems. I can run the GPUs. I can run Small FFT torture test, but not the others. This suggests a problem in the memory part of the system. I have stripped down to 1, and then 2 DIMMs, in different slots, but problems persist. I am suspecting that this "Open Box" mobo has problems. I can trade it in at Microcenter, but I am debating whether to try once more with the AMD CPU, or to make the leap to Intel.

I know prices are likely to drop when AMD Zen comes out, but I don't know if I can hold out that long. I have started releasing assignments I haven't started. Unfortunately, there are two which were running on the CPU which are at about 60%. I may have to throw that work away. :sad:

It does sound like the machine is broken. I would replace the motherboard if possible at no charge. It will be easier to sell when new stuff comes out.

Also, Zen might not be the best for Prime95. The desktop Ryzen chips apparently have only two memory channels and possibly half-speed AVX2, so we'll have to see how the price compares to an i5-6500/7500. So it may be that going Intel now wouldn't be a bad thing.

[QUOTE=Mark Rose;449575]It does sound like the machine is broken. I would replace the motherboard if possible at no charge. It will be easier to sell when new stuff comes out.

Also, Zen might not be the best for Prime95. The desktop Ryzen chips apparently have only two memory channels and possibly half-speed AVX2, so we'll have to see how the price compares to an i5-6500/7500. So it may be that going Intel now wouldn't be a bad thing.[/QUOTE]
An i5 and DDR4 dual channel would be a great step up. I lust for a hex i7 with quad channel, but that adds ~50% to the price of the basic components.

[QUOTE=kladner;449594]An i5 and DDR4 dual channel would be a great step up. I lust for a hex i7 with quad channel, but that adds ~50% to the price of the basic components.[/QUOTE]

Even an i3 with DDR4 dual channel would be a decent setup. You'll get >> 50% of i5 performance at roughly half the cost. You'll have to do your homework to see if it actually makes sense compared to an i5.

As a point of reference, my i3 skylake does about 28 Gd/d running 1st time LL on 4 (hyper)threads.

[QUOTE=kladner;449594]An i5 and DDR4 dual channel would be a great step up. I lust for a hex i7 with quad channel, but that adds ~50% to the price of the basic components.[/QUOTE]

Yeah, when I built my mini cluster, I went with i5-6600 along with DDR-2133. The memory can't keep up with the CPU, but having the high clocked part should help with resale value. The hex core is much faster (I have access to such a system) with the doubled memory bandwidth, but it isn't the cheapest way to do LL.

I should really look at optimizing the power consumption of the cluster via BIOS settings, under locking, etc.

[QUOTE=axn;449598]
As a point of reference, my i3 skylake does about 28 Gd/d running 1st time LL on 4 (hyper)threads.[/QUOTE]
What is the core clock on the i3?

[url]http://ark.intel.com/products/93366/Intel-Core-i3-6098P-Processor-3M-Cache-3_60-GHz[/url]

[QUOTE=axn;450939][URL]http://ark.intel.com/products/93366/Intel-Core-i3-6098P-Processor-3M-Cache-3_60-GHz[/URL][/QUOTE]
Thanks! That is pretty impressive. ECC memory, even.

I confess that I went with the whole enchilada: i7 6700K, on an Asus Sabertooth Z170 Mk1. It has taken some effort to adjust to the Intel environment after 20+ years of AMD, and vast development by Intel since the Pentium Pro Overdrive chip: last Intel I ran.

I think I have CPU stability pretty well worked out. Early issues were actually glitches in my Win 7 transition. Beyond those, it seems that this chip is happy at 1.30-1.31 Vcore, though Auto in the BIOS puts it over 1.35 V.

Memory is 2 x 8MB, Corsair Vengeance LPX, rated 2133 MHz, 2400 XMP. It seems very well behaved on the XMP settings.

Default settings also put the CPU in Sync All Cores mode, which runs at Turbo speed on all the cores. I turn that off, so far. It runs hot, and I have not had time to evaluate stability. I am happy to have it produce good DCs at 4.0 GHz, while staying under 60° C. :smile:

EDIT: It does 58 G-D in 30 hours, which I take to be about 46 G-D/D, on 39M DCs.

If it's not too late, I would switch your RAM for something higher clocked. Your CPU is very much memory starved. Usually DDR4-3000 is reasonably priced compared to DDR-3200.

