pyrojoe34

So I've upgraded my Server (Dell PowerEdge T610) from a single E5620 (Quad core) to two E5645 (hexa core) CPUs. I'm wondering what the best way to config my RAM is between the two systems. I have 2 kits: 3x2GB ECC kit and 2x8GB non-ECC. Does each CPU need to have a matching RAM config? The way I see it these are my options: Option 1: Each CPU gets one 8GB stick (total 16GB) Option 2: Each CPU gets one 8GB stick and one 2GB stick (total 20GB) Option 3: CPU1 gets 2x8GB and CPU2 gets 3x2GB (total 22GB) Is option 3 even possible? Can I mix and match like that and if I do will CPU2 still run tri-channel while CPU1 runs dual channel? If not, I'm thinking option 2 is the best balance between capacity and balance. Thanks

This is highly dependent on the algorithms and datasets you use. There is no answer to this question. The bottom line is: where is the bottleneck? Do you run out of RAM when loading assets? If not then capacity is not the issue. Are your cores skipping cycles while waiting to load data from the RAM to the cache? If not then bandwidth is not the issue. Is your GPU skipping cycles while it waits for instructions from the CPU? Then your CPU is the limitation. There is not good way to answer this without either running diagnostic codes during a run or just trying different configs and seeing what matters most. You might also want to see if your algorithms can use mixed precision (i.e. tensor cores) which would give you a >10x improvement in FLOPS over traditional comp methods. If you can use that tech and can implement it efficiently you might want to consider Volta-based cards (Tesla V100 or Titan V) for much faster training/execution.

I've been pretty happy with the CyberPower 1500PFCLCD. 900W max, pure sine wave rather than simulated, 1500VA, 5 battery outlets + 5 surge outlets + 2 USB ports. Also reports power input/output and other info.

I'm using an old single core Atom netbook w/ 2Gb RAM as a (kinda) low-power Plex server for my music library. Old, slow laptops still have their uses for some people.

Okay, so it seems like your computer is not shutting off or freezing, it's just that your peripherals are "disconnecting". In that case it sounds like an issue with your motherboard/USB ports. First, reset BIOS to defaults and see if it still happens. Then, try to install the newest USB drivers for the board from the manufacturer's site. If that doesn't do it you might need to try flashing the BIOS to the newest revision. If all of those fail to fix it it might mean the problem is hardware-based and the board needs to be RMAed.

What? That's a bit confusing... when you say your character keeps running you mean that's what your friends see, not you right? Okay so your computer completely turns off, your screen is black, fans all off, and the power light on your PC is off? That's most likely a PSU issue then, it might be faulty. Do you have another you can test out?

Does it shut off or does it freeze? Because you said it still displays the image on the screen which means it's not off but frozen. Try reinstalling your GPU drivers.

Sounds like it's crashing. Is anything overclocked? It might be unstable.

Yea, resolution scale is your issue. You were basically playing at 1440p, not 1080p. In most cases there's no reason to go above 1 (also listed as 100% in some games) for resolution scaling, you should just stick with your monitor's native resolution. It is basically the most taxing/inefficient version of anti-aliasing.

Just let the system manage it and don't worry about it. Your page file is not only used for "extra" memory but it's a copy of what is in your RAM at the moment. This is because RAM is wiped when it loses power so in the event of a sudden crash there is a copy of your RAM backed up to the drive so you don't lose everything or corrupt your files. The page file usage is almost always larger than RAM use, this is normal, don't worry about it, it is not impacting your performance.

A screenshot doesn't give a great idea of overall load since it might not be representative of the actual usage when you took the screenshot. Do you have a tool that will graph per thread use over a period of time? How often are individual threads hitting 100%? Thread 1 looks to be over 75%, is getting to 100% often? How's your drive use during those moments? Is it bottle-necked trying to load assets to RAM?

Look at the task manager. What is using the disk? Are updates running? It's possible the disk is faulty, check the SMART status and drive health.

