[Conumdrum]

Link:
http://skinneelabs.com/gtx480-fc.html

Thanks Skinnee, thats a TON of work.

I have 470s, so bit different. Middle of the pack in all areas is plenty good for me, and the middle is pretty high up the curve. Amazing that the most expesive block is the worst performer.

[Owenator]

Originally Posted by Conumdrum

Link:
http://skinneelabs.com/gtx480-fc.html

Thanks Skinnee, thats a TON of work.

I have 470s, so bit different. Middle of the pack in all areas is plenty good for me, and the middle is pretty high up the curve. Amazing that the most expesive block is the worst performer.

Interesting read. I'm glad I got the EVGA blocks fo rmy 470's. They were the best performing!

I don't a gree with his weighted chart. Flow restriction had a positive impact on cooling performance (probably desinged that way on purpose) and the VRM's don't need to be water cooled. Those two items "weighted" made the best performing block look second to last. That makes zero sense to me?

When you design something you pick the paramreters you want to optimize - in this case GPU cooling. To penalize your results by adding in factors that don't apply to you goal is just bad engineering IMHO.

[Conumdrum]

I disagree somewhat. These are 480's, not 470's. VRM's are smoking with volt mods and overclocks. So VRM cooling matters to some, not all, but for the big clockers it's a very important part of the test.

The GPU cooling except for the BP one, is within 3.2C. GPU chips don't care about that small diff. Even if all the GPUs were at 70C average, we really don't need to cool our GPUs except for the crazy noise they make.

But, looking at the VRM temps, wow! 33.8C difference, thats huge!

Depending on rig setup, the flowrate chart shows that picking the right block can make the difference between buying one pump or putting two in series.

I think he covers all 3 issues that matter. One part might be more important to someone. The last chart means nothing to me, a quick glance chart. I'd rather dig in and read the data.

[Owenator]

Originally Posted by Conumdrum

I disagree somewhat. These are 480's, not 470's. VRM's are smoking with volt mods and overclocks. So VRM cooling matters to some, not all, but for the big clockers it's a very important part of the test.

The GPU cooling except for the BP one, is within 3.2C. GPU chips don't care about that small diff. Even if all the GPUs were at 70C average, we really don't need to cool our GPUs except for the crazy noise they make.

But, looking at the VRM temps, wow! 33.8C difference, thats huge!

Depending on rig setup, the flowrate chart shows that picking the right block can make the difference between buying one pump or putting two in series.

I think he covers all 3 issues that matter. One part might be more important to someone. The last chart means nothing to me, a quick glance chart. I'd rather dig in and read the data.

You make good points.

I am not a GPU overclocker so the VRM cooling is not something I focus on. I did add a 80mm fan blowing across the heatsink secion of my two cards with the EVGA coolers.

I agree that the flow restriciton can be a factor in pump sizing although most of the pumps used in water cooling today could deliver good flow even with two of the EVGA blocks in the loop. I have two blocks in series in my system and they get good flow from a single MCP350.

Bottom line though, you are right about the charts versus the data. Charts are meaningless unless you understand the data behind them.

Sorry, to come across as panning those tests. I just didn't like the "weighting".

EDIT: I just remembered that I have my two EVGA blocks flowing in parallel. That makes my pressure drop much less than two in series.

[Grosjambon]

LOoking Good! Won't Be usefull for now, since these gpu cost a lot,
Keep up the great work.

[EarthDog]

Cool review, thanks Conun, and skinee...

470/480... potato/potatoe. If you are cooling your GPU with water, you had darn well better cool those VRM's with it too.

[Owenator]

Originally Posted by EarthDog

...470/480... potato/potatoe. If you are cooling your GPU with water, you had darn well better cool those VRM's with it too.

The VRMs are cooled on all of the blocks. Just not actively water cooled on the EVGA one. It has what looks like a heatpipe to cool the VRM's which is in-turn connected to the water block.

I would like to know what they normally operate at and what "too hot" is for them. So far everthing I've found today says that VRM temp has to be monitored with a separate meter and is not built in on the 400 series cards.

[EarthDog]

Check GPUz.. wont that tell you VRM temps? I swear they are on there...

Conundrum hit it right on the head though...they need cooled by water or there is no point in getting the water as you will likely be limited in your overclocking b/c of the VRM's not directly getting cooled down by the block/water, especially when you volt mod.

[Owenator]

Originally Posted by EarthDog

Check GPUz.. wont that tell you VRM temps? I swear they are on there...

Conundrum hit it right on the head though...they need cooled by water or there is no point in getting the water as you will likely be limited in your overclocking b/c of the VRM's not directly getting cooled down by the block/water, especially when you volt mod.

