How Does One Choose a Windows PC Computer ?

1235»

Comments

  • Thread again trimmed. If the covnersation continues to fall short of acceptable degrees of civility towards and respect for others it will be locked.

  • bluejauntebluejaunte Posts: 1,990

    To the OP I'd say try to find a local PC builder with a good reputation and enough hardware options and good warranty. This way you can pretty much get the benefits of having state of the art components without having to build anything yourself.

  • ebergerlyebergerly Posts: 3,255
    edited January 2019
    FWIW, I threw some more reasonable numbers into the PS calcs, and if you assume a more reasonable 2% efficiency difference at 400 watts, and assume youre operating at full load for 5 hours a day all year, the total yearly wasted $$ due to efficiency differences is less than $2. Again, pretty much irrelevant. And assuming computers generally hang around for only 5 years or so, thats only like $10 over its lifetime. I think you'd save a LOT more electricity $$ by turning off the 100watt bathroom light when youre not using it.
    Post edited by ebergerly on
  • kenshaw011267kenshaw011267 Posts: 3,805
    edited January 2019

    Actually at a 2% difference running 24/7 which is completely reasonable for a render box, mine does that, it adds up to more than the cost difference of the more efficient higher capacity psu. 

    But what do I know I just build out datacenters for a living.I

    and you low balled the power consumption and used average cost per kw/h so if the user is over that low end number during rendering or lives in almost any urban area, since the US average is based of much lower rural rates, then the savings will be higher and then the savings over the life of the psu will be higher.

    But again what do I know I just have an entire industry backing me up on this.

    Post edited by kenshaw011267 on
  • hjakehjake Posts: 1,274
    edited January 2019

    Every computer I've ever bought has been an Apple Mac of one kind or another - since they began making them.  Now I suppose I need a Windows PC with an nvida card just to do iRay renders.  How do you choose which Windows PC to buy?  With Apple there is usually about 12 choices.  With windows machines it looks like there are 1200 choices.  Is there one brand that is overwhelimingly superior or reliable?

    Thanks!

    Late to the conversation but hopefully not too late. If you are an Applephile, then coming over the to the light side of bright open windows world can be daunting.  My first recommendation would be to educate yourself about computer hardware and assemble it yourself or have a local computer store assemble it for your. My experience is that people who buy Apple PCs aren't willing to invest the time required to learn what they will need and would prefer to pay someone else to tell them what will meet their needs and have it just work. For those people, who are also focused buying on a limited budget will be best served by Dell/Alienware. The closest experience you will get to buying Apple.

    If you require a quality machine that can meet your requirements and are willing to pay a premium for a top notch PC then here are some companies that can cater to your requirements. People here can tell which parts to pick. In the end this PC will cost you like an Apple PC but you will get a quality PC with a solid warranty.

    Origin PC, Puget Systems, Falcon Northwest, Digital Storm, Xidax, Cyberpower, Velocity Micro, and others.

    .*********************************

    On another note. No matter how powerful the PC with uber Nvidia cards you will not create a render farm for quick 5 minute renders in your home. You can do more of "press button and make art" with Iray than you could before but I feel people oversell the fast render times. Faster, yes. Lightning fast, no.  Scene optimization, rendering in layers, efficient use of lights, and postwork in GIMP/Photoshop are still WAY more important than an uber PC. I mention this because I see people spend fortunes on PCs to uber render and are disappointed because it can still take hours to render a scene.

     

     

     

     

    Post edited by hjake on
  • Takeo.KenseiTakeo.Kensei Posts: 1,303
    edited January 2019

    Actually at a 2% difference running 24/7 which is completely reasonable for a render box, mine does that, it adds up to more than the cost difference of the more efficient higher capacity psu. 

    But what do I know I just build out datacenters for a living.I

    and you low balled the power consumption and used average cost per kw/h so if the user is over that low end number during rendering or lives in almost any urban area, since the US average is based of much lower rural rates, then the savings will be higher and then the savings over the life of the psu will be higher.

