Destroy my build - $3,000 gaming setup

[mfenn]

I got my 550D today, and the engineering on it is fantastic. I can see the logic behind everything they did.

Here's a diagram of the fan setup I'm planning, and some thoughts below it:

• The 120mm front intakes would both be NF-S12A fans - these are low impedance fans that should simply bring air into the case.
• The 140mm bottom intake would be an NF-A14. The blade design on this should create a more focused, straight-line type of flow. So, not only would it draw cool air in from the bottom, it should also direct the cool air from the 120mm front fans towards the top of the case.
• The rear 140mm exhaust would also be an NF-A14. Not much to say here. My NH-D14 CPU cooler should pull in the cool air from the front, exhaust it towards the back, and then the NF-A14 rear exhaust should pull it out of the case.

There we go. Is this logical?

I'm running 3 NF-S12A FLX's with the 7V adapters in the standard spots (read: a very quiet but hardly any airflow) and don't have any cooling issues.

Is this going on a soft surface? If so, the bottom fan isn't going to do anything because the feet on the 550D aren't terribly tall. Otherwise, it sounds fine, but the bottom fan isn't really necessary IMHO. No reason not to run the fan if you already have it though.

 

[PCJake]

I'm running 3 NF-S12A FLX's with the 7V adapters in the standard spots (read: a very quiet but hardly any airflow) and don't have any cooling issues.

Is this going on a soft surface? If so, the bottom fan isn't going to do anything because the feet on the 550D aren't terribly tall. Otherwise, it sounds fine, but the bottom fan isn't really necessary IMHO. No reason not to run the fan if you already have it though.

It's going on a wood floor. It should be able to pull air from the bottom OK.

I'll test it with and without the bottom fan. If it makes no difference then I'll leave it off, just for the sake of simplicity and tidiness.

 

[PCJake]

Update: I've moved all my components into my new 550D case, and I couldn't be happier with the results; temps, noise, and cable management are all improved compared to my Cosmos 1000 setup. The Seasonic X650 PSU is also perfect for me. I've updated my OP with additional thoughts and details on these new parts.

 

[mfenn]

Why did you take the side panel off? That pretty much completely negates the positive pressure design (hard to maintain a pressure differential with a wall missing!).

 

[Termie]

I'm really glad to hear that the 550D worked out so well for you and that it didn't seem like a downgrade from the Cosmos. Just goes to show that a solid design counts for a lot.

I will agree with mfenn, however, that removing the side panel probably doesn't help, but only you can be sure. Have you tested temps with and without the panel?

 

[PCJake]

Why did you take the side panel off? That pretty much completely negates the positive pressure design (hard to maintain a pressure differential with a wall missing!).

Just to be sure we're on the same page: I took the side panel fan cover off, not the panel itself.

Temps seem to be lower by a couple degrees with the cover off (though, admittedly, I haven't tested this thoroughly yet), and I'm just going for lots of airflow and ventilation; I'm not really concerned with having one sort of pressure or another, as long as temps and noise are well balanced.

I'll do CPU and GPU stress tests, both with and without the cover on. Then I shall analyze the results...with science!

I'm really glad to hear that the 550D worked out so well for you and that it didn't seem like a downgrade from the Cosmos. Just goes to show that a solid design counts for a lot.

Thanks - I agree. Being smart about a build is much more effective than buying the largest, most expensive case or getting a 1200W PSU just because you can. It's more satisfying to know that your setup is perfect for your own purposes.

I will agree with mfenn, however, that removing the side panel probably doesn't help, but only you can be sure. Have you tested temps with and without the panel?

See above Further testing will definitely take place.

 

[PCJake]

Update: I've run temperature tests with and without the side panel fan cover on:

CPU load test (Prime 95, run three times with the cover on, and three times without; each time for five minutes - temps listed are averaged from hottest core)

Fan cover on: 71.3°C avg. min., 77.3°C avg. max.
Fan cover off: 70.3°C avg. min., 76.3°C avg. max.

GPU load test (Unigine Heaven, run three times with the cover on, and three times without; each time for five minutes)

Fan cover on: 85°C avg. max., 64% avg. fan speed
Fan cover off: 85°C avg. max., 61% avg. fan speed

So, there we have it; with the fan cover off, CPU load temps dropped in every test by 1°C, and the GPU fan didn't have to work as hard. Noise is only slightly more noticeable with the cover off. If the noise starts to bother me, I'll put the cover on it, otherwise I'll just leave it off.

 

[Sleepingforest]

Yay, science!

 

[mfenn]

Just to be sure we're on the same page: I took the side panel fan cover off, not the panel itself.

Oh, I thought you took the whole panel off. It should indeed be slightly cooler with that panel off, at the cost of more dust and noise.

