Advice on a new computer system

[jonasthomas]

Hello,

 

I would like to buy a new computer and I was wondering if someone could help me out in this jungle of numbers. :-)

 

My budget maximum for everything (including two screens, a mouse and a keyboard) is about 2000 euros.

The tasks I want to use the computer for are (and in this order):

 

 

- Sketchup

- Maxwell Render

- 3DS Max (I am not using this right now, but probably in the future)

- Vray

- Vue

- maybe some games too, but that's not important :-)

 

 

 

This is the setup that I have come up with so far:

 

- Processor: Intel Core i5 3570K Boxed

- Processor cooler: Gelid Solutions Tranquillo Rev.2

- Motherboard: Gigabyte GA-Z77X-D3H

- Internal memory: 4 x 4GB PC3-12800 (what brand?)

- Solid state drive: Samsung 830 series 256GB

- Hard drive: Seagate Barracuda 7200.14 ST2000DM001, 2TB

- Video card: Nvidia GTX 670 (what architecture brand?)

- Case: Antec Eleven Hundred

- Power: Corsair Enthusiast TX750M

 

What do you guys think?

 

Many thanks in advance!

 

Pieter

Edited September 30, 2012 by jonasthomas

[Dimitris Tolios]

Overall looks fine.

If you don't plan on overclocking - or at least to push a lot ouf of your chip, everything looks fine.

 

-Cooler: If you don't want to OC at all, then the stock intel cooler will serve you just fine.

 

-Memory: I would get 2x8GB sticks, to allow for headroom in upgrading to 32GBs - if ever needed. Preferably low profile or no Heatspreaders at all, to allow for CPU cooler clearance with no problems. PC 1600 like the one you suggest will do fine, 1866 is also pretty close in prices (at least in the US), but I would not go higher than that and honestly, you won't see performance increases outside of benchmarks etc going little bit higher than that. G.Skill Ares, Corsair Vengeance LP, or crucial and Kingston...all brands are producing reliable products (its RAM, if it works, it works!).

 

-SSD: Great choice. 840 is on the way, so keep your eyes open for dealer offers on the 830. At least those (the offers) are awaited here in the states for October-November.

 

-HDD: Get the best warranty - all disks (WD/Seagate/Samsung) are pretty close in performance.

 

-GPU: don't overspend. The GPU is the part most likely to be the least utilized by the programs you've mentioned outside of games. The 670 will be kinda overkill for most of them: no real performance benefits over slower GTX cards, no real need for Games unless you ABSOLUTELY have to have the best etc...for 1080p-1200p (single monitor) gaming, I believe you will have no issue opting for a GTX 660ti 2GB or AMD 7950 3GB. Both those are 20-25% cheaper than a 2GB 670.

The 7950 will have probably better performance in both Sketchup and Vue, as the Radeons like OpenGL viewports much better than Geforce cards. In gaming it is also a beast.

 

By saving those $100/€100 off the card, you might shoot for a 2600K/3770K i7, and overall you will have a better all-around system for CPU rendering.

[jonasthomas]

Hello Dimitri,

 

Thanks a lot for replying! It is much appreciated!

 

Cooler: Does using the "Turbo" to 3,8 Ghz count as overclocking?

 

Memory: Do you think 2 x 8GB is as fast as 4 x 4 GB of RAM, or is it slower?

 

GPU: If the Radeons are better for OpenGL I will opt for this. Do you know if 3D applications benefit when there are two graphics cards installed please?

 

Thanks!

[Dimitris Tolios]

- Tubro boost is overclocking, but not past factory specs - as it is defined by the factory and the factory cooler then sized accordingly. So in the scope of my comment, it is not.

 

- 2x8 sticks are at least as good as 4x4, usually better - at least with the LGA 1155 platform where chipsets support up to dual channel ram. You usually won't have a issue running them both at the factory timings etc, but if you would overclock, most motherboards/memory controllers can achieve more aggressive timings the fewer sticks you are trying to push.

