- Dec 29, 2015
- 2
- 0
- 4,510
GORY DETAILS OF A FUTURE PROOF VR BUILD
or
FLEXIBLE BUDGET AND FUTURE-PROOF VR BUILD
Who am I and what do I want?
Lawyer and future gamer living in Toronto who is looking forward to VR Gaming. I don’t have an unlimited budget, but also don’t want to spend less and find that the new demands of VR render my PC quickly redundant. So I want to scale my budget based on how important any particular component is for future-proofing and performance. I have some hard limits (unless they’re stupid and you feel I should revise them). So spending between $1.5-$2k (not including Oculus/Vive price) would be nice. Spending less would be much better. I want advice on what to buy. My ultimate goal is to do office work on this too and move to a VR Office when that sort of thing is possible…
Primary Question?[/u]
With a budget that scales somewhat, what components should I buy to build a high-end (but not 100% cutting edge) VR-ready PC, keeping in mind my predilection for space sims and strategy games and the physical space (in my living room) for a full ATX board?
1. SUGGESTED BUILD & PRICE (CANADIAN PRICES)
a. Total Build Price
i. ~$2309.69 + Tax.
b. Separate Component Prices
i. Processor – Core i7-6700k - $536.98
ii. CPU Cooler – Cooler Master Hyper 212 EVO 82.9 CFM Sleeve Bearing CPU Cooler - $36.99
iii. Motherboard – MSI Z170A GAMING M5 ATX LGA1151 - $219.98
iv. Memory – G. Skill Ripjaws V Series 16 GB (2x8B) DDR-3200 - $140.98
v. SSD Drive - Samsung 850 EVO-Series 500GB 2.5” SSD - $178.98
vi. Sapphire Radeon R9 Fury X 4GB - $829.00
vii. Fractal Design Case - $173.98
viii. EVGA SuperNOVA G2 650W - $128.80
ix. Microsoft Windows 10 Home OEM (64-bit) $120.98
x. Discounts – Approx. $75.
2. QUICK AND DIRTY CONCLUSIONS (Detailed Discussion Below)
a. i7-6700. Mid-Range Motherboard.
b. Don't buy a GPU until you pre-Order a Rift/Vive. See benchmarks then. Or get an R9 Fury c. Once Pascal/Greenland benchmarks come out, either sell it and buy one or buy a second car for a multi-GPU set-up.
3. SPECS
a. Should this be based on the Oculus Rift/HTC Vive recommended/minimum specs?
i. Fact: It is difficult for developers to know exactly what will be required for playable VR, as it is new.
ii. Fact: VR will demand powerful machines but the exact requirements will vary per game.
iii. Fact: Games have already been released with VR minimum specs greater than set hardware minimum specs (e.g. - Elite Dangerous: Horizons for example).
iv. Conclusion: To be a high-end, future-proof, VR build, it should exceed recommended/minimum specs for VR hardware.
4. GRAPHICS CARD
a. How important is this decision?
i. Fact: Most likely, the GPU will be the most expensive component of a build.
ii. Fact: A GPU or GPUs will most almost certainly be the factor limiting VR, even to the point that a less powerful CPU might be chosen to afford second graphics card (with LiquidVR or Gameworks VR, this now makes great sense for a VR set-up).
iii. Conclusion: The decision of what graphics card to buy will be your most important decision, get it right.
b. Can I take purchase risks or buy a temporary card to be replaced later?
i. Fact: Resale of a decent card and the purchase of a better one can remedy buyers’ remorse, with a penalty.
ii. Fact: Replacing two cards would incur a higher penalty.
iii. Conclusion: You can take risks or buy a temporary graphics card, but it’s NOT advisable to take risks if you’re going all in on a multi-card set-up.
c. Should I prefer Nvidia because it has 81% of the market and AMD is on its last legs?
i. Fact: Remember the Voodoo by 3dfx? In the late 90’s, 3dfx had 80-85% of the market. Better products were released, and Nvidia bought 3dfx. ATI released great products until it put out an over-hyped R8500 (driver problems contributed to its woes then, as they do now) which hurt it. Nvidia released better products (Titanium line) which were easily the best value for money. ATI released better products (9700 Pro). and the companies kept exchanging blows. ATI merged with AMD at one point.