I've found that 4-core Skylake is memory starved above 3.3 GHz with dual channel DDR-2133: clocking faster for LL just makes more heat with only tiny performance increases. If you're getting 4.5 GHz out of your chip then DDR4-3000 will be about the sweet spot.

[QUOTE=kladner;450940]
I confess that I went with the whole enchilada: i7 6700K, on an Asus Sabertooth Z170 Mk1. It has taken some effort to adjust to the Intel environment after 20+ years of AMD, and vast development by Intel since the Pentium Pro Overdrive chip: last Intel I ran.
[/QUOTE]Oooh, that brings back memories. My PPro 225 is stashed away in the loft and was still working last time it was powered up.

[QUOTE=xilman;450945]Oooh, that brings back memories. My PPro 225 is stashed away in the loft and was still working last time it was powered up.[/QUOTE]
A single PPro ran just fine for me. However, when I attempted to run 2 of the overdrive version together it burned the board power connector. :picard:

[QUOTE=kladner;450947]A single PPro ran just fine for me. However, when I attempted to run 2 of the overdrive version together it burned the board power connector. :picard:[/QUOTE]Mine was a H-t-G PPro 200MHz device, suitably over-clocked. It was the 256k cache version; the two 512K chips I had available wouldn't run reliably at 225MHz. The extra cache helped at 200MHz over the 256K version but couldn't compete with the 225MHz, 256K chip.

I still have ~10 of those chips. I believe they are now worth USD 20 or so each as scrap Au. Those were the days.

[QUOTE=Mark Rose;450943]If it's not too late, I would switch your RAM for something higher clocked. Your CPU is very much memory starved. Usually DDR4-3000 is reasonably priced compared to DDR-3200.

I've found that 4-core Skylake is memory starved above 3.3 GHz with dual channel DDR-2133: clocking faster for LL just makes more heat with only tiny performance increases. If you're getting 4.5 GHz out of your chip then DDR4-3000 will be about the sweet spot.[/QUOTE]
I certainly noticed the difference when I switch it from 2133 to 2400 MHz. I will have to look at the receipts.

[QUOTE=Mark Rose;450943]If it's not too late, I would switch your RAM for something higher clocked. Your CPU is very much memory starved. Usually DDR4-3000 is reasonably priced compared to DDR-3200.

I've found that 4-core Skylake is memory starved above 3.3 GHz with dual channel DDR-2133: clocking faster for LL just makes more heat with only tiny performance increases. If you're getting 4.5 GHz out of your chip then DDR4-3000 will be about the sweet spot.[/QUOTE]

Generally agreed with the above. I run my quad Skylakes at 4.2 GHz and try to get as fast ram I can in them, but also finding that ram compatibility isn't as great as it could be once you reach around 3000+ speeds. Also note dual rank ram can offer significant performance over single rank in the same speed/setting configuration. Where ram limited, I found dual rank gives about 20% more performance. Unfortunately it is hard to find this info listed, but as a rule of thumb 4GB DDR4 modules are single rank, 8GB may be single or dual, with higher performing ones generally moving towards single.

Note my observations apply to the case of one core per worker as that is essentially my use case with LLR. I've not really looked into multiple cores per worker.

[QUOTE=mackerel;450981]Generally agreed with the above. I run my quad Skylakes at 4.2 GHz and try to get as fast ram I can in them, but also finding that ram compatibility isn't as great as it could be once you reach around 3000+ speeds. Also note dual rank ram can offer significant performance over single rank in the same speed/setting configuration. Where ram limited, I found dual rank gives about 20% more performance. Unfortunately it is hard to find this info listed, but as a rule of thumb 4GB DDR4 modules are single rank, 8GB may be single or dual, with higher performing ones generally moving towards single.

Note my observations apply to the case of one core per worker as that is essentially my use case with LLR. I've not really looked into multiple cores per worker.[/QUOTE]

Good to know. My Skylake systems are all running 2 x 16GB of G.SKILL Aegis DDR4-2133 F4-2133C15D-32GIS. From what I can tell, those are dual rank modules.

[QUOTE=Mark Rose;450990]Good to know. My Skylake systems are all running 2 x 16GB of G.SKILL Aegis DDR4-2133 F4-2133C15D-32GIS. From what I can tell, those are dual rank modules.[/QUOTE]
I take it that you are running this somewhat faster than the label suggests?