It is not an all or nothing situation. The VRM's are cooled by all of the blocks just apparently not as good by each.

No GPUz won't do it the sensors aren't there. Skinnee added the temperature monitoring for his tests. But looking at where he placed them I am not sure what he was monitoring. I always thought that the little squares in the "four packs" were the voltage regualting IC's not the larger six ones where he put the sensors between. I used one GPU cooler where you attached heatsinks to the foutr packs but never those larger six things. It does look like some blocks did touch the sensors but he didn't use any TIM so I can't see if there was any contact.

I'd like to see this addressed because the sensors appear to just be touching some waterblocks which is of course going to be cooler but that's not measuring the VRM temp at all.

Again, I don't overvolt/overclock video cards so I'm no expert there. Just looking at the tests and the results and wondering. It's the scientist in me. When I see one data point way out there I want to know "why"?

[m0r7if3r]

Originally Posted by EarthDog

Check GPUz.. wont that tell you VRM temps? I swear they are on there...

not on the 480...some cards do have it though. EDIT: appearantly not...I'm crazy...

Originally Posted by Owenator

I'd like to see this addressed because the sensors appear to just be touching some waterblocks which is of course going to be cooler but that's not measuring the VRM temp at all.

Again, I don't overvolt/overclock video cards so I'm no expert there. Just looking at the tests and the results and wondering. It's the scientist in me. When I see one data point way out there I want to know "why"?

Skinnee hits on this a little in there about how he tried to avoid contact...I don't remember what page. I do have to agree about the breakdown chart at the end, I think a performance/dollar chart would be much more handy...then you have to design performance though...now I'm wondering how much difference varying amounts of contact would make.

[thorilan]

skinnee might want to avoid a performance dollar chart because as a review of the cards uses data . prices change and for him to get the parts he needs from the companies he has to be un biased.
once he starts showing dollar signs it wont matter if he is biased or not , the companies will feel like they are being attacked and may not give him test material free again unless they feel they have a real flagship product.
i have seen this a few times. the companies want you to have to interpret the data and not consider price before deciding what you think of the performance.

[EarthDog]

Must have been some other card... or I am crazy. both likely.

Why is a great question to ask, especially if the results are anomolous in this case.

[Conumdrum]

Originally Posted by Owenator

The VRMs are cooled on all of the blocks. Just not actively water cooled on the EVGA one. It has what looks like a heatpipe to cool the VRM's which is in-turn connected to the water block.

I would like to know what they normally operate at and what "too hot" is for them. So far everthing I've found today says that VRM temp has to be monitored with a separate meter and is not built in on the 400 series cards.

No card has a VRM sensor. In Skinnees review and much discussion 90C on a VRM is okay.

Skinnee actually put HQ temp probes right next to the VRMs on the card. It's as good as he could do. He was able to compare VRM cooling somewhat.

If the VRM overtemps it goes into a phenomena known as thermal runaway and becomes a short and burns up itself and other fun things usually.

[Conumdrum]

Skinne went into discussion in the 1400+ long thread why he's trying the comparison chart.

[hokiealumnus]

I handled this privately not once, but twice now, so I'm going to put it out for everyone to see.

If you have a problem with skinnee's past issues (if you don't know of which I speak, don't stress it), just plain stay out of the thread. Posting one thing that has absolutely nothing to do with the other will lead to strong action against you. This goes not only for this thread, but for any thread involving said person.

If it happens again in this particular thread after this warning, regardless of person, you will be given a pleasant, sunny vacation from the forums to think about why you shouldn't do it again.

-hokie

[Owenator]

Originally Posted by Conumdrum

No card has a VRM sensor. In Skinnees review and much discussion 90C on a VRM is okay.

Skinnee actually put HQ temp probes right next to the VRMs on the card. It's as good as he could do. He was able to compare VRM cooling somewhat.

If the VRM overtemps it goes into a phenomena known as thermal runaway and becomes a short and burns up itself and other fun things usually.

I'll do some more searching on VRM temps then, thanks.

I never did more than a few trial gpu and ram overclocks on video cards. I water cooled my cards more to get the temps lower and reduce the noise. I doubt I'll be overvolting any time soon but good info if I do.

My main concern with the VRM temps, as published, was how well the method really shows their temp versus external factors. I agree that if applied the same to all cards it is "fair" but I am not sure how accurate.

[m0r7if3r]

Originally Posted by Owenator

I never did more than a few trial gpu and ram overclocks on video cards. I water cooled my cards more to get the temps lower and reduce the noise. I doubt I'll be overvolting any time soon but good info if I do.

cmon, you know you want to, it's free frames