ii. Fact: The discrete GPU market is fluid. Dominance depends on the release of better hardware, software support of that hardware and good strategic moves. The first two are most important.
iii. Fact: 2016 will bring Artic Islands (AMD) and Pascal (NVidia), radically powerful and different architectures, and the die will be thrown again. If AMD releases a dominant, well-supported card then its GPU fortunes will change. With LiquidVR (AMD) and Gameworks VR (NVidia), the software die will also be cast again.
iv. Fact: LiquidVR has received much traction from the VR community (anecdotally more than GameworksVR). The “Ashes of the Singularity” hullaboo, while exaggerated for current titles, seems to suggest that the asynchronous compute architecture of AMD may be better implemented for DX12 and VR titles. This may result in better VR and/or DX12 performance by AMD cards.
v. Fact: AMD’s technology is NOT on its last legs. If there are any more missteps, AMD or its discrete graphics card division may be bought out by a major player (like the Microsoft rumor early in 2015), but the technology will continue to be supported for some time. Also, discrete graphic cards for PCs is not AMDs only market.
vi. Conclusion: Make your GPU purchase based solely on the effectiveness of the hardware and its software support. Consider both AMD and Nvidia.
d. Should I get one top-notch graphics card or use multiple cheaper cards?
i. Fact: Top-end graphics cards entail a premium cost, an extreme one.
ii. Fact: The new LiquidVR and GameworksVR initiatives suggest that a dual-card set-ups, with one card rendering for each eye, may be the only way VR will be playable at sufficient frame rates.
iii. Fact: Though new APIs and drivers may change things, it still seems unlikely that multi-GPU set-ups will be optimized for tri or quad-GPU set-ups.
iv. Conclusion: Plan on a dual-GPU set-up as your end-goal.
e. So what discrete graphics cards should I get and when?
i. Fact: R9 295X2 and GTX Titan Z are the top cards. Generally a premium is paid for same and the premium on the Titan Z make it a ridiculous purchase decision. The R9 295x2 also has a substantial premium that prices it out of range (though far less than the Titan Z)
ii. Fact: Going down just one level in the performance hierarchy will sacrifice very little (to none for some games) in terms of performance but be far, far, cheaper than the top-end cards.
iii. Fact: Both the R9 Fury X and the 980Ti can be doubled-up in a multi-GPU system.
iv. Fact: The R9 Fury X and 980 Ti are both powerful second-tier graphics cards which, on their own, will surpass any current known recommended requirements for VR.
v. Fact: Either the R9 Fury X or 980 Ti will play nearly any game currently made with high settings in HD. The sole question for this build, then, is which will be better for VR.
vi. Fact: With Greenland and Pascal being released next year, the landscape for graphics cards may change drastically. It is also possible, though unlikely, that current performance levels of these (and all lesser) cards will be rendered obsolete.
vii. Conclusion #1: No purchase decision on a graphics card should be made until pre-orders for your preferred VR HMD are made and a shipping date known. A single 980Ti or R9 Fury X should then be purchased based on any data out at the time. If performance is acceptable (which it almost certainly will be), then the purchase of a sister card should be delayed until Greenland or Pascal price points and roadmap are released. At this point, a decision can be made whether to sell the single card and buy a single Greenland/Pascal or to buy a sister card immediately.
viii. Conclusion #2: If you cannot wait, buy a R9 Fury X now. Unless you must play current titles in 4K, then buy a 980Ti. Follow the upgrade path in Conclusion #1.
ix. Conclusion #3: If you can compromise enough to make it really affordable, go ahead and get a dual R9 Fury X set-up now.