EDIT: Or maybe not, on re-reading. I have seen Corsair 3000 modules for about what I paid for nominal 2133. As near as I can tell from the returns policy as stated online, I can still exchange the ones I have. It does seem that more of the parts at those speeds are 1.35 V, instead of 1.2.

This really makes me wonder if the RAM I have will run reliably above DDR4-2400. However, getting parts which are rated for the speed, and have specified timings, has a lot of appeal.

I am running at the stock clock of 2133. Getting overclocking motherboards, CPUs, and memory didn't make economic sense.

It's unlikely your RAM will run reliably faster. RAM is extensively binned.

Thanks, Mark. I really appreciate your evaluations. I will talk to Micro Center tomorrow about trade-ins.

So looking at the local offerings at Microcenter, there are a couple of possibilities. I can move up the line with Corsair for a reasonable price bump. They also have the G.Skill Aegis for a bit less. Again, I am looking at 2 x 8GB, so I don't know how the dual rank question shakes out.

The only other difference I see is that the Corsair seems to be 1.2V versus 1.35V. In reading around the web I have seen some folks preferring lower voltage so as to go easy on the memory controller. I would appreciate any views on this issue.

The lower voltage is probably better, but it's the 3000+ speed that will matter more.

You'll probably end up with single rank RAM. If you have the specific model numbers it may be possible to figure it out before buying.

The voltage usually depends on the speed. 1.35v kicks in with some lower latency 2666 kits, and pretty much everything 2800 upwards. I wouldn't worry about it either way.

Most recently I got Kingston HyperX 2x8GB 2666 because they were dual rank, and I didn't want the higher speed compatibility problems. You can look up the part number and they do list the rank configuration.

I did come across some Crucial Ballistix 4GB modules which were mis-labelled as dual rank though.

The G.Skill part is F4300C16D16GISB

The Corsair is CMK16GX4M2B3000C15
[URL]http://www.corsair.com/en-us/vengeance-lpx-16gb-2x8gb-ddr4-dram-3000mhz-c15-memory-kit-black-cmk16gx4m2b3000c15[/URL]

Looking at the Corsair site, it looks like the 1.2V listing at Micro Center may be an error. 1.35 V looks to be pretty much universal at DDR4-3000.

EDIT: Cross post. I have not found if Corsair lists the rank config. However, the 1.2 V rating only seems to apply at SPD 2133. XMP bumps it to 1.35.

Sometimes the listings can get confused, and they might list 1.2v which the modules could do at basic 2133 speed.

[QUOTE=mackerel;451054]Sometimes the listings can get confused, and they might list 1.2v which the modules could do at basic 2133 speed.[/QUOTE]
Thanks! That does seem to be the case, here.

And yes. Micro Center will still exchange what I have. :smile:

1.35v is probably the max safe voltage(some would say 1.4v). Given that there are so many 1.35v kits it can't be that damaging. 3000 was the sweet spot when I bought.

I have the trade lined up. I just have to drive over to the store. Thanks for all the responses.

I exchanged the XMP 2400 parts for XMP 3000. Strangely, running a single, four thread worker is slightly (0.10 to 0.20 ms) slower, now. Would the increased speed show up if I were running more than one worker?

EDIT: I did notice last night that the current RAM shows Single ranks in CPUZ>SPD. I don't know what the previous RAM indicated. I am supposing that it would be the same as these are from the same Corsair Vengeance LPX product line.

[QUOTE=kladner;451113]I exchanged the XMP 2400 parts for XMP 3000. Strangely, running a single, four thread worker is slightly (0.10 to 0.20 ms) slower, now. Would the increased speed show up if I were running more than one worker?[/QUOTE]
One worker wouldn't max the memory bandwidth

[QUOTE=henryzz;451114]One worker wouldn't max the memory bandwidth[/QUOTE]
Thanks! I'll move toward multiple workers.

[QUOTE=henryzz;451114]One worker wouldn't max the memory bandwidth[/QUOTE]

One worker with four threads would max the memory bandwidth.

It's kind of obvious, but have you confirmed the XMP profile is enabled and working?

[QUOTE=Mark Rose;451125]It's kind of obvious, but have you confirmed the XMP profile is enabled and working?[/QUOTE]
'Fraid so.
EDIT: On the bright side, the parts trade only set me back 45 minutes, and $3.45 price difference.

