Bulldozer coming 19/09/2011

[Bungral]

AlienALX, providing your rads have cooling capacity, then a standard heatsink and fan isn't going to outperform properly done watercooling.

Only way it would happen is a perfectly designed heatsink that is the size of a case.

Even then, both methods of cooling are limited by ambient temperature.

 

[alienware]

Yeah.

I did see a machine put together by Scan recently. Scorpion or something? and with a triple double height rad it was hot.. On the I7 2600k @ around the 4.7 mark. In all honesty it was no better than the Noctua. I think it's more so on the SB chips which are not really limited by the heat as much as CPUs used to be.

http://www.bit-tech.net/hardware/pcs/2011/03/24/scan-3xs-scorpion-x-pc-review/1

There it is.



The board also offers more overclocking options than we could shake an arachnid at, which is appropriate considering the 3.4GHz Intel Core i7-2600K CPU has been overclocked to 4.8GHz. This is more than a 40 per cent increase in clock speed, and matches the best overclocks we’ve managed with the same CPU.

However, all this performance comes at a price: heat, and lots of it. Running our standard 24 hour stress test of Prime95 and Unigine Heaven, we saw the CPU temperature reach 95°C in our 21°C labs, which is just 3°C shy of the CPU’s Tjmax thermal limit. We wonder how sensible it is to run a CPU so close to its thermal limits. The GPUs of the GTX 590 3GB were much better behaved though, and peaked at just 44°C above room temperature.

 

[Bungral]

Well something is seriously wrong there. I wouldn't really call that a double height rad either... I'd call that normal and the thinner ones, slim rads. No point in doing it if you go with anything thinner than that.

I have a PA120.3 with an XSPC Delta V3 and the DDC Ultra with Res Top along with a 2500K running at 4.7GHz 24/7 with 1.4v and it doesn't get above 58 - 60 degrees under Prime95. I know because I left it for a day to stress test.

I had it running at 4.949GHz with higher V's and it was still only high 60s.

For them to hit 95 degrees is ridiculous.

 

[Ya93sin]

They have all the exhausted air going out via the bottom, so the heat is just going to rise again...What a dodgy design.

 

[Bungral]

Yip... Following on from Ya93sin, there is nowhere in that design where cool air is being blown / sucked into that case either so the rad is only being supplied toasty air.

 

[Bungral]

Oh yeah, just noticed.... The high temps might, just might have something to do with that sodding great GTX590 (that is also overclocked) dumping heat into the loop???

Like to see a NH-D14 cool a SB clocked that high as well as a GTX590?

 

[alienware]

Sadly all the time GPU manus keep releasing their cards so quickly, on different boards with different layouts we will never get really good air coolers for GPUs. It's not like you can fit an NH to a GPU is it? That's a shame.

What I said in my initial mentioning of the subject was -

High end air = better than low end water, by a bleeding long way too.

Which is completely factually correct. I note you are running a WC system and have your I5 at 4.7ghz. I've seen the same I5 clocked to 4.7ghz here -

http://3xs.scan.co.uk/ShowSystem.asp?SystemID=1266

On a cooler master Hyper 212 plus. An air cooler which, compared to the NH pales in comparisson.

And when you couple that with this very good article -

http://www.bit-tech.net/blog/2011/07/13/is-there-still-a-need-for-water-cooling/

Which very correctly states -



Even if the next generation of hardware doesn't notably benefit from water-cooling, there's always a small gap between air cooling and extreme cooling, and there will still be a huge market for it, for the simple reason that it's cool.



Which is what led me to my conclusion. On (at least) the latest SB CPUs water cooling is nowhere near as functional as it was on preceding CPUs. So much so that water cooling is now being looked upon as 'cool' and 'for looks' and 'an art form' rather than a way to gain massive overclocks that just aren't possible on air. Whether the trend will continue? well, I guess that would be up to AMD and Intel on how they make their chips. But with NM (and therefore heat) dropping by the month then on a CPU at least it will eventually become a white elephant.

Now on GPUs however? I stand by my statement. It is important and it is FAR better than any air cooler can do. That won't ever change either, due to the constantly different sized and length boards and GPU and memory placement.

 

[Bungral]

Not overly impressed by that article really. He just says his opinion with no facts or stats or even detailing an explanation why.

Nobody really sees my PC apart from my and my girlfriend and maybe a few mates. My watercooling certainly isn't just for looks. It's functional and is superior to any heatsink/fan combo out there whilst staying silent with my CPU clocked to 4.7GHz (with room for more baring chip limits) at very good temperatures. Quite literally, with my case 1 foot away and the side panel off, I can only hear my PSU fan.

I agree with what you say about high end air being better than low end water definitely but it definitely isn't better, and won't ever be better (in it's current form) than high end water done properly.

This isn't taking into account the ease of a heatsink/fan compared watercooling of course. If you happen to be a reviewer / tester or someone who changes components every couple of weeks, then watercooling is a pain in the arse.

 

[alienware]

Not overly impressed by that article really. He just says his opinion with no facts or stats or even detailing an explanation why.

Most probably because he spends his days, every day, reviewing and setting up hardware. It isn't just him either. I starkly remember Tom making such a comment very recently. Something along the lines of "I am beginning to wonder if water cooling is even worth it any more".

Nobody really sees my PC apart from my and my girlfriend and maybe a few mates. My watercooling certainly isn't just for looks. It's functional and is superior to any heatsink/fan combo out there whilst staying silent with my CPU clocked to 4.7GHz (with room for more baring chip limits) at very good temperatures.

The NH is silent, and I would bet it could keep a I5 @ 4.7ghz cool enough too. On 1366 water was king. On 1156 water was king, as it was on 775. Chip technology is moving on, and heat is not going to be the object for much longer. When CPUs reach 20nm or less the brick walls will not be caused by heat. They will be caused by the CPU simply blowing up due to the voltage needed to try and get them stable.

I agree with what you say about high end air being better than low end water definitely but it definitely isn't better, and won't ever be better (in it's current form) than high end water done properly.

This isn't taking into account the ease of a heatsink/fan compared watercooling of course. If you happen to be a reviewer / tester or someone who changes components every couple of weeks, then watercooling is a pain in the arse.

It won't be better. But, it will be sufficiently good enough to perform the same function. Air coolers were not really focused on a couple of years ago like they are now. Again, back then air coolers were a bit pants. We had nothing like the NH on 775. I'd actually be interested to see how they do on 775. Maybe some day I'll shove mine on my core 2 duo

 

[Ya93sin]

I think the CPUs are running cooler and cooler, so the need for watercooling is less. I suppose atm watercooling is still better performance wise than air cooling, but more expensive.

For GPUs though, watercooling is definitely the thing to do to maximise performance. Even the best air coolers will struggle to get near a watercooled GPU setup.

 

[alienware]

Oh absolutely. Mind you having said that GPU nm is dropping like crazy too. I read some specs on the 6 series Nvids last night and it was pretty sick !

It may not be long until we have nice cool GPUs in our mitts too !

 

[Ya93sin]

I suppose eventually we will be able to put a high end gaming rig in a HTPC case With decent temps too!

I wonder whether passive cooling might make a return too?

 

[Bungral]

While they are lowering the NM, in turn they are increasing the number of transistors per chip. Have you noticed that even with the lowering of NM the TDP's have remained similar on the performance parts?

They can still kick out the same heat.

In fact, my 2500K kicks out just as much, if not more heat than my i5 750 / i5 760 and i7 860.

All of the above chips have the same TDP of 95w even though the 750, 760 and 860 are 45nm and the 2500k is 32nm. Reason being the Lynnfields have 774m transistors compared to the 995m in the Sandy Bridge.

So they aren't necessarily going to get that much cooler in the high end, but they will get more powerful. I'm not saying they wont get cooler at all as I expect they will but they gobble up the gains of a smaller process by increasing the performance.

 

[dugdiamond]

this thread is getting off-topic from the OP

 

[alienware]

Not really. We're talking about future chip technology, heat, and cooling.

Very interesting IMO.

Ed.

Actually take a look around for the info on Intel's new 3d tri-gate transistors.

50% power saving over planar transistors and considerably reduced power leakage. They reakon that they offer 37% more speed with less power leakage on a smaller die. Apparently they only cost 2% more than current transistors, though I would bet my bum that wouldn't be the price increase.

Oddly enough their debut will be on Ivy Bridge, 22nm.

 

[Ya93sin]

While they are lowering the NM, in turn they are increasing the number of transistors per chip. Have you noticed that even with the lowering of NM the TDP's have remained similar on the performance parts?

They can still kick out the same heat.

In fact, my 2500K kicks out just as much, if not more heat than my i5 750 / i5 760 and i7 860.

All of the above chips have the same TDP of 95w even though the 750, 760 and 860 are 45nm and the 2500k is 32nm. Reason being the Lynnfields have 774m transistors compared to the 995m in the Sandy Bridge.

So they aren't necessarily going to get that much cooler in the high end, but they will get more powerful. I'm not saying they wont get cooler at all as I expect they will but they gobble up the gains of a smaller process by increasing the performance.

Hmm, I wonder if they will get to a point (imagine 1nm or so) where they decide they need to start making chips smaller. We would also see smaller motherboards too at that point

Speaking of which, I wonder how hot these Bulldozer cores will run?

 

[alienware]

AMD usually run as cool as cucumbers. Heat wasn't really the problem with those in honesty, it was just more than AMD had squeezed more out of them for sale. Intel leave the headroom in, but you pay for it.

I just can't see a 22nm CPU that uses half the power getting very hot at all really. And as I mentioned earlier, I think that heat will no longer be a barrier, water cooling or not. They will simply brick wall or blow up.

 

[KiiYzOo]

I think they'll move onto pm rather than reduce sizes

also I wonder how the new noctua air coolers will fare with new cpus, as they can only get better than the nh-d14

also I don't think that watercooling technology can go any further, as there isn't really much you can improve over water cillers etc, which are too expensive to even consider for mainstream gamers

 

[Ya93sin]

If Noctuas become even more overkill then they may run these chips nearly as cool as watercooling, but I think the D14 successors may have found a home on the LGA 2011 socket. Let's see.

 

[alienware]

Indeed. I know it was mentioned that the NH cools so well due to its massive size. I won't deny or debate that, I totally agree.

However, most coolers look the same. There is bound to be other ways of getting them down in size with amazing performance. GPUs run hotter than CPUs. Look what Nvidia's vapour chamber did for the 5 series. And that was some tiny little thing. Imagine that was the size of say, the dark rock pro, cooling a CPU that is way cooler than any GPU.

But there are always walls. Obviously as voltages decrease then heat will decrease with it.. I mean from what I can gather the Ivy will run on 3/4 of a volt

That's friggin insane !

Heat is, and always was the enemy.. Right up until something craps out through just not being able to go further. I've seen a GPU for example go from 700 odd mhz to 1ghz, then refuse to go higher. And it's not down to heat at all, simply that the technology won't let it go any further.