    But again what do I know I just have an entire industry backing me up on this.

    I'm with ebergerly on this point. Personnal use and industry use are different case. First, the max efficiency at 50% load rule you apply is not always true. Also most people certainly don't render 24/7. We're not talking about large economy scale with 1000+ PCs here. And if you'd really like to nitpicking, you'd have to control your PSU every year to be sure that it still works at it's best efficiency and buy a new one if it's not

    Really the discussion about the PSU should stop because it is really not relevant to the thread question which is not "which PSU to buy for unkown working condition". A 80+ PSU with enough power for the system and eventually 250W margin if planning an other gfx card in the future should be good. I personnaly don't buy under gold certification and not under 700W but that's personnal taste and not a necessity for most people

    Back to the thread, any big name like ASUS, MSI, Dell (especially Alienware), HP should be OK if you buy products in the gaming area. Special attention should be given to the PSU and motherboard specs if planning future upgrade. For two gfx card, a +700W should be OK. Special mention for RTX cards for people who want to SLI two cards to (perhaps) benefit of RAM spooling : you need a motherboard that can do 8x/8x for the PCIe 3.0 x16 slots. The simple rule would be to check for the SLI certificate. Despite what some people think, all motherboard that have two or more PCIe x16 slots are not all capable of SLI (some have a mix of PCIe 2.0 and PCIe 3.0 or some only do 8x/4x). If SLI is not planned then any board with adequate number of PCIe x16 slot should do.

    Some people also mentionned eGPU. That is in my POV a valid solution especially for Mac users because they won't have to learn a new OS (and they usually don't have the desire to do so because they think Windows is inferior). There are a lot of solutions nowadays, from cheap (less than 20$) to expensive (~1000$). There are a lot of infos here https://egpu.io/external-gpu-buyers-guide-2019/#tb3-enclosures. There are also lots of example builds with explanations and scripts to male it work for MacOS here https://egpu.io/build-guides/ . Some Mac users reported to have Iray working. One of the drawback is that depending on the system, data transfer speed may be average, but that point is not that important for Iray outside of SLI. The other big drawback is that Apple is actally blocking driver development for it's new OS and may be also for RTX cards, so you can only buy cards up to Pascal. That may not be a long term solution if Apple stays on it's position

    Post edited by Takeo.Kensei on
  • On what to buy. A 2080 ti suggests a 600w psu minimum. Each can draw up to 250w. So if your going with 2 cards I think an 850w is a sensable precaution.
  • Actually at a 2% difference running 24/7 which is completely reasonable for a render box, mine does that, it adds up to more than the cost difference of the more efficient higher capacity psu. 

    But what do I know I just build out datacenters for a living.I

    and you low balled the power consumption and used average cost per kw/h so if the user is over that low end number during rendering or lives in almost any urban area, since the US average is based of much lower rural rates, then the savings will be higher and then the savings over the life of the psu will be higher.

    But again what do I know I just have an entire industry backing me up on this.

    I'm with ebergerly on this point. Personnal use and industry use are different case. First, the max efficiency at 50% load rule you apply is not always true. Also most people certainly don't render 24/7. We're not talking about large economy scale with 1000+ PCs here. And if you'd really like to nitpicking, you'd have to control your PSU every year to be sure that it still works at it's best efficiency and buy a new one if it's not

    Really the discussion about the PSU should stop because it is really not relevant to the thread question which is not "which PSU to buy for unkown working condition". A 80+ PSU with enough power for the system and eventually 250W margin if planning an other gfx card in the future should be good. I personnaly don't buy under gold certification and not under 700W but that's personnal taste and not a necessity for most people

    Back to the thread, any big name like ASUS, MSI, Dell (especially Alienware), HP should be OK if you buy products in the gaming area. Special attention should be given to the PSU and motherboard specs if planning future upgrade. For two gfx card, a +700W should be OK. Special mention for RTX cards for people who want to SLI two cards to (perhaps) benefit of RAM spooling : you need a motherboard that can do 8x/8x for the PCIe 3.0 x16 slots. The simple rule would be to check for the SLI certificate. Despite what some people think, all motherboard that have two or more PCIe x16 slots are not all capable of SLI (some have a mix of PCIe 2.0 and PCIe 3.0 or some only do 8x/4x). If SLI is not planned then any board with adequate number of PCIe x16 slot should do.