One factor which might be in play here is that the original XMP-2400 RAM was CAS 14, while the XMP-3000 is CAS 15. The difference in ms/it is 2.628 versus ~2.877.

[QUOTE=kladner;451217]One factor which might be in play here is that the original XMP-2400 RAM was CAS 14, while the XMP-3000 is CAS 15. The difference in ms/it is 2.628 versus ~2.877.[/QUOTE]

14/(2400/2) MHz = 11.7 nanosecond latency
15/(3000/2) MHz = 10.0 nanosecond latency

Try running the new RAM at 2400 & 14? The slowness may be the single/dual rank thing mackerel stumbled upon.

[QUOTE=Mark Rose;451222]14/(2400/2) MHz = 11.7 nanosecond latency
15/(3000/2) MHz = 10.0 nanosecond latency

Try running the new RAM at 2400 & 14? The slowness may be [U][B]the single/dual rank thing[/B][/U] mackerel stumbled upon.[/QUOTE]
That seems a possibility. Another is that the earlier run at DDR4-2400 may also have had the CPU cores locked into Turbo at 4200 MHz. Switching to XMP does that, and I did not include CPUZ in that screen grab to show the core clocks.

I wish I had paid more attention when I had the previous set so I could give a positive answer. It is a relatively small difference in any case: 0.1 to 0.15 ms/it, more or less.

To update: With regard to the relative speed of different memory modules, I have come to the conclusion that there are too many variables at play to clearly assign causes for the numbers that turn up in a particular LLDC run. The earlier "slower" RAM had the advantage of fewer cycle hogging apps, especially performance monitoring apps. Earlier in the rebuild, things were more basic.

I can see tenths of milliseconds difference per iteration by shutting down and restarting things like HWiNFO64. Worse yet are things like Corsair Link. Worst of all, so far, is Asus AI Suite III. Even if the main program is shut down, there are 3-4 services which keep running. These can be disabled, but not stopped without a reboot.

The current RAM, rated XMP 3000, seems to be happy at 3200 with timings relaxed a bit. I am seeing the kind of 2.6x ms/it that I screen grabbed at least once from the "old" RAM. especially if I shut down the cycle hogs. The other biggie in this category is [STRIKE]a browser[/STRIKE] Firefox with multiple tabs and windows open.

I now close everything I can if the system is going to run otherwise unused for some hours (like work or sleep.)

Well, I have the opportunity to exchange RAM again. One of the current modules has gone flaky on me. It throws about 500 errors in the first second of Memtest86 v7.2.

Now I have been digging through web sites trying to find dual rank information. Crucial gives the info for some product lines, but not others. As near as I can tell, in the following part numbers, the "D" preceding the last numeric string signifies Dual.
CT2K8G4DF[U][B]D[/B][/U]824A ** Crucial
CT2K16G4DF[B][U]D[/U][/B]8213 ** Crucial
CT2K16G4DF[U][B]D[/B][/U]824A ** Crucial

On the other hand, CT2K8G4DF[B][U]S[/U][/B]8213 is listed as Single. There are other numbers with a "D" in that position which do not list Rank.

I have a question out to Corsair regarding Dual rank modules.

[QUOTE=kladner;453149]Well, I have the opportunity to exchange RAM again. One of the current modules has gone flaky on me. It throws about 500 errors in the first second of Memtest86 v7.2.[/QUOTE]

Please forgive me for this, but have you run further experiments?

For example, are you absolutely _sure_ that that one module is at fault? Perhaps it's the motherboard or the controller. Or, perhaps, it simply isn't seated properly. Have you tried switching them around and see what happens?

My apologies if I am telling you how to chew gum. But I have found that crying wolf is not wise if a wolf is not present.

[QUOTE=chalsall;453154]Please forgive me for this, but have you run further experiments?

For example, are you absolutely _sure_ that that one module is at fault? Perhaps it's the motherboard or the controller. Or, perhaps, it simply isn't seated properly. Have you tried switching them around and see what happens?

My apologies if I am telling you how to chew gum. But I have found that crying wolf is not wise if a wolf is not present.[/QUOTE]
I appreciate shared experience, which such questions represent.