 

- 3D applications do not care (yet) for more than one GPU, unless it is for computation tasks, so again we are talking GPU accelerated renderers and not Sketchup / 3DS / Maya etc. None of these cares about SLI/ Crossfire X, etc: will only use your primary card.

[jonasthomas]

Thanks Dimitris! That's of great help.

 

One more question: I can see that these HD 7950 GPU cards only have 1 DVI-I out port. I want to hook up two monitors. Is this possible?

[Dimitris Tolios]

Thanks Dimitris! That's of great help.

 

One more question: I can see that these HD 7950 GPU cards only have 1 DVI-I out port. I want to hook up two monitors. Is this possible?

 

Yes. Most 79xx reference Radeons have one DVI, along with 2x mini display ports (DP) and 1x HDMI. You can hook one monitor @ each port, so up to 4 monitors are simultaneously supported. Adapters and/or cables with different ends exist for HDMI -> DVI / DVI -> HDMI and miniDP/DP for both, so you can do a lot of combinations.

Edited October 1, 2012 by dtolios

[jonasthomas]

Ok, thanks, another thing I know.

 

Do you know where I can find some information about the OpenGL performance/ benchmarks of the different GPU cards?

[Dimitris Tolios]

That's a hard subject. Almost no benchmark collection/review/site database etc objectively tests consumer GPUs on viewport performance, especially in OpenGL, as when they do, they do 3DS. Maya, Blender, Sketchup etc is not comparably tested most of the times.

 

OpenGL gaming performance is also something you cannot directly correlate to view-port performance in 3D modelling programs. i.e. : if say the X card is Y% faster than the Z card in an Unreal Engine based game for example, doesn't mean it will be close to that faster in Maya, and vise versa.

 

You can read this forum post for OpenGL viewport performance and the 7970.

 

 

I am of the opinion that professional cards are nothing really special as far as hardware goes: yes, the GPU processor might be handpicked after QC a production batch and it might be a slight tad more stable than the average consumer version of the same chip, but if you consider what % of the QC certified production ends up in high-end Quadro's or FireGLs, with 1000s of the latter sold vs. millions of the high-end gaming cards, it is safe to conclude that the chance of you getting a "less stable" consumer card is pretty slim, unless nothing but the top 1/1000 % of chips out there is actually "good".

 

It is like the "Starbuck stores have sourced and sell only the best coffee beans in the world" claim: with a franchise that grows 100s if not 1000s of stores every year, it is safe to assume that by comparison, the coffee will be getting worse and worse every year, as the "best coffee beans" are no longer enough, and compromises have to be made.

 

Some - not all - pro cards do use ECC ram. Maybe critical for certain extremely sensitive computation processes (aka irrelevant to mainstream 3D modelling before 2-3 years or less), doubtfully of any real meaning for 3D application viewport performance and most likely GPU rendering aswell, were the result is the sum of multiple passes for each pixel were random mistakes get averaged out.

 

In a nutshell, variation exists (in absolute stability/heat production/speed), but it is not tangible, and ofc exists within the Pro cards whether we like it or not. It's all in the set tolerances, as nothing is perfect or even exactly the same down to each silicon molecule.

 

What really sets the cards apart, is software: BIOS and Drivers recognize the software and treated accordingly. Thus gaming cards with virtually immensely better technical characteristics, get outclassed in view-port performance by Pro cards, while when under unbiased rendering computation tasks, the gaming cards resume their role and smoke everything but the best of the "pro" cards which might barely catch up (simply because even if the Quadro/FireGL card is based on the same core architecture, is usually a tad slower in clock speeds to ensure "improved stability" - ofc the gaming cards would have increased stability and better longevity running @ lower speeds, but I haven't seen anybody complaining about gaming cards dying prematurely, even tho such speculations have been made in bulk when studios started implementing GPU renderings through GTX cards few years ago).