    Some people also mentionned eGPU. That is in my POV a valid solution especially for Mac users because they won't have to learn a new OS (and they usually don't have the desire to do so because they think Windows is inferior). There are a lot of solutions nowadays, from cheap (less than 20$) to expensive (~1000$). There are a lot of infos here https://egpu.io/external-gpu-buyers-guide-2019/#tb3-enclosures. There are also lots of example builds with explanations and scripts to male it work for MacOS here https://egpu.io/build-guides/ . Some Mac users reported to have Iray working. One of the drawback is that depending on the system, data transfer speed may be average, but that point is not that important for Iray outside of SLI. The other big drawback is that Apple is actally blocking driver development for it's new OS and may be also for RTX cards, so you can only buy cards up to Pascal. That may not be a long term solution if Apple stays on it's position

    ?

    Yes, the 50% load rule is always the case. Every PC PSU sold is most efficient at 50% load, period. This is the second time someone has claimed this isn't true in this thread and it's ridiculous that people are claiming this. Not only does every PSU peak in efficiency at 50% max load every PSU has so close to the same efficiency curve that the same curve is used for all of them.

    The reason folks at datacenters know so much about this is because we have thousands of machine to monitor and track the data from and we pay the power bill for all of that and how minor inefficiencies show up very strongly when you have hundreds of identical machines running side by side. That doesn't mean that doesn't matter for the home PC user. It just means it will take the lifetime of the system to matter.

    You then recommend buying a prebuilt. Please just stop. Nothing good can come from that.

  • On what to buy. A 2080 ti suggests a 600w psu minimum. Each can draw up to 250w. So if your going with 2 cards I think an 850w is a sensable precaution.

    250 + 250 + 100 (about the minimum for the rest of a system) = 600 watts. 850 watts would be a 70.5% load. That's not going to be good if you run at that load full time. 1000 or 1100 watt platinum or titanium would be better. Once you're spending $2600 on graphic cards an extra $30 on a decent PSU seems pretty trivial to me.

  • Takeo.KenseiTakeo.Kensei Posts: 1,303

    Actually at a 2% difference running 24/7 which is completely reasonable for a render box, mine does that, it adds up to more than the cost difference of the more efficient higher capacity psu. 

    But what do I know I just build out datacenters for a living.I

    and you low balled the power consumption and used average cost per kw/h so if the user is over that low end number during rendering or lives in almost any urban area, since the US average is based of much lower rural rates, then the savings will be higher and then the savings over the life of the psu will be higher.

    But again what do I know I just have an entire industry backing me up on this.

    I'm with ebergerly on this point. Personnal use and industry use are different case. First, the max efficiency at 50% load rule you apply is not always true. Also most people certainly don't render 24/7. We're not talking about large economy scale with 1000+ PCs here. And if you'd really like to nitpicking, you'd have to control your PSU every year to be sure that it still works at it's best efficiency and buy a new one if it's not

    Really the discussion about the PSU should stop because it is really not relevant to the thread question which is not "which PSU to buy for unkown working condition". A 80+ PSU with enough power for the system and eventually 250W margin if planning an other gfx card in the future should be good. I personnaly don't buy under gold certification and not under 700W but that's personnal taste and not a necessity for most people