5. PROCESSOR
a. How many cores should my CPU have? 4? 6? 8?
i. Fact: Very few gaming PCs will have more than 4 cores. Even enthusiast rigs will usually only have 4 cores.
ii. Fact: For cores to be useful, they must be optimized.
iii. Fact: Almost no game developers optimize games for rare rig set-ups.
iv. Fact: Most game developers will optimize for a maximum of 4 cores.
v. Conclusion: Buy a processor with 4 cores. Any more will not be utilized anywhere near efficiently.
vi. Counters-Arguments:
1. Argument: What about strategy or multi-player games with hundreds to thousands of units on screen?
2. Response: Same logic applies, and if the extra cores were necessary for the game, then it would not have been commercially viable to produce. If it has been produced, it’ll be optimized for 4 cores.
3. Argument: What about VR, will it require more cores?
4. Response: Same argument applies even more strongly, as some developers will fear slow adoption of VR and an already limited market so they’ll optimize for as many as possible.
5. Argument: Early VR adopters will have multi-core PCs already, so won’t VR be targeted towards them?
6. Response: See above. Even VR early adopters are unlikely to have more than 4 cores. VR hardware and software developers more than anyone will want to include as much as possible of a very limited early adopter market or the launch of VR will not be successful. 4 cores will be sufficient.
7. Argument: Don’t higher end 6+ core processors and their chipsets allow for more PCI 3.0 lanes, which would allow for more lanes to be dedicated to GPU’s for future SLI upgrades?
8. Response: While more PCI 3.0 lanes are better than fewer, modern GPU’s will not push enough data to require more lanes than available in common gaming chipsets (supporting 4 cores) and so, as long as there are sufficient lanes for a full SLI configuration, that will be sufficient.
9. Argument: Shouldn’t I plan for a future 3 or 4 card GPU set-up by purchasing 6+ core processors which can handle a tri or quad-SLI configuration?
10. Response: The same logic re CPU core-optimization applies to multi-GPU core optimization. Additionally, following this guide, by the time funds are available for a high-end tri-or quad card set-up, it will likely more efficient simply to sell and upgrade to a newer single or dual-card set-up rather than adding even more mid-level (by that time) cards.
b. What brand of 4-core processor should my PC have?
i. Fact: For some games at the highest resolutions, processors can be a bottleneck, so purchasing a good processor is important.
ii. Fact: The demands of VR are still uncertain, so purchases should err in favor of better processors.
iii. Fact: Strategy games have many units in play, and may require heftier processors, but in general CPUs are only bottlenecks in quite demanding games.
iv. Fact: Intel processors far outclass AMD processors for gaming, the latter which have been known to bottleneck performance. Still, specific AMD processors (such as a higher clocked FX perhaps) can be found that are cheaper and keep up with high-end Intel processors.
v. Conclusion: It’s not exactly clear, so be safe and buy an Intel processor or carefully select a specific very high-end AMD processor (for convenience, I won’t be discussing AMD processors myself).
c. What model of 4-core Intel processor should my PC have?
i. Fact: CPU’s can be a serious bottleneck for gaming framerates. (See Fallout Benchmarks with different CPUs on gamersnexus or techspot SW Battlefront bechmarks).
ii. Fact: VR will require serious processing power, and their CPU processing requirements may be roughly comparable to 4k ready set-ups of today. (See excellent logicalincrements article.)
iii. Fact: VR games recommended requirements are generally higher end processors, but 4k benchmarks seem to indicate that GPU speed will be most important bottleneck, not the CPU. This may be especially true with DX12, which may allow more efficient use of multiple CPU cores.
iv. Fact: If GPUs are the most substantial 4k bottleneck, then it is unlikely that CPU requirements can be specifically determined until higher end GPUs (perhaps Pascal or Greenland) become available to remove the GPU bottleneck.
v. Conclusion: While it is quite likely to be overkill, a very high-end quad-core i7-6700k is commonly recognized as one of the top-end gaming processors. At the least, this overspending will help to keep the system “future-proof”. A compromise would be unlikely to kill the build or its future-proof nature, similar framerates could probably be obtained with an i5 6600k on the same board, or even a mid-range set-up. This would be worth considering especially if the extra funds permitted a dual GPU set-up. The larger cache on i7, however, may make speed up general desktop use and so be otherwise useful.
vi. Counter-Arguments:
1. Argument: Shouldn’t we go for an enthusiast build that would provide for more lanes for peripheral use, like “Haswell,” rather than a “Skylake” i7-6700 which would be limited?
