Really dire temps for new loop - advice...

[Michaeljcox24]

The long and short of it, is that nobody knew what was causing the high temps on the cards. I even tried the hottest one on it's own, and it still went up to 65 odd degrees.

People (including TTL) told me that blocks can't be faulty, they're machined to precision sizes, so it must be the cards. However, the cards were fine on air, so I don't know. Perhaps they can't swim very well or something.

 

[Scoob]

Hi Michael,

Sorry, I missed your last-but-one post where you say you'd gotten the 690! Seems your issues have been most odd! It's possible something was partially blocked somewhere, I assume you flushed everything out properly before using? I'd have to agree with Tom re: blocks, unless they'd somehow been damaged - in which case they'd likely have leaked - then there's not much to go wrong with them.

Do you think you might have made a mistake fitting the blocks originally? I've heard of people forgetting (or simply not realising) to remove the plastic layer (both sides!) from the TIM pads - i.e. they removed one layer, and stuck the TIM to the card, but left the other layer intact...not good & would mean v. high temps. Cutting to shape & attaching the pads is fiddly & quite frustrating when doing two cards, so it's understandable I guess. Also, with EK blocks at least, they only ship just enough TIM Pad material with minimal spare in case of mistakes.

It's one of those things where a second pair of eyes actually with you and your kit would help. Forums and the community are great, but hands on is unmatched.

On an aside...I admit, I'm tempted by a 690 myself at the mo, £720 for the card (using a promo code) and around £100 for the block. How do you find your 690 vs. a pair of 580's? If that's not drifting too off-topic.

Scoob.

 

[Michaeljcox24]

I did everything according to the book with the EK blocks, didn't leave a single stone un-turned. The ONLY thing I would say I could have possibly have done wrong, was the stand offs. They were incredibly fiddly, to the point of pulling your hair out. They all tightened up properly, but so utterly fiddly to fit that that's the only part of the process I envisage I could have done wrong.

As for the 690, well, it's a two sided coin. It's listed and advertised as a 4GB card, but it's actually only 2GB usable. I didn't know until I started going on Skyrim and had huge problems with lag. I'm debating whether to RMA it and get a pair of 7970's with more VRAM.

The otherside of the coin, fares better. She overclocks like a beaut, I got 1230MHz out of it today. Temps didn't go above 40 (which I'm still struggling to believe are true after a lifetime on air) and thus, performance to noise ratio is monstrous.

So yeah, apart from the 'bit of a con listing it as a 4GB card' marketing plot, it's fine.

 

[Michaeljcox24]

P.S Your rig is almost identical to mine, bar the card. Good call.

 

[Scoob]

Odd you struggled with the stand-offs - I don't recall having any issues with them other than being slightly fiddly and the included mini-spanner allowed me to tighten them nicely. I found doing the pads far harder. Incidentally, my first GTX570 had the same PCB as a 580, hence my old block is a direct fit for a 580...I mean, it should be the same as your block.

Re: 690 VRam...that's the norm, they list total for both GPU's rather than the effective amount. Still, 2gb per-GPU should be fine*, or are you a 2560x1440/1600 gamer? My 570's only have 1.25gb each of course, yet Skyrim @ 1920x1200 with LOTS of visual mods (way beyond the "Ultra" preset") plays like a dream - largely 60fps everywhere with the exception of a few external locations.

Note: there is an odd engine bug in Skyrim which manifests its self when you turn on the spot - you get momentary judder. I've observed that vRam was fine as is GPU load, yet I'd see a spike in CPU during those moment - effectively maxing a core for a fraction of a second. It ONLY happens when I turn, if run/walk normally anywhere things are really smooth. It could be some texture swapping going on I guess, but it only happens in one or two places - notably outside the Whiterun Blacksmiths - I have quite a heavily modded Whiterun with loads of extra trees and clutter.

If you're really unhappy with the 690's performance, then maybe a pair of the 7870's with the new core would be worth considering - assuming blocks are available, I'm not sure if the PCB has changed...I'd be surprised if it hadn't. Still, I'd expect better from a 690! Maybe hang fire for now and see if the 700 or 8000 series from NV and AMD respectively deliver.