As far as I know that's a feature that would have to be in the HDMI switch and not a thing you can change in the software. The HDMI switch would have to change which input gets sent to the output without turning off the other inputs. Basically, it's a hardware thing... I don't think you can fix that.

Install Tensor Flow and run some massive datasets through that sucker. Or better yet download some RNAseq data from the NCBI database and see how long it takes to run sequence alignment on a gamut of total RNAseq FASTA datasets. I'll have some RNAseq experiments coming up in the next month or so (should have 24-52 mRNAseq profiles to run), and you can put my own system to shame with the alignment runtime on those. You can use the results in this study to compare your system in million read pairs/hr.

Hey so those motherboard temps over 100 are not real sensors. The only real sensors on the board is the CPU socket and the Aux temp. I also have an MSI board that shows one sensor at like 125C and I confirmed with MSI that there are only 2 hardware temp sensors on the board and the others are software bugs. Nothing to worry about. Having said that, you should really have at least one case fan in there to keep your chipset and VRMs cool.

Your issue lies somewhere else, the cooler should be just fine for that. I'm guessing you have a bad mount on the block or possibly your pump is failing. The delid will help but if you're getting 80+C at 1.375V then you have bad heat transfer to the block as well. Make sure the pump is running and try remounting the block.

To be fair, people upgrade monitors much less often than GPUs. If I buy a nice monitor I want it to last 5-10yrs before replacement/upgrade. The idea is to buy a monitor that will work for the next 3-4 systems. So even if OP can only push 60fps with a modern build, at least this way you don't have to buy a new monitor in a couple years when you can get higher framerates.

Thanks, I'll give it a try. Follow-up question: Do you think I can mix and match the two sets (the board has 6 DIMM slots per socket, I am only using one CPU for now) to get 22GB? I realize it won't be ECC but do you think it will work? Would it still run dual channel or would it force default to single-channel? I know I can just test it myself but I figure I'd ask. Even that is more than I want to spend right now... I'll wait til I can get another 6+gb ECC (in tri-channel) for <$15. I can add another e5620 to the board for about $6 to make it an 8c/16t setup (actually the most expensive item would be a T610 compatible heatsink that fits in the airflow shroud). I'm a grad student and money is tight...

I have a poweredge T610 (w/ e5620) and I am using as a NAS (running FreeNAS 11.1) but it only has 3x2gb ECC which is very quickly saturated (and doesn't really leave me much room to run any VMs), I have 2x8gb non-ECC sticks laying around unused and I'm wondering if it's worth switching to those or if it's too risky (all the data on this NAS is backed up in at least 2 other locations so this is not mission critical, just a backup/file share server). What do you guys think? If RAM prices ever drop I'll upgrade the RAM with ECC but for now the prices are just dumb (and I got the system for free and it's slightly older tech so I don't want to put any significant $ into it).

That's pretty normal no need to worry.

Yes, That CPU will be a bottleneck on basically any modern games.

No that's the speed of the integrated GPU (on the CPU). There is no limit to the GPU speed on a dedicated GPU. You are fine to pair it with any card.

Which is also only true if you are comparing cores of identical design. I can show you a 3.5Ghz quad core that will completely destroy a 5Ghz 8-core in every performance test. Processor performance is so much more complex than a simple reductionist equation using frequency and core/thread count.

You always do remote desktop to raspberry pi/old computer or just use something like the Steam Link setup. The problem with display signals over ethernet is they are very compressed signals. Typical ethernet gives you a max bandwidth of 1Gb/s (up to 10Gb/s if your network can do it but most can't and every device would have to be able to do it for it to work) and an display signal running 1080p/60 is ~3.8Gb/s, 1440p/60 is ~7Gb/s and 4k/60 is ~16Gb/s. So basically you will have to compress the video if you use ethernet (plus the large increase in latency).

I would personally wait but it depends on how desperate you are for a card. IMO even if you don't end up getting next gen (expected in April I believe), the market will be flooded with cheap Pascal cards when they release so you'll come out on top either way.