What I can say is that I started getting BSODs for Memory_Management. This led to web searches, and running Memtest with both sticks in, which caused an instant flood of errors. I tried first one DIMM and then the other in the first slot of the two I use: the second and fourth, as advised in the manual. One could run Memtest through a full cycle with no problem. The other delivered many errors instantly.

I didn't really see the point in going further, since one seemed pretty surely to be the problem.

[QUOTE=kladner;453166]I didn't really see the point in going further, since one seemed pretty surely to be the problem.[/QUOTE]

Agreed. It seems you have run the appropriate tests appropriately.

Sorry for questioning you. But far too often people (including myself) make conclusions without fully exploring the possibility space. You clearly have.

Namastey.

I truly appreciate the thought and effort. You never know if something might have been overlooked. Doesn't hurt to check.

EDIT: I do all the driving in this family. I like to, and Dan hates to drive. However, I work strongly to shut down any irritation I might feel if I have already seen something that Dan points out. I encourage him to say something on the premise, "Four eyes are better than two." Attention can wander on long drives. I much prefer warnings that might be unneeded, to having a crash.
This mindset carries over to other areas. It comes of trying to reduce irritability in general.

[QUOTE=kladner;451113]I exchanged the XMP 2400 parts for XMP 3000. Strangely, running a single, four thread worker is slightly (0.10 to 0.20 ms) slower, now. Would the increased speed show up if I were running more than one worker?

EDIT: I did notice last night that the current RAM shows Single ranks in CPUZ>SPD. I don't know what the previous RAM indicated. I am supposing that it would be the same as these are from the same Corsair Vengeance LPX product line.[/QUOTE]
Short update: it finally hit me that all I had to do was look at my order history. The original Kingston 2133 MHz parts are Dual Ranked. It seems that the Corsair are mostly Single. I am now trying to find something that Microcenter carries, or will order for me. There are Kingston 2400 MHz parts that are dual, though not listed with MC. Certain Crucial parts give the Rank information, and some of those are Dual. Micron makes some, but hardly anyone carries them.

Here is a list I compiled of published Dual rank specs. Double asterisk is Dual.
Kingston HyperX Fury:

HX424C15FBK2/16 **
HX421C14FBK2/16 **
HX424C15FB/16 **
HX424C15FB2K2/16 *
HX424C15FBK2/16 **
HX421C14FB2K2/1 *
HX426C15FBK2/16 **
HX424C15FB/8 **
HX424C15FB/16 **
HX424C15FB2/8 *
HX424C15FB/8 **
HX421C14FB/16 **
HX426C15FBK4/32 **
HX421C14FBK2/32 **
HX426C15FBK4/32 **
HX424C15FBK4/32 **
HX421C14FBK4/32 **
HX424C15FB/16 **
HX421C14FB/16 **
HX426C15FBK2/16 **
HX421C14FBK2/16 **
HX426C15FB/8 **
HX424C15FB/8 **
HX421C14FB/8 **


CT2K8G4DFD824A ** Crucial
CT2K16G4DFD8213 ** Crucial
CT2K16G4DFD824A ** Crucial

HX424C15FBK2/8 * Kingston
KVR21N15D8K2/32 ** Kingston
HX424C15FBK2/8 *
KVR21N15D8K2/32 **
KVR24N17D8/16 **
KVR21N15D8/16 **
KVR24N17S8/8 *

More Dual Rank info
 

Well, Microcenter does not have any of the known dual model numbers posted above. Worse, they don't order stuff that's not already stocked. The only ordering they do is for a listed item that is out of stock.

That said, I am back to ordering from NewEgg, or direct from either Crucial or Kingston. Kingston seems to be the fastest choice available with specific information on Rank. Basically, the Crucial (brand) DIMMs top out at 2400 MHz. Kingston has the same sized parts going to 2666. That, with the dual-rank boost should be impressive.

From earlier in the thread:
[QUOTE]Originally Posted by [B]kladner[/B] [URL="http://www.mersenneforum.org/showthread.php?p=451217#post451217"][IMG]http://www.mersenneforum.org/images/buttons/viewpost.gif[/IMG][/URL]
[I]One factor which might be in play here is that the original XMP-2400 RAM was CAS 14, while the XMP-3000 is CAS 15. The difference in ms/it is 2.628 versus ~2.877.[/I]
[/QUOTE]
I now believe that the real reason is that the very first RAM I got, Kingston DDR4-2133, XMP2400, was dual ranked. I have to surmise that the Corsair modules I replaced them with were single rank.