    Back to the thread, any big name like ASUS, MSI, Dell (especially Alienware), HP should be OK if you buy products in the gaming area. Special attention should be given to the PSU and motherboard specs if planning future upgrade. For two gfx card, a +700W should be OK. Special mention for RTX cards for people who want to SLI two cards to (perhaps) benefit of RAM spooling : you need a motherboard that can do 8x/8x for the PCIe 3.0 x16 slots. The simple rule would be to check for the SLI certificate. Despite what some people think, all motherboard that have two or more PCIe x16 slots are not all capable of SLI (some have a mix of PCIe 2.0 and PCIe 3.0 or some only do 8x/4x). If SLI is not planned then any board with adequate number of PCIe x16 slot should do.

    Some people also mentionned eGPU. That is in my POV a valid solution especially for Mac users because they won't have to learn a new OS (and they usually don't have the desire to do so because they think Windows is inferior). There are a lot of solutions nowadays, from cheap (less than 20$) to expensive (~1000$). There are a lot of infos here https://egpu.io/external-gpu-buyers-guide-2019/#tb3-enclosures. There are also lots of example builds with explanations and scripts to male it work for MacOS here https://egpu.io/build-guides/ . Some Mac users reported to have Iray working. One of the drawback is that depending on the system, data transfer speed may be average, but that point is not that important for Iray outside of SLI. The other big drawback is that Apple is actally blocking driver development for it's new OS and may be also for RTX cards, so you can only buy cards up to Pascal. That may not be a long term solution if Apple stays on it's position

    ?

    Yes, the 50% load rule is always the case. Every PC PSU sold is most efficient at 50% load, period. This is the second time someone has claimed this isn't true in this thread and it's ridiculous that people are claiming this. Not only does every PSU peak in efficiency at 50% max load every PSU has so close to the same efficiency curve that the same curve is used for all of them.

    Sure, check these one and tell us that again. https://www.tomshardware.com/reviews/pc-power-and-cooling-silencer-1050w-psu,5871-5.html

    https://www.tomshardware.com/reviews/thermaltake-toughpower-grand-rgb-850w-gold-psu,5822-5.html

    https://www.tomshardware.com/reviews/antec-hcg1000-extreme-1000w-psu,5701-5.html

    There are many PSU for which your rule is not true. If I'm the second one to tell you you're wrong, there is a reason.

    The reason folks at datacenters know so much about this is because we have thousands of machine to monitor and track the data from and we pay the power bill for all of that and how minor inefficiencies show up very strongly when you have hundreds of identical machines running side by side. That doesn't mean that doesn't matter for the home PC user. It just means it will take the lifetime of the system to matter.

    You then recommend buying a prebuilt. Please just stop. Nothing good can come from that.

    I'm not recommending a prebuilt, but that's what the OP asked. So I answer his question...

    Is there one brand that is overwhelimingly superior or reliable?

  • Actually at a 2% difference running 24/7 which is completely reasonable for a render box, mine does that, it adds up to more than the cost difference of the more efficient higher capacity psu. 

    But what do I know I just build out datacenters for a living.I

    and you low balled the power consumption and used average cost per kw/h so if the user is over that low end number during rendering or lives in almost any urban area, since the US average is based of much lower rural rates, then the savings will be higher and then the savings over the life of the psu will be higher.

    But again what do I know I just have an entire industry backing me up on this.

    I'm with ebergerly on this point. Personnal use and industry use are different case. First, the max efficiency at 50% load rule you apply is not always true. Also most people certainly don't render 24/7. We're not talking about large economy scale with 1000+ PCs here. And if you'd really like to nitpicking, you'd have to control your PSU every year to be sure that it still works at it's best efficiency and buy a new one if it's not

    Really the discussion about the PSU should stop because it is really not relevant to the thread question which is not "which PSU to buy for unkown working condition". A 80+ PSU with enough power for the system and eventually 250W margin if planning an other gfx card in the future should be good. I personnaly don't buy under gold certification and not under 700W but that's personnal taste and not a necessity for most people