 

Still it doesn't change the fact that consumer gaming cards are not really tested by pro reviewers on professional applications, and when they do test them, it is usually vs. a pro card of similar or better specs on paper.

Few amateurs can assemble a solid testing process with multiple scenarios, multiple design suites and more than a couple - at best - pro cards, and usually both nVidia and AMD (ATI) will not supply them for free if you express the will to assemble such a test.

Edited October 1, 2012 by dtolios

[jonasthomas]

Thanks for your thorough explanation Dimitris! If I understand this well, it's very hard to compare graphic cards (consumer and pro) due to a lack of professional benchmarks. I think I will go for a consumer card then, like a GTX or a Radeon, because at least I then know that it's good for games and there is larger "support" for them (if I run into technical troubles for example)

[Dimitris Tolios]

Thanks for your thorough explanation Dimitris! If I understand this well, it's very hard to compare graphic cards (consumer and pro) due to a lack of professional benchmarks. I think I will go for a consumer card then, like a GTX or a Radeon, because at least I then know that it's good for games and there is larger "support" for them (if I run into technical troubles for example)

 

I would not go to a Pro card unless I know it will increase my productivity that much that it will pay for itself.

For unpaid 3D stuff, it is nearly impractical, as the cards that will give you enough "horsepower" for that usually are in the $600+ range (mid-range Pro cards) as the entry level ones do not pull ahead of the good consumer enough to justify the expense.

 

Also keep in mind that viewport performance does not scale well with the consumer cards: e.g. a 560ti being 2 times slower than a 680 as games would suggest (and theoretical computation power alike), translates to almost no difference in viewport performance. Computation is better (Mari/VRay RT etc will like the upgrade) but practically the drivers/way the program interacts with the card is the limiting factor and that caps well before the cards are maxed.

 

For your application, a consumer card is def. recommended, especially if you plan on gaming: i.e. unless you want to spend the same money you would spend on a 7970 and get a Quadro 2000 which is a more or less a GT 450 hardware wise.

[Tom Bussey]

I'm looking to buy a similarly specced system, so would appreciate any input. Here's a list as I have it so far. Any obvious compatability issues / total overspecing would be good to have pointed out to me. I'll be using it to train myself up on 3DS + Vray which I use quite lightly at the moment. I'll also be running Photoshop and AutoCAD. The prices are rough estimates collected from Google, I'm pretty much at the top of my budget here.

 

One thing I don't know is where do audio outputs and USB sockets come in, I'm not interested in surround sound or anything but could do with a few ins and outs, also would like USB 3 / Firewire.

 

CPU: INTEL CORE I7-3770 - £200

GPU: Nvidia GTX 670 4gb - £330

RAM: Corsair Memory Vengeance Low Profile Black 16GB DDR3 1600 MHz - £60

Motherboard: ASUS Maximus IV Extreme -Z - £200

PSU: Antec True Power Quattro 1200 - £150

Case: Antec Eleven Hundred - £90

Internal HD (I have a 3TB external for archiving): Western Digital 1tb internal - £45

SSD: Crucial RealSSD M4 256GB - £55

OS: windows 8 - £45

 

Any help is very much appreciated.

[Dimitris Tolios]

A few tips here and there:

 

GPU - I feel kinda guilty as I've mentioned the GTX 670 4GB way to often in these forums. Unless you will be rendering crazy models through VRay RT GPU, Mari or other GPU accelerated rendering or computing apps (never tried CS6 to know if even that will use it), the only reason to get a card with more than 2GB of VRam is for gaming in some crazy multi-monitor setup with max FXAA settings etc.

 

For "ACAD and Photoshop", most likely any card in the $150 range with 1GB of RAM or more would do just fine, even with dual 1080p/1200p or even dual 1440p monitors.