Microcenter only carries a few Kingston items. I picked that model for low latency. They did not even have it as a kit. I got two individual DIMMs. This is not an issue with either NE or the manufacturer-direct vendors. The kits do exist.

Some links follow. If they end in PDF they are spec sheets for relevant parts.
[url]http://www.kingston.com/dataSheets/HX426C15FBK4_32.pdf[/url]
[url]http://www.kingston.com/dataSheets/HX426C15FBK2_16.pdf[/url]
[url]http://www.kingston.com/dataSheets/HX424C15FB_8.pdf[/url]

This is the part I originally bought 2 of:
[url]http://www.kingston.com/dataSheets/HX421C14FB_8.pdf[/url].

Finally, some screen shots of Compare tables from various sources.

The conclusion I offer is that the HyperX Fury parts are a better choice. The Crucial DIMMs are ultra-retro bare chips on green circuit boards. The Kingstons have nice heat spreaders. :smile:

It ended up being NewEgg. Cheaper and faster than OEM.
HyperX Fury 16GB (2 x 8GB) DDR4 2666 Model# HX426C15FBK2/16-SLV

This is listed twice on NE, with ~$20 difference. I hope my assumption that Sleeved is the equivalent of a bare HDD is correct. That is, same product, less packaging.

I hope to see them tomorrow. I paid for faster shipping because I am coming up on the 30 day limit for store return of the pair with a failure.

If my previous performance comparison of DDR4 2400 dual rank to DDR4 3000 single rank DIMMs is real, then I have some hope that P95 at 2666 MHz will perform on par with 3200 MHz on single rank RAM. :w00t:
I wonder if this carries over to P-1.

Woohoo! Got the RAM, but delivered at work. Have to be patient until I get home ~3:30 PM :whee:

EDIT: Hold the presses! I just realized that the first two examples are of 39.9M exponents. The most recent is 42.5M. If I am not mistaken, this tips the balance even more toward the HyperX 2666 RAM.
See below:

This particular HyperX Fury 2666 is interesting, in that JDEC and XMP specs are identical.

"Sleeved" turned out to be typical rack-hanger packaging, complete with HyperX sticker for your case, if you are so inclined. :smile:

It is running quite well in P95. I have done some Memtest, mainly to see if it crapped out instantly or at least early. ATM, it is running Blend torture test with 15.2 GiB in use overall out of 16 GiB.

I ran it briefly in P95 DC, and it seems to live up to my hopes.

First, here results I can tie to the first pair of 2133, XMP 2400 (dual rank) DIMMS, running DC at 2400 MHz.

Then, there are the best results I could squeeze out of the single rank Corsair at XMP-3000, with the CPU at 4200 MHz instead of 4000. This was pretty much in Clean Boot mode to limit competition.

Finally, there is a quick grab of the current RAM at 2666 MHz with the CPU at 4000.:tu: (...on a larger exponent! :smile:) [with Windows in its typical loaded up state.]

Kingston 2666M dual rank +GTX1060 results
 

attachments:
[I][U]Kingston 2667M.JPG[/U][/I] shows the RAM at its sole published speed.

[I][U]start P95.JPG[/U][/I] shows the change in the change in performance of the 1060 in mfaktc when Prime95 is started, as indicated by the red mark. With previous GPUs (580), and CPUs (AMD FX) starting P95 caused a noticeable [U]decrease[/U] in G-D/D. Here, mfaktc speeds up.
[I][U]
hyperX-2666 at 3000.JPG [/U][/I]shows a snap of P95, with the RAM running at 3000M, at one-clock relaxation of basic timings, and V[SUB]mem[/SUB] bumped from 1.20 to 1.22 V.
This setup ran Blend torture test for ten hours. The RAM amount was set to 13 G (out of 16). FFTs were set at 2048 to 4096.

The final ms/it numbers are great!

Here's the latest tweak. Now the CPU is running at 4200 Mhz, up from 4000. The RAM is at 3000.

This really strange: GTX1060 and 460 together
 

Shortly after I built this system, I decided that running a GTX 580 was getting really tiresome. I looked at the current lineup, and the GTX 1060 really jumped out on price vs production. It is way cheaper than its larger cousins, and is supposed to use less that half of what the 580 sucks down. It also produces substantially more mfaktc G-d/d than the old fire breather.