    Back to the thread, any big name like ASUS, MSI, Dell (especially Alienware), HP should be OK if you buy products in the gaming area. Special attention should be given to the PSU and motherboard specs if planning future upgrade. For two gfx card, a +700W should be OK. Special mention for RTX cards for people who want to SLI two cards to (perhaps) benefit of RAM spooling : you need a motherboard that can do 8x/8x for the PCIe 3.0 x16 slots. The simple rule would be to check for the SLI certificate. Despite what some people think, all motherboard that have two or more PCIe x16 slots are not all capable of SLI (some have a mix of PCIe 2.0 and PCIe 3.0 or some only do 8x/4x). If SLI is not planned then any board with adequate number of PCIe x16 slot should do.

    Some people also mentionned eGPU. That is in my POV a valid solution especially for Mac users because they won't have to learn a new OS (and they usually don't have the desire to do so because they think Windows is inferior). There are a lot of solutions nowadays, from cheap (less than 20$) to expensive (~1000$). There are a lot of infos here https://egpu.io/external-gpu-buyers-guide-2019/#tb3-enclosures. There are also lots of example builds with explanations and scripts to male it work for MacOS here https://egpu.io/build-guides/ . Some Mac users reported to have Iray working. One of the drawback is that depending on the system, data transfer speed may be average, but that point is not that important for Iray outside of SLI. The other big drawback is that Apple is actally blocking driver development for it's new OS and may be also for RTX cards, so you can only buy cards up to Pascal. That may not be a long term solution if Apple stays on it's position

    ?

    Yes, the 50% load rule is always the case. Every PC PSU sold is most efficient at 50% load, period. This is the second time someone has claimed this isn't true in this thread and it's ridiculous that people are claiming this. Not only does every PSU peak in efficiency at 50% max load every PSU has so close to the same efficiency curve that the same curve is used for all of them.

    Sure, check these one and tell us that again. https://www.tomshardware.com/reviews/pc-power-and-cooling-silencer-1050w-psu,5871-5.html

    https://www.tomshardware.com/reviews/thermaltake-toughpower-grand-rgb-850w-gold-psu,5822-5.html

    https://www.tomshardware.com/reviews/antec-hcg1000-extreme-1000w-psu,5701-5.html

    There are many PSU for which your rule is not true. If I'm the second one to tell you you're wrong, there is a reason.

    The reason folks at datacenters know so much about this is because we have thousands of machine to monitor and track the data from and we pay the power bill for all of that and how minor inefficiencies show up very strongly when you have hundreds of identical machines running side by side. That doesn't mean that doesn't matter for the home PC user. It just means it will take the lifetime of the system to matter.

    You then recommend buying a prebuilt. Please just stop. Nothing good can come from that.

    I'm not recommending a prebuilt, but that's what the OP asked. So I answer his question...

    Is there one brand that is overwhelimingly superior or reliable?

    Did you miss that those are the same curves? The exact peaks of the curve are just whether the PSU is a gold, platinum or whatever. I mean really did you look at them at all? I didn't expect you to do regression analysis on them but just glabcing at them shows that they are the same shape.

  • Takeo.KenseiTakeo.Kensei Posts: 1,303

    Yes I did and the point is not the shape of the curve but the 50% peak rule. Tell me for each of them if that applies or not

  • It applies to all of them. You're not actually talking about their low power performance are? You notice Tom's results weren't consistent run to run right? That the numbers were off by 10% run to run? PSU's simply cannot be reliably measured that low. But sure run a 1k PSU at 10% if it makes you happy.

    You might want to actually look into what they're all certfiied to do. There's a reason the 80+ certs start at 20% load and always set the 20% efficiency equal to the 100% efficiency. But go ahead keep telling me my field some more.

  • kyoto kidkyoto kid Posts: 41,861

    ...umm, I sense Ricahrd heading this way with lock in hand.

  • Locked.

This discussion has been closed.