I love a low-noise PC, so I would gladly trade 1 deg C for any reduction in noise level. Everybody's different though.

 

[PCJake]

Well, everyone, I've been doing some hardcore research on the 4770K, and for the first time in my history of building computers, I am NOT excited about this CPU. The heat issues are just unbelievable, assuming even half of what people are saying is true. I know it would be faster per clock than my i7-930, but it looks like I wouldn't be able to get much more than the 4.2GHz I'm already getting on my 930--and, so, here's the big picture, in terms I've what I'd be spending vs. what I'd be getting:

• $350 for a processor that would be slightly faster and definitely hotter than my current one.
• $200 for Z87 mobo. As far as I can tell, the only advantage of switching to Z87 from my X58 (besides Haswell support) is native SATA GB/s ports (which I actually really, really wanted) for my Samsung 840 Pro 256GB SSD.
• $100 for a different sound card. Didn't know they were going to phase out PCI slots completely, though it makes sense.
• $100 for a new Windows 7 key (maybe, depending on whether or not Microsoft lets me use my old one after switching mobos. Kind of depends on what sort of mood they're in, I think).

So, there you have it: $750. For that, the only real benefits are a faster CPU (a benefit that is pretty much negated by the extra heat output, as far as I'm concerned) and a better chipset that natively supports Sata 6GB/s.

On the bright side, my new Seasonic PSU and Corsair 550D case are awesome, and I'll be grabbing a superclocked GTX 780 as soon as they're in stock. But the Haswell thing is still bumming me out. Here's what I want to ask you guys:

1. I'm currently using my 840 Pro SSD on my mobo's native 3.0GB/s ports, because random reads and writes were very poor when I tested it on the Marvell 6.0GB/s ports. Running it on the 3.0GB/s ports improves random speeds, but they still bottleneck this particular SSD overall. Is there anything I can do to improve this situation, short of getting a different chipset? Perhaps something along the lines of installing some Marvell drivers (haven't tried that before)?
2. Speculation time: anyone think Intel will release a real enthusiast chip (one with heat and overclocking as top priorities) within the next year/year-and-a-half? Perhaps the Haswell-E?

 

[crashtech]

The only thing on the horizon is IB-E, which many have preemptively dismissed, but now that Haswell has shown such minimal improvements over IB, perhaps IB-E's third quarter launch will get more attention. It's got me curious. I may just get an LGA2011 system if IB-E makes a good showing.

As far as overclocked quad cores go, the 3770K is still the king of the hill, unless some binned 4770K chips come out that can start making some better numbers.

 

[mfenn]

For that, the only real benefits are a faster CPU (a benefit that is pretty much negated by the extra heat output, as far as I'm concerned) and a better chipset that natively supports Sata 6GB/s.

This is hysteria speaking IMHO. A couple of points:

1. That your 920 is a 130W chip. When overclocked, I would not be surprised if you were dumping 160W+ into the case.

2. Temperature is not the same thing as heat. The die of Haswell is much much smaller than the die of your Nehalem. Simple physics tells you that it will run hotter (more energy per unit area). A new chip is not going to solve this.

Also, you can get to 4.2 on Haswell without doubling the TDP. A Haswell is something like 50% faster than your Nehalem at equivalent clocks.

 

[crashtech]

Anything looks good against what he has, and if the OP wasn't going to overclock, the decision would be somewhat simpler, definitely in favor of some Haswell flavor.

Haswell's got the IPC, but does it have the speed? It doesn't look like it (so far).

 

[PCJake]

The only thing on the horizon is IB-E, which many have preemptively dismissed, but now that Haswell has shown such minimal improvements over IB, perhaps IB-E's third quarter launch will get more attention. It's got me curious. I may just get an LGA2011 system if IB-E makes a good showing.

I actually hadn't considered the September IB-E. I would definitely be willing to get an LGA2011 system. It's worth $200 extra to get a CPU that's actually designed for desktops. It'll all come down to the overclocking and heat performance, so we'll see!

Temperature is not the same thing as heat. The die of Haswell is much much smaller than the die of your Nehalem. Simple physics tells you that it will run hotter (more energy per unit area). A new chip is not going to solve this.

You're saying there aren't issues with the TIM and IHS? If it were just the smaller die, then how come people are taking the freakin' lid off 'em right away to get decent temps? "Simple physics," *harrumph*...

But word it however you want; I just don't want a processor that gets so hot it throttles or crashes under a mild overclock, and I'm not going to put my CPU in a clamp and pull the top off.

Also, you can get to 4.2 on Haswell without doubling the TDP. A Haswell is something like 50% faster than your Nehalem at equivalent clocks.

Double the TDP? Who's doubling the TDP?...

I looked at those charts. The 4770K is 25-50% faster when the i7-920 is at 2.66GHz and the 4770K is at 3.5GHz..where are you seeing them compare the two CPUs at equivalent clocks?

And I've already made the point about the Haswell being faster at equivalent frequencies, but it's still a mobile-oriented processor. I'm simply holding out to see if Intel is going to release a more desktop-oriented processor.

Anything looks good against what he has

Wait a sec, I don't follow. Are you talking about my i7-930 (I'm not being sarcastic, just curious)?

WHAT THE #$@* IS WRONG WITH MY 930 :twisted::twisted::twisted::twisted::twisted::hmm:

 

[crashtech]

Well, I didn't mean it like it came out, sorry about that, lol. I didn't mean "anything," but Intel has continuously improved IPC. mfenn comparing the IPC of Haswell to your several generations old i7 is just beside the point, don't you think? Of course the last several i7s are going to have better IPC. Your hangup at this point seems to be Haswell's lack of overclockability, not any deficit in IPC. Some think that is a problem, others don't. If I was asked to spec a system for a hardcore overclocker today, the choice would not be easy. Even 2700K and 3930K still look pretty good in this regard.

 

[Essence_of_War]

But word it however you want; I just don't want a processor that gets so hot it throttles or crashes under a mild overclock, and I'm not going to put my CPU in a clamp and pull the top off.

IVB and Haswell both get hot fast on an overclock. Neither of them is going to thermal throttle on a mild overclock, and de-lidding is hardly necessary unless you intend to chase the 5 GHz dragon.

It IS a little upsetting that IVB/Haswell seems to be less efficient at dissipating heat than SB, and it does seem likely that in addition to the die shrink, the replacement of the IHS binder w/ a paste rather than solder has something to do with it, and worse, that this decision may have been conscious on intel's part. But even so, you're going to be totally fine on a mild overclock with a decent HS+F, and decent case airflow.

If you're willing to wait for IVB-E, that's an option too, but I'd caution you that:
1) You'll spend a lot more money for marginal improvements to gaming over the i7.
2) You'll spend a lot more money on the MoBo, for again, at best marginal improvements
3) Due to the IVB die shrink, and depending on whether or not IVB-E uses solder or paste on the IHS, you may still run into all of the same "hot overclock" issues.

 

[mnewsham]

Okay at 2.8Ghz (the i7-930 stock clocks) your current CPU pulls 130w

Now the i7-4770k at 3.5Ghz (stock 4770k clocks) the CPU will pull 84w

that's 46w less power and 600Mhz more clock speed, not to mention at IDENTICAL clock speeds, the 4770k is ABOUT 30-50% faster than your i7-930.

So lets get this right, we lower power draw by a good bit (46w), increase performance by 600Mhz (at stock) and at the same time we have increased IPC improvements too.

Now, most reliable reviews I have seen have no trouble getting the 4770k up to 4.2Ghz, in fact I have seen plenty of people get theirs up to 4.5Ghz on good air cooling.

Now lets be conservative, let's say you only overclock to 4.2Ghz, you are now at the same clock speed as your current CPU, however you will be pulling less power (probably around 130w for the 4770k vs 240w for the i7-930) AND you must take into account the fact that there have been massive IPC improvements since nehalem, so you are still 40% faster while using less power, integrated graphics, integrated voltage regulation, more SATA and USB 3 lanes, etc.


I don't understand why you think your current i7-930 is really comparable to the 4770k at all...

 

[Essence_of_War]

Also, you can read a bit about the paste v. solder debacle here:

http://www.overclockers.com/ivy-bridge-temperatures

Unless you're buying for the pure fun-sies of getting the highest stable overclock you can, there really isn't a contest. The Haswell i7 smokes the Nehalem at every metric, and is just fine on a modest overclock.

And I've already made the point about the Haswell being faster at equivalent frequencies, but it's still a mobile-oriented processor. I'm simply holding out to see if Intel is going to release a more desktop-oriented processor.

The i7-4770k is not a mobile processor, it's a desktop processor. You might claim that the Haswell architecture is more exciting for mobile users than for desktop users, and many people would agree, but claiming that the 4770K is a mobile chip doesn't really make much sense to me.

 

[mnewsham]

I'm simply holding out to see if Intel is going to release a more desktop-oriented processor.

Let me just burst your bubble now, do NOT expect a massive overclocking SKU, or huge performance jump from intel any time soon, they have been king of the x86 hill for YEARS, they are not losing that title any time soon, it is in their best interest to make headway into the ultramobile market (tablets, smartphones, etc). To do this they are making power improvement changes across the architecture as a whole, not just for individual SKUs this allows them to have much more massive changes quickly as they aren't doing completely different R&D than the larger CPU's, just slightly different. This also gives the added benefit of making desktop processors more power friendly, some would argue that they would rather have more overclocking headroom, but intel disagrees, and frankly, I agree with their choice if it lets them quickly get into fighting ARM.


TL;DR
You wont see nehalem type overclocking CPU's coming out for years in the desktop consumer market.

 

[pcunite]

I personally believe that as Intel moves to smaller and smaller architectures, the headroom for overclocking is going to be lost. The margin for error is changing gentlemen. There is always the chance they could release an advertised lower speed that internally is known to test higher. More likely though, they're doing whatever it takes to keep heat and clock speed in balance.

 

[mfenn]

You're saying there aren't issues with the TIM and IHS? If it were just the smaller die, then how come people are taking the freakin' lid off 'em right away to get decent temps? "Simple physics," *harrumph*...

I never said that there weren't any issues with the TIM between the IHS and the die for extreme overclockers. Please don't put words in my mouth. Note that this is nothing new, Ivy Bridge faced the same issues. Intel has shown that cutting production costs is more important to them than appeasing extreme overclockers (most of whom will delid as a matter of course). Don't expect a new chip to change this trend.

What I did say is that simple physics says that a die with 1/4 the area but 1/2 the heat output will necessarily run hotter. Intel isn't switching back to 45nm any time soon.

But word it however you want; I just don't want a processor that gets so hot it throttles or crashes under a mild overclock, and I'm not going to put my CPU in a clamp and pull the top off.

This is an overreaction, plain and simple. No, you're not going to get to 5 GHz without delidding. Yes, you can do 4.2 GHz on fairly pedestrian cooling. Plenty of review sites have done it. TechReport even did 4.7 on an H80. Actual data from ASUS indicates that 50% of 4770Ks will do 4.5 GHz.

Double the TDP? Who's doubling the TDP?...

Nobody, that's the point. You would have to nearly double the TDP of a Haswell to get it to the same power draw as your 920 at stock.

I looked at those charts. The 4770K is 25-50% faster when the i7-920 is at 2.66GHz and the 4770K is at 3.5GHz..where are you seeing them compare the two CPUs at equivalent clocks?

Huh? I linked you to a chart of a 3.5 GHz Haswell and a 3.33GHz Nehalem. There are benchmarks where the Haswell is twice as fast and benchmarks where it is only 25% faster. I was conservative and said 50%.

So yeah, I agree mnewsham and Essence of War here. Stop spending so much time in the extreme overclockers' echo chamber.

 

[PCJake]

There's been lots of good comments since I posted last - rather than multi-quote everything, I'll just sum up where I'm at:

I'm coming around to the idea of getting a Haswell, even though its depressing to think about how much Intel could have done to better serve enthusiasts. They're a business, though, and I'm sure they're doing the right thing for the business. I am still considering waiting until September, though, just in case they go back to soldering on the IB-E. Most of you will probably tell me not to bother - but I have hope, given that IB-E is obviously going to be more of a desktop/workstation-oriented CPU, and logic should lead Intel to put more effort into making it a more heat-efficient design because of that. I also expect that games will become more optimized for six cores over the next two years, which is roughly how often I upgrade.

If the IB-E doesn't excite me any more than Haswell has, then I'll just get a 4770K in September. Done deal.

 

[mnewsham]

If the IB-E doesn't excite me any more than Haswell has, then I'll just get a 4770K in September. Done deal.

As far as I know the base level IB-E chip is going to be the i7-4820k, quad core, 8 threads and cost $300-400. To get a 6 core (12 thread) IB-E will cost you probably around $600.

Not including cost of a 2011 motherboard, which will set you back another $200-300


So minimally you are looking at $550 for an i7-4820k which is still just a quad core. For 6 cores you will be looking at $700 as the minimum.

 

[PCJake]

As far as I know the base level IB-E chip is going to be the i7-4820k, quad core, 8 threads and cost $300-400. To get a 6 core (12 thread) IB-E will cost you probably around $600.

Not including cost of a 2011 motherboard, which will set you back another $200-300


So minimally you are looking at $550 for an i7-4820k which is still just a quad core. For 6 cores you will be looking at $700 as the minimum.

Yep, if it's not worth it then I'll get a Haswell.

 

[mnewsham]

Yep, if it's not worth it then I'll get a Haswell.

Fair enough, if you have a microcenter near by (not sure if mentioned earlier or not) you can pick up an i7-4770k for $279.99 and if you buy a motherboard with it, they knock $40 off the M/B price. So you could get a full haswell setup for $400
$280(CPU) + ~$120(M/B)

Seems like a pretty decent deal, but if you still want to wait feel free, just trying to make sure you have all the information you need to make an informed decision.