* While more vRam is certain the future and cannot hurt, there appear to be a lot of people online saying things like "Game X NEEDZ at least 2.5gb of vRam as that's what I see it using nowz!" Erm, no. Like windows, many games will cache things in vRam if you have the capacity. This is a good thing in that if those resources are needed again the GPU has them on hand. However, it does not mean that the GPU HAS to have all this in vRam at all times to render the current scene. My 570's have "only" 1.25gb of vRam, very low compared to the current stuff from AMD, yet it plays this VERY well at the settings I want to use. I generally max everything, but only use 4xAA or FXAA if available - such as in Skyrim - and things play great.

Modded Crysis 2 (high-res textures, DX11) is a fluid and costant 60fps on my rig @ 1920x1200 with everything maxed. The most vRam I've ever seen used is 1260mb out of a total 1280mb, so my mere 1.25gb copes just fine in a title where people have screamed that it MUST HAVE at least 2gb.

I will add that vRAM does help with high levels of AA but, to me at least, regular 4xAA (with 16xAniso) does the business.

Scoob.

How OT did I go?

 

[Michaeljcox24]

I know what you mean about the turning on the spot thing in Skyrim. That's exactly what I am referring to when I talk about the lag. With the 2K textures in installed from the Nexus, I used to hit 2.5GB ram usage in some areas, sometimes 2.7GB. It's not as if the game is unplayable or anything, it's just when it spikes, and judders, that it can become a noticeable distraction from the norm. Otherwise, it flows perfectly, and I can't fault it. I was used to it not spiking like that, and being the main game i play, it's just annoyed me a bit.

I can't go over to the red team because I'm running a 3D Vision monitor, and play a few games in 3D (AC Rev, Batman AC etc) so would loose the functionality of that. The only thing i could go over to would be a 670 with more VRAM, but that would mean draining the loop, RMA'ing, pissing about, and after the huge amount of fault finding I did the last fortnight, I just can't be arsed. At the end of the day, it's only ONE game that goes over the 2GB, all the rest are between 1 and 1.5. Trying to justify the whole RMA process to myself, just for one game, is hard.

Oh yeah, and the thermal pads on the card. When you put yours on, were there two different thickness pads you had to put on, or were they all the same thickness? The instructions that came with the blocks stated some of the VRMS needed a thicker pad, which seemed weird, because they were the same thickness as the others.

 

[Scoob]

Yeah, a friend and myself have done some testing on that - he has a pair of 1.5gb 480's on water @ 850mhz - we both experience the same thing with the same build.

Ah, you play in 3d! Yes, my friend with the 480's also plays in 3d @ 1920x1080 for Skyrim and everything else he can these days. We've noticed that 3d only has about a 20% or so GPU/CPU overhead, but it can push you out of the 60fps zone easily as of course we need 120fps for a perceived 60fps now...that's too much for a pair of 480's. Plus, really, some more CPU grunt over his 2600k @ 4.8 would be useful too, Skyrim does use more CPU than you'd expect.

I do get the whole hassle factor thing. My loop was working perfectly, however the (external) rad had a manufacturing fault so had to be returned - major hassle. Then, after fitting the replacement rad, one of my EK Compression fittings failed (I'll never buy another EK compression fitting) killing my upper GTX 570. Major hassle once again. While I, like yourself no doubt, do enjoy building a rig, I do want to get some play time too! One of the reasons I did an external assembly is that I can easily transfer it to another build...it takes 2 minutes to unplumb it, if that. Draining is easier too, as I'm just draining the coolant in the case, so what's contained in the pipes and blocks - takes a couple of minutes.

Thermal pads - I just dug out the boxes to check this...

My old GTX 570 rev 1.0 block - so same PCB as a 580 - had THREE thicknesses of pads, a 0.5mm, a 1.5mm and a 2mm. The last one was for one component only.

My newer GTX 570 rev 2 (the SE 1.1 block) for the new, shorter "HD" GTX 570 had TWO thicknesses of thermal pad, a 0.5mm and a 1mm.

I still have the remaining off-cuts of pad, so I can see the difference between them quite readily. I wonder if your blocks only came with the ONE thickness? That would be really bad as, if too thin, you'd not get the contact - hot hot hot, and if too thick, well, the extra clearance that would potentially give you on some components, would mean zero contact on others. That could be the source of all your troubles!

Scoob.

 

[Michaeljcox24]

I can only think its the pads. There looked hardly any difference in the thickness, and it's the only thing that would explain the high temps - bad contact on the gpu itself.

 

[Scoob]

Yeah. 0.5mm isn't a lot, but if they've given you the correct pads then you should be able to tell them apart from each other. Once you're in the modd for having another look, I'd double check that if I were you - for peace of mind if nothing else.

Scoob.