 

A valuable tool for seeing what you will most likely be using, is TechPowerUp GPU-Z. You can load it in your current system, fire up one or even more of your most demanding files and watch how much VRam is required in a typical workflow scenario you will subject your card to. Most likely you won't see more than 200MB of use for heavy ACAD drawings (correct me if I'm wrong), and even 3DS won't really care for more or even close to 1GB unless you do really demanding stuff.

 

Windows 8: I would not rush to get those before a few patches at least...waiting for the SP1 is maybe too much, then again I would wait for at least a few months after those are out for fully validating their compatibility.

 

Mobo: Maximus 4...wow...where did you come up with that? First of all it is 2+ years old tech. Z77 is much more feature packed than both the Z68 and the P67. I would not get a P67 mobo for a brand new 3770K even if I was saving a lot of money on it - paying premium price for an outdated "enthusiast" mobo is no-no for me. Any P8Z77 mobo will do fine and will have equal or better features than the "high-end" P67 mobo with the 2-3 extra controllers natively through the Z77.

 

If you want to O/C decently, go for a P8Z77-V Pro.

If you want slightly more, a Gene V will do great. Yes it is mATX and you don't need more than that, unless you plan on using more than 2x PCIx 16x...highly doubt it.

 

PSU: 1200W...crazy overkill. Any good 550-650W will work with your system just great, with potential even for 2x GPUs if not O/Ced. A 750W will allow easily for 2x high-end modern GPUs. 1200W would be needed if you were using an overclocked 3930K with Tri-SLI 580 GTX or something equally propostrusly energy hogging like that. Newer systems get lower in energy requirements. Getting a 1200W PSU is not future-proofing. If you have the money opt for a high quality 80+ Gold or Platinum 650-800W PSU and you have more than enough headroom (most likely even the 750-800W will be working @ 50-60% with 2x GPUs).

[Tom Bussey]

Brilliant stuff, thanks for that, very helpful stuff. As said I am totally ignorant when it comes to a lot of these things, motherboards especially confuse the hell out of me as there is not just a comparable set of numbers in the name. I'll digest all of that and have another go. Thanks again!

[dhammikaherath]

Processor

i7-3770K Ivy Bridge 3.5 GHz (3.9 Turbo) Quad-Core with 4000 Graphics

Motherboard

Asus P8Z77-I DELUXE Z77/SATA3 & USB3.0/WiFi & Bluetooth Mini-ITX

Memory

Mushkin Enhanced 32 GB (4x8GB) PC3-12800 (10-10-10-27)

Heatsink

Asetek 570 LC Extreme Liquid CPU Cooler + Dual BitFenix 120’’ Fans

Video Card

EVGA GeForce GTX 660 Ti 2 GB GDDR5 PCI Express 3.0 SLI Ready SLI

Primary Storage

Mushkin Chronos Deluxe 120 GB SATA III Solid State Drive (SSD)

Secondary Storage

WD Caviar Black 1 TB 7200 RPM/64 MB Cache SATA 6.0Gb/s

DVD

ASUS DRW-24B1ST 24X Internal DVD+/-RW Drive (Black)

Power Supply

CORSAIR GAMING SERIES GS800 800W ATX12V V2.3 SLI READY CROSSFIRE READY 80 PLUS CERTIFIED ACTIVE PFC HIGH PERFORMANCE POWER SUPPLY

Chassis

BITFENIX COLOSSUS BLACK ATX FULL TOWER WINDOW CASE RED LED EXTERIOR

OS

Windows 7 Pro 64 Bit

 

I think above spec will fit with your budget...!

[Tom Bussey]

Looks good, thanks, I'll total that up and see how it comes out when I get home. Out of interest, what would be the main advantage of the P8Z77-I Deluxe over the P8Z77-V Pro?

[Dimitris Tolios]

Size...it is a mITX board (smallest possible capable of a i7...).

I would say it was a mistake, otherwise why chose a mITX board + a full ATX tower?

[dhammikaherath]

I think you will have to go with ASUS P8z77-V Pro. That supports upto 32 GB.

 

http://www.asus.com/Motherboards/Intel_Socket_1155/P8Z77V_PRO/#specifications

 

http://www.asus.com/Motherboards/Intel_Socket_1155/P8Z77I_DELUXE/#specifications

 

in here price is around $250. But you will have to check the price in UK. I'm recommending full size tower for reason of future upgrading & better ventilation for long hour rendering.

[Tom Bussey]

OK, that looks good to me, and should come in at about £1100 if I decide to get someone else to build it. I'm going to go with Kingston KHX1600C9D3K8 Ram as it's on the recommended list for the motherboard (32Gb almost certainly overkill, but it can't hurt), and Antec Eleven Hundred case as I prefer it's plain look. Now to try and find as few shops as possible to source the bits from....

 

 

Thanks for you help.

[Dimitris Tolios]

More cliche SSD recommendations:

I would highly recommend a Crucial m4 256SSD - here in the US we get amazing deals on it @ $160 quite often lately.

King of the hill is still the Samsung 830 as the all around most reliable and fast SSD, a bit pricier than $160 (more in the $200 range, but there are offers for it lower than that), but still solid.

 

256GB SSDs are not only faster than the 120/128 variants, but allow you a lot of headroom to install pretty much everything on them, including your design suites and a few current projects for a complete storage boost without micromanaging few GBs here and there.

[jonasthomas]

By saving those $100/€100 off the card, you might shoot for a 2600K/3770K i7, and overall you will have a better all-around system for CPU rendering.

 

Hey Dimitris,

 

I wished I would have followed your advice more closely. I have bought the i5 3570 processor, but I saw later on the Maxwell Render Benchwell website that the i7 3770k does the same render about 25% faster. I guess replacing the CPU is not really an option since it is not easy to do and I lose the money I spent on the i5 3570. I should have listenend. :-)

[Dimitris Tolios]

Weeeeellll,

Thought we have established that the i7 is rarely that much faster than the i5, but when it comes to rendering or video encoding/transcoding it simply is! =)

 

Upgrading the CPU is not that difficult - especially with intel CPUs where there are no CPU pins to bend and damaging something is very difficult - as long as you realize the basic concept in PC components coming together: those are delicate instruments, precisely made - if they don't snap together effortlessly, usually it means you are doing something wrong/the other way around - it almost NEVER means "push a bit harder to get it in"

 

In you case, the upgrade procedure should be:

Open case -> remove heatsink -> open socked -> remove i5 -> drop in i7 correctly aligned -> close socket -> re-install heatsink (or just install the new stock one).

 

The i5 could be re-sold as used in ebay or in classified/used ads in forums etc, or returned for a refund if possible.

 

I know it is hard to "follow" advice, and by no means me or anybody else in forums is infallible, but I do my best of unbiased answers/advice.

[jonel]

You'll need a double ammount of RAM for i7 vs. i5.

[Dimitris Tolios]

You'll need a double ammount of RAM for i7 vs. i5.

To run 8 vs 4 threads? I don't think so...

[jonel]

Try it, don't think

[Dimitris Tolios]

Well, sorry but what you claim is extremely exaggerated.

Yes, some renderers like Mental Ray do get a mem-usage increase for every thread you add, but that's not linear. If it was, how could we possibly utilize GPU rendering with cards that have hundreds or thousands of processing threads fed by 1-2GB of RAM?

 

The actual RAM overhead for extra threads is a pretty small percentage of the total.

For most renderers, and I believe Vray it is not even an issue. The Vray frame buffer alone eats so much more memory in large resolutions that doesn't even make sense talking about overhead due to extra threads.

 

Truth is that for most rendererers, the amount of RAM needed for processing is simply much the same, regardless of the processes happening at a given moment, as the bits being processed are amazingly small in comparison with the textures, light cache and geometry information that accounts for the bulk of the committed RAM. There are dozens of techniques recommended for optimizing the RAM usage when rendering, none of them involves "disabling HT" or reducing your threads at any way.