2. Response: A single PCI 3.0 lane can send approx.. 985 MB/s. 8 lanes would have a thoroughput of almost 8 GB/s. In 2013, Anandtech tested whether performance would scale with higher thoroughput and concluded that even a tri-Titan X set-up would not gain much from more lane thoroughput. The same conclusions were reached by guru3d and techpowerup in 2015, for the majority of games. This means that an 8x8 or 16x8 PCIe 3.0 set-up will generally be sufficient, leaving enough lanes over for any desired peripherals even on a Skylake-compliant chipset like Z170 (20 lanes) rather than an X99 (40 lanes). You might leave 5% of performance on the table (on very few games significantly more) but you’ll get more out of the higher processor clock (at a more reasonable price). Generally the requirement for thoroughput would scale most with the frame-rate, which will be difficult to push very much in a demanding VR title.
6. MOTHERBOARD
a. What chipset should I use?
i. Fact: Only the Skylake will support an i7-6700 or i5-6600.
ii. Conclusion: If compromises are NOT being made on the CPU (to purchase an SLI set-up) or extra lanes a priority (see x99 discussion above) then get a Skylake, which will need an LGA 1151 Socket. So you’re looking at a Z170 chipset.
b. What particular features should I look for in a motherboard?
i. M.2 Slots: These are standard now in many motherboards but the drives themselves are prohibitively expensive. Since at some point in the future the price may come down (and you may wish to double, triple or quadruple the speed of your current SSD) the slots should be available for that.
ii. USB 2.0: You’ll need one of these to use the Oculus Rift, HTC Vive will probably have similar requirements.
iii. USB 3.0: Oculus and its touch controllers will together need 3 of these, so they’re recommended. Sometimes (strangely) referred to as USB 3.1 Gen 1.
iv. USB 3.1: Useful as they transfer double the data of a 3.0. Also referred to as USB 3.1 Gen 2 (weird).
v. Space: Motherboard’s, even ATX ones, can be laid out badly so it’s hard to put multiple GPUs on them. You should be careful to ensure this doesn’t become a problem.
c. What motherboard should I get?
i. Fact: Unless you make a really bad error (like picking a motherboard that doesn’t fit your GPUs or one that doesn’t have enough slots or USB 3.0 & 2.0 connections) your VR experience won’t be terribly affected by your choice here. You can save money though by going cheap.
ii. Conclusion: MSI’s Z170A Gaming Pro, MSI's Z170A Gaming M5, ASRock Z170 Extreme6, MSI Z170 Krait Gaming ATX LGA1151 Motherboard, all acceptable. I’d pick the M5 because comparing them is time-consuming.
7. HARD DRIVE
a. Should I buy an SSD or a regular spinning platter drive?
i. Fact: SSD’s are far more reliable and faster than traditional platter drives.
ii. Fact: The price/performance point of the Samsung 850 EVO is superior.
iii. Fact: 500 GB is the minimum storage required for a decent number of games.
iv. Conclusion: Buy a Samsung 850 EVO with at least 500 GB capacity.
8. MEMORY
a. How much and what kind of memory should I buy?
i. Fact: Ram (memory) is now cheap and plentiful.
ii. Fact: DDR4 memory will work on Z170 Boards.
iii. Fact: Modern games will not need more than 16 GB of memory now or in the near future.
iv. Conclusion: Get 16GB of DDR4 RAM.
9. CASE
a. Given a living room set-up, what Case should I buy?
i. Fact: Living room acceptable cases are rare, but Fractal Design makes good ones.
ii. Fact: A case dressed up like a media center could be a good fit for a living room.
iii. Conclusion: Buy a Node 605 from Fractal Design or some similar media center case.
10. POWER SUPPLY
a. What Power Supply should I get?
i. Fact: The noise, temperature and build quality of power supplies are important facts to consider.
ii. Conclusion: Get an EVGA Supernova 850 G2 Power Supply for its silence, or compromise and find a cheaper one.
or
FLEXIBLE BUDGET AND FUTURE-PROOF VR BUILD
Who am I and what do I want?
Lawyer and future gamer living in Toronto who is looking forward to VR Gaming. I don’t have an unlimited budget, but also don’t want to spend less and find that the new demands of VR render my PC quickly redundant. So I want to scale my budget based on how important any particular component is for future-proofing and performance. I have some hard limits (unless they’re stupid and you feel I should revise them). So spending between $1.5-$2k (not including Oculus/Vive price) would be nice. Spending less would be much better. I want advice on what to buy. My ultimate goal is to do office work on this too and move to a VR Office when that sort of thing is possible…
Primary Question?[/u]
With a budget that scales somewhat, what components should I buy to build a high-end (but not 100% cutting edge) VR-ready PC, keeping in mind my predilection for space sims and strategy games and the physical space (in my living room) for a full ATX board?
1. SUGGESTED BUILD & PRICE (CANADIAN PRICES)
a. Total Build Price
i. ~$2309.69 + Tax.
b. Separate Component Prices
i. Processor – Core i7-6700k - $536.98
ii. CPU Cooler – Cooler Master Hyper 212 EVO 82.9 CFM Sleeve Bearing CPU Cooler - $36.99
iii. Motherboard – MSI Z170A GAMING M5 ATX LGA1151 - $219.98
iv. Memory – G. Skill Ripjaws V Series 16 GB (2x8B) DDR-3200 - $140.98
v. SSD Drive - Samsung 850 EVO-Series 500GB 2.5” SSD - $178.98
vi. Sapphire Radeon R9 Fury X 4GB - $829.00
vii. Fractal Design Case - $173.98
viii. EVGA SuperNOVA G2 650W - $128.80
ix. Microsoft Windows 10 Home OEM (64-bit) $120.98
x. Discounts – Approx. $75.
2. QUICK AND DIRTY CONCLUSIONS (Detailed Discussion Below)
a. i7-6700. Mid-Range Motherboard.
b. Don't buy a GPU until you pre-Order a Rift/Vive. See benchmarks then. Or get an R9 Fury c. Once Pascal/Greenland benchmarks come out, either sell it and buy one or buy a second car for a multi-GPU set-up.
3. SPECS
a. Should this be based on the Oculus Rift/HTC Vive recommended/minimum specs?
i. Fact: It is difficult for developers to know exactly what will be required for playable VR, as it is new.
ii. Fact: VR will demand powerful machines but the exact requirements will vary per game.
iii. Fact: Games have already been released with VR minimum specs greater than set hardware minimum specs (e.g. - Elite Dangerous: Horizons for example).
iv. Conclusion: To be a high-end, future-proof, VR build, it should exceed recommended/minimum specs for VR hardware.
4. GRAPHICS CARD
a. How important is this decision?
i. Fact: Most likely, the GPU will be the most expensive component of a build.
ii. Fact: A GPU or GPUs will most almost certainly be the factor limiting VR, even to the point that a less powerful CPU might be chosen to afford second graphics card (with LiquidVR or Gameworks VR, this now makes great sense for a VR set-up).
iii. Conclusion: The decision of what graphics card to buy will be your most important decision, get it right.
b. Can I take purchase risks or buy a temporary card to be replaced later?
i. Fact: Resale of a decent card and the purchase of a better one can remedy buyers’ remorse, with a penalty.
ii. Fact: Replacing two cards would incur a higher penalty.
iii. Conclusion: You can take risks or buy a temporary graphics card, but it’s NOT advisable to take risks if you’re going all in on a multi-card set-up.
c. Should I prefer Nvidia because it has 81% of the market and AMD is on its last legs?
i. Fact: Remember the Voodoo by 3dfx? In the late 90’s, 3dfx had 80-85% of the market. Better products were released, and Nvidia bought 3dfx. ATI released great products until it put out an over-hyped R8500 (driver problems contributed to its woes then, as they do now) which hurt it. Nvidia released better products (Titanium line) which were easily the best value for money. ATI released better products (9700 Pro). and the companies kept exchanging blows. ATI merged with AMD at one point.
ii. Fact: The discrete GPU market is fluid. Dominance depends on the release of better hardware, software support of that hardware and good strategic moves. The first two are most important.
iii. Fact: 2016 will bring Artic Islands (AMD) and Pascal (NVidia), radically powerful and different architectures, and the die will be thrown again. If AMD releases a dominant, well-supported card then its GPU fortunes will change. With LiquidVR (AMD) and Gameworks VR (NVidia), the software die will also be cast again.
iv. Fact: LiquidVR has received much traction from the VR community (anecdotally more than GameworksVR). The “Ashes of the Singularity” hullaboo, while exaggerated for current titles, seems to suggest that the asynchronous compute architecture of AMD may be better implemented for DX12 and VR titles. This may result in better VR and/or DX12 performance by AMD cards.
v. Fact: AMD’s technology is NOT on its last legs. If there are any more missteps, AMD or its discrete graphics card division may be bought out by a major player (like the Microsoft rumor early in 2015), but the technology will continue to be supported for some time. Also, discrete graphic cards for PCs is not AMDs only market.
vi. Conclusion: Make your GPU purchase based solely on the effectiveness of the hardware and its software support. Consider both AMD and Nvidia.
d. Should I get one top-notch graphics card or use multiple cheaper cards?
i. Fact: Top-end graphics cards entail a premium cost, an extreme one.
ii. Fact: The new LiquidVR and GameworksVR initiatives suggest that a dual-card set-ups, with one card rendering for each eye, may be the only way VR will be playable at sufficient frame rates.
iii. Fact: Though new APIs and drivers may change things, it still seems unlikely that multi-GPU set-ups will be optimized for tri or quad-GPU set-ups.
iv. Conclusion: Plan on a dual-GPU set-up as your end-goal.
e. So what discrete graphics cards should I get and when?
i. Fact: R9 295X2 and GTX Titan Z are the top cards. Generally a premium is paid for same and the premium on the Titan Z make it a ridiculous purchase decision. The R9 295x2 also has a substantial premium that prices it out of range (though far less than the Titan Z)
ii. Fact: Going down just one level in the performance hierarchy will sacrifice very little (to none for some games) in terms of performance but be far, far, cheaper than the top-end cards.
iii. Fact: Both the R9 Fury X and the 980Ti can be doubled-up in a multi-GPU system.
iv. Fact: The R9 Fury X and 980 Ti are both powerful second-tier graphics cards which, on their own, will surpass any current known recommended requirements for VR.
v. Fact: Either the R9 Fury X or 980 Ti will play nearly any game currently made with high settings in HD. The sole question for this build, then, is which will be better for VR.
vi. Fact: With Greenland and Pascal being released next year, the landscape for graphics cards may change drastically. It is also possible, though unlikely, that current performance levels of these (and all lesser) cards will be rendered obsolete.
vii. Conclusion #1: No purchase decision on a graphics card should be made until pre-orders for your preferred VR HMD are made and a shipping date known. A single 980Ti or R9 Fury X should then be purchased based on any data out at the time. If performance is acceptable (which it almost certainly will be), then the purchase of a sister card should be delayed until Greenland or Pascal price points and roadmap are released. At this point, a decision can be made whether to sell the single card and buy a single Greenland/Pascal or to buy a sister card immediately.
viii. Conclusion #2: If you cannot wait, buy a R9 Fury X now. Unless you must play current titles in 4K, then buy a 980Ti. Follow the upgrade path in Conclusion #1.
ix. Conclusion #3: If you can compromise enough to make it really affordable, go ahead and get a dual R9 Fury X set-up now.
5. PROCESSOR
a. How many cores should my CPU have? 4? 6? 8?
i. Fact: Very few gaming PCs will have more than 4 cores. Even enthusiast rigs will usually only have 4 cores.
ii. Fact: For cores to be useful, they must be optimized.
iii. Fact: Almost no game developers optimize games for rare rig set-ups.
iv. Fact: Most game developers will optimize for a maximum of 4 cores.
v. Conclusion: Buy a processor with 4 cores. Any more will not be utilized anywhere near efficiently.
vi. Counters-Arguments:
1. Argument: What about strategy or multi-player games with hundreds to thousands of units on screen?
2. Response: Same logic applies, and if the extra cores were necessary for the game, then it would not have been commercially viable to produce. If it has been produced, it’ll be optimized for 4 cores.
3. Argument: What about VR, will it require more cores?
4. Response: Same argument applies even more strongly, as some developers will fear slow adoption of VR and an already limited market so they’ll optimize for as many as possible.
5. Argument: Early VR adopters will have multi-core PCs already, so won’t VR be targeted towards them?
6. Response: See above. Even VR early adopters are unlikely to have more than 4 cores. VR hardware and software developers more than anyone will want to include as much as possible of a very limited early adopter market or the launch of VR will not be successful. 4 cores will be sufficient.
7. Argument: Don’t higher end 6+ core processors and their chipsets allow for more PCI 3.0 lanes, which would allow for more lanes to be dedicated to GPU’s for future SLI upgrades?
8. Response: While more PCI 3.0 lanes are better than fewer, modern GPU’s will not push enough data to require more lanes than available in common gaming chipsets (supporting 4 cores) and so, as long as there are sufficient lanes for a full SLI configuration, that will be sufficient.
9. Argument: Shouldn’t I plan for a future 3 or 4 card GPU set-up by purchasing 6+ core processors which can handle a tri or quad-SLI configuration?
10. Response: The same logic re CPU core-optimization applies to multi-GPU core optimization. Additionally, following this guide, by the time funds are available for a high-end tri-or quad card set-up, it will likely more efficient simply to sell and upgrade to a newer single or dual-card set-up rather than adding even more mid-level (by that time) cards.
b. What brand of 4-core processor should my PC have?
i. Fact: For some games at the highest resolutions, processors can be a bottleneck, so purchasing a good processor is important.
ii. Fact: The demands of VR are still uncertain, so purchases should err in favor of better processors.
iii. Fact: Strategy games have many units in play, and may require heftier processors, but in general CPUs are only bottlenecks in quite demanding games.
iv. Fact: Intel processors far outclass AMD processors for gaming, the latter which have been known to bottleneck performance. Still, specific AMD processors (such as a higher clocked FX perhaps) can be found that are cheaper and keep up with high-end Intel processors.
v. Conclusion: It’s not exactly clear, so be safe and buy an Intel processor or carefully select a specific very high-end AMD processor (for convenience, I won’t be discussing AMD processors myself).
c. What model of 4-core Intel processor should my PC have?
i. Fact: CPU’s can be a serious bottleneck for gaming framerates. (See Fallout Benchmarks with different CPUs on gamersnexus or techspot SW Battlefront bechmarks).
ii. Fact: VR will require serious processing power, and their CPU processing requirements may be roughly comparable to 4k ready set-ups of today. (See excellent logicalincrements article.)
iii. Fact: VR games recommended requirements are generally higher end processors, but 4k benchmarks seem to indicate that GPU speed will be most important bottleneck, not the CPU. This may be especially true with DX12, which may allow more efficient use of multiple CPU cores.
iv. Fact: If GPUs are the most substantial 4k bottleneck, then it is unlikely that CPU requirements can be specifically determined until higher end GPUs (perhaps Pascal or Greenland) become available to remove the GPU bottleneck.
v. Conclusion: While it is quite likely to be overkill, a very high-end quad-core i7-6700k is commonly recognized as one of the top-end gaming processors. At the least, this overspending will help to keep the system “future-proof”. A compromise would be unlikely to kill the build or its future-proof nature, similar framerates could probably be obtained with an i5 6600k on the same board, or even a mid-range set-up. This would be worth considering especially if the extra funds permitted a dual GPU set-up. The larger cache on i7, however, may make speed up general desktop use and so be otherwise useful.
vi. Counter-Arguments:
1. Argument: Shouldn’t we go for an enthusiast build that would provide for more lanes for peripheral use, like “Haswell,” rather than a “Skylake” i7-6700 which would be limited?
2. Response: A single PCI 3.0 lane can send approx.. 985 MB/s. 8 lanes would have a thoroughput of almost 8 GB/s. In 2013, Anandtech tested whether performance would scale with higher thoroughput and concluded that even a tri-Titan X set-up would not gain much from more lane thoroughput. The same conclusions were reached by guru3d and techpowerup in 2015, for the majority of games. This means that an 8x8 or 16x8 PCIe 3.0 set-up will generally be sufficient, leaving enough lanes over for any desired peripherals even on a Skylake-compliant chipset like Z170 (20 lanes) rather than an X99 (40 lanes). You might leave 5% of performance on the table (on very few games significantly more) but you’ll get more out of the higher processor clock (at a more reasonable price). Generally the requirement for thoroughput would scale most with the frame-rate, which will be difficult to push very much in a demanding VR title.
6. MOTHERBOARD
a. What chipset should I use?
i. Fact: Only the Skylake will support an i7-6700 or i5-6600.
ii. Conclusion: If compromises are NOT being made on the CPU (to purchase an SLI set-up) or extra lanes a priority (see x99 discussion above) then get a Skylake, which will need an LGA 1151 Socket. So you’re looking at a Z170 chipset.
b. What particular features should I look for in a motherboard?
i. M.2 Slots: These are standard now in many motherboards but the drives themselves are prohibitively expensive. Since at some point in the future the price may come down (and you may wish to double, triple or quadruple the speed of your current SSD) the slots should be available for that.
ii. USB 2.0: You’ll need one of these to use the Oculus Rift, HTC Vive will probably have similar requirements.
iii. USB 3.0: Oculus and its touch controllers will together need 3 of these, so they’re recommended. Sometimes (strangely) referred to as USB 3.1 Gen 1.
iv. USB 3.1: Useful as they transfer double the data of a 3.0. Also referred to as USB 3.1 Gen 2 (weird).
v. Space: Motherboard’s, even ATX ones, can be laid out badly so it’s hard to put multiple GPUs on them. You should be careful to ensure this doesn’t become a problem.
c. What motherboard should I get?
i. Fact: Unless you make a really bad error (like picking a motherboard that doesn’t fit your GPUs or one that doesn’t have enough slots or USB 3.0 & 2.0 connections) your VR experience won’t be terribly affected by your choice here. You can save money though by going cheap.
ii. Conclusion: MSI’s Z170A Gaming Pro, MSI's Z170A Gaming M5, ASRock Z170 Extreme6, MSI Z170 Krait Gaming ATX LGA1151 Motherboard, all acceptable. I’d pick the M5 because comparing them is time-consuming.
7. HARD DRIVE
a. Should I buy an SSD or a regular spinning platter drive?
i. Fact: SSD’s are far more reliable and faster than traditional platter drives.
ii. Fact: The price/performance point of the Samsung 850 EVO is superior.
iii. Fact: 500 GB is the minimum storage required for a decent number of games.
iv. Conclusion: Buy a Samsung 850 EVO with at least 500 GB capacity.
8. MEMORY
a. How much and what kind of memory should I buy?
i. Fact: Ram (memory) is now cheap and plentiful.
ii. Fact: DDR4 memory will work on Z170 Boards.
iii. Fact: Modern games will not need more than 16 GB of memory now or in the near future.
iv. Conclusion: Get 16GB of DDR4 RAM.
9. CASE
a. Given a living room set-up, what Case should I buy?
i. Fact: Living room acceptable cases are rare, but Fractal Design makes good ones.
ii. Fact: A case dressed up like a media center could be a good fit for a living room.
iii. Conclusion: Buy a Node 605 from Fractal Design or some similar media center case.
10. POWER SUPPLY
a. What Power Supply should I get?
i. Fact: The noise, temperature and build quality of power supplies are important facts to consider.
ii. Conclusion: Get an EVGA Supernova 850 G2 Power Supply for its silence, or compromise and find a cheaper one.
Facebook
Twitter
Instagram