All that came to pass, with great results, except for one thing. I want to keep running my venerable Gigabyte GTX 460. After some puzzling and testing, I have found that when the Gigabyte 1060 is in residence, the voltage control for the 460 no longer works, in either Afterburner or the Gigabyte competing program. Take out the 1060, and control returns. Worse, at least in Afterburner, saved settings don't carry over when the hardware config changes. So, no getting the 460 set up solo, then plugging in the 1060.

This all means that I can run the 460 at738 MHz, max. ('Stock' OC is 715 MHz.) Under the old regime, I could crank the card to the upper 800 MHz range, for mfaktc. With CUDALucas, I held that to no more than 848 MHz, with more voltage than mfaktc at the same speed, and the VRAM clocked 200 MHz under stock.

I can live with this, but it is an odd hardware interaction.

GeForce® GTX 1060 WINDFORCE OC 6G
 

This GPU surpasses my fondest expectations.

Example: Gigabytes clock specs:
[CODE]Boost: 1797MHz/ Base: 1582MHz in OC Mode
Boost: 1771MHz/ Base: 1556MHz in Gaming Mode[/CODE]I have not altered any settings on the card. Out-of-the-box, it has run mfaktc at 1860 MHz, giving a steady state output of 570 G-D/D. When it starts, it turns out ~595 G-D/D, which decreases over the first few minutes. I have not been able to tune out the decrease with ini tweaks.

Also, in the context of my fan-loaded case, it is nearly inaudible. If I set the fan to 25% in Afterburner, and then run it up to 100%, I hear a slight change in the tone of the overall noise. It never gets close to 100% in everyday use. It is currently at 68%, and the case fans completely cover it. It is running at 70 C, which is a little higher than usual. The apartment is fairly warm at the moment. In cooler conditions it runs 2-3 C cooler.

Kil-a-watt shows that starting mfaktc on the 1060 increases consumption by about 120 W. The GTX 460 shows 140 W on the same test, producing 215 G-D/D2. That is 17% more power for 38% of the output.

Could you post GPU-Z pictures of the first two tabs?

Our reference model runs quite a bit slower and hotter, but we do not have control over any of the parameters. (We could, but we are trying to keep things simple.)

We wish there was a simple fan-speed-only utility. That is all we would change.

[QUOTE=kladner;453961]I have not been able to tune out the decrease with ini tweaks.[/QUOTE]We suspect that the decrease is due to the card warming up. Ours does the same thing.

Here you go!
Correction:
[QUOTE]I have not altered any settings on the card.[/QUOTE]Its fans are under Afterburner control. Return it to its own control, and the fans drop from 70% to 52%, and then step back up to 60% as the temperature rises about one degree C.

[QUOTE=Xyzzy;453972]We suspect that the decrease is due to the card warming up. Ours does the same thing.[/QUOTE]
Indirectly, yes. I found that turning the Power Limit to the max (116%) brought it back a good way (573 v 593 GD/D) toward the startup number. The Temp limit was set to 83 C, and the card never went over 74 C. I turned the limit down to 82 just for good measure.

[QUOTE]We wish there was a simple fan-speed-only utility. That is all we would change.
[/QUOTE]
See if the Nvidia procedure on this page helps. I can't check right now, as I am feeding the monitor from the built-in Intel graphics. Wait.....You run Linux, don't you?
[url]http://smallbusiness.chron.com/adjust-graphics-card-fan-speed-57415.html[/url]
Here it is. I don't suppose the same options are available, but it might give you ideas.

OK. This is instructions for Ubuntu/Linux Mint-
[url]http://www.upubuntu.com/2015/05/how-to-controladjust-gpu-fan-speed-for.html[/url]

This seems to be a productive search area, and I was using DuckDuckGo, not the Goog, which would likely turn up even more.
<graphics fan control [linux] (if appropriate)>

We gave up and installed the EVGA OC tool.

This is interesting:

Set the card to 50% TDP.

430 GHz-d/day at ~125W for the whole computer, with the display asleep. Plus, low noise and nearly no heat.

Now there's something that had not occurred to me. Instead of adjusting core clock to control temperature, as with older cards, just tell it to throttle sooner. "COOL!" :smile: