![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Swiftech Apogee Review.
- Thread starter Bio-Hazard
- Start date
Top Nurse said:I do get it, but I think you guys don't. Obviously there are some here who care about the finite numbers that one can get in a laboratory and there are others who care about what you get in real life. Seems to me I am in the latter category so I am not too concerned with what you guys think unless you can come up with a testing methodology that will have data that will easily transfer into real life situations. When you get there please announce it.
TN, no offense, but I am starting to think that you argue just to argue. If you can not see the real world benefits of using bare die over ihs then...
I don't think anyone is argueing the benefits of using bare die but what do you tell all the Intel users who can't remove the IHS? The fact is there are two types of purchasers, ones that buy based on non real world results and ones that buy based on real world results.plywood99 said:
Erasmus354
[H]F Junkie
- Joined
- Mar 12, 2004
- Messages
- 9,450
R1ckCa1n said:
What good are looks in a waterblock? You dont use your computer by staring at it's innards...
Everything is relative. Some might want to spend more on looks that dont really get them anything. Some may want to spend more on cooling performance, whether it nets them something or not. If you look at the test the guy did on XS it shows that the block which performed best on the die sim also performed best in overclocking (shocker!).
As was said earlier, unless he were to test each block and each overclock on a large number of different CPU's the test means nothing. Each CPU is different, and each block will perform slightly differently on each CPU. So how about we stop bickering about testing and just stick to the best thing we have right now.
When you buy a car and they tell you it has X horsepower, how do you think they measure that? By going out on the Beltway and driving around a bit and saying "this feels like 250HP to me." NO! They put it in a lab and ACCURATELY measure the performance. Do you drive the car in a lab? No, but that is where the performance is measure anyways. The real world contains too many irregularities to make it a valid testing area. Sorry.
DFI Daishi
2[H]4U
- Joined
- Feb 14, 2005
- Messages
- 2,839
sorry, i might be mixing up my terms here. calibration and accuracy are two different things. on refers to internal consistancy and the other referst to argreement with accepted values. i can't remember which is which. a combination of those two factors gives you variance.Top Nurse said:
if i remember correctly, the accepted variance in a CPU diode is 5C, as that is good enough for thermal protection purposes, and that is all that the diode is really necessary for as far as the manufacturer is concerned.
If you are referring to the fact my Cuplex XT performs just as well as my Storm, you are right the XT looks alot better.Erasmus354 said:
Exactly what blocks have you tried in your system for comparison? Please enlighten us to your testings.
Erasmus354
[H]F Junkie
- Joined
- Mar 12, 2004
- Messages
- 9,450
R1ckCa1n said:
Read the full post.....taking quotes out of context is convenient. If you were to read the entire post and take everything in its full context it would be relatively simple to see where I was going with the comment
It is good for you to mix old threads and include them in current onesErasmus354 said:Read the full post.....taking quotes out of context is convenient. If you were to read the entire post and take everything in its full context it would be relatively simple to see where I was going with the comment
Again, try to stay on topic to the thread and don't let your fustration mix up what thread you are posting in. I addressed your concerns above.Flawed yes but the link provided by Cathar speaks volumes about how testing might go in the future, I hope. It is a matter of time before AMD attachs the IHS to their cores too and then we will be left with a testing plateform that is completely based on theory.plywood99 said:
What I am saying is if I was in the market for a new block the results in that link would hold more water IMHO than anything a die simulator based test says.
plywood99 said:
In my dream-world, AMD and Intel will eventually admit the total thermal folly that IHS's are, and go back to shipping bare-die CPU's and quality heatsinks and properly designed mounting mechanisms that don't apply angled pressure on CPU's, just like they do with mobile CPU's. There are no IHS's on mobile CPU's. Why? Because IHS's suck for thermal performance. The CPU developers know this. They just use IHS's on CPU's because they can't trust Joe Sixpack not to crack a core messing up a bodgy install of a cheap arse low-quality heatsink, and then attempt to claim that it came shipped to him like that. Mobile CPU's they can trust that at least someone with the proper equipment and half a brain is installing a quality controlled heatsink.
Erasmus354
[H]F Junkie
- Joined
- Mar 12, 2004
- Messages
- 9,450
R1ckCa1n said:It is good for you to mix old threads and include them in current ones
Eh? My post was DIRECTLY related to the ongoing discussion in this thread of testing of waterblocks. Specifically testing a waterblock based upon overclocking results on a CPU. I dont know what thread you think I am referring to, or how it was off topic. I think you need to put down the pipe my friend.
R1ckCa1n said:
Hmmm, not quite sure I agree with you there.
It's like this. IHS's can and do vary from CPU to CPU. Unless you do a sample size of 10 installs on perhaps 50 different CPU's, there's no real way to generate a clear statistically probable picture of what will happen on a "real world" setup, just because there are too many variations from CPU to CPU.
The "theoretical" value speaks volumes. It is measuring and telling you how well the block cools, independently of the varying nature of IHS's. Sure, stick the block on a few different CPU's and it may not appear to perform as the theory states. To assume that the lab testing is invalid as a result is a false presumption. The issue here is the uncontrolled and unforeseen variations on a per-CPU basis of the IHS. The theory is telling you what should be happening. The reasons why you wouldn't see that in the "real world" is because of the IHS. However, if you did actually do the 10 installs on 50 CPU's to obtain a statistically confident ranking, what you would find is that the theoretical value would then be providing pretty much the exact same information.
This is the problem. "Real World" testing would involve a ton of work to generate a statistically confident estimation of the waterblock performance because of all the in-built variations in IHS's, whereas the "theoretical" testing would be giving you the exact same information, but we were able to derive it without having to deal with and attempt to quantify IHS variations.
Now, if you instead told me that what you really want to see is a measure of just how much IHS's and CPU thermal probes can vary "in the real world", then we could probably justify spending the effort on that. If the result of that research then came out to say that in the "real world" that things vary by +/- 5C and +/- 200MHz overclock across a range of CPU's, what would that tell you about someone who provided a statistical sample of 1 when they provided "real world" test results?
Erasmus354
[H]F Junkie
- Joined
- Mar 12, 2004
- Messages
- 9,450
Cathar said:
Exactly what I have tried to say earlier, except like always you put it so much more eloquently (not to mention that people listen to you instead of assuming what my point is).
Well let's see, you KNOW I can run at 5Ghz with my P4 on my water setup. You live in So Cal as well no? We should get together and put my PSU, mobo, ram and CPU in YOUR system and see how far we can go on your AC stuff. How does that sound? Then, when you are 400 Mhz behind, you call sell your AC stuff and get a real big setup like mine, since as you've stated so many times, only Max OC matters to youR1ckCa1n said:
This here is a challenge that I doubt you will accept.Part of the problem with overclocking for testing is ensuring that the overclock is purely CPU & temperature bound. Overclocking ability is very often limited by the motherboard's power regulation capabilities.
I have had CPU's that will overclock well and respond like a dream to better cooling in one type of motherboard. Move the exact same CPU to a different board with mediocre power regulation and suddenly everything "flat-lines", and typically at a much lower overclock. The issue being that the board can't supply the required juice to the CPU to sustain the CPU's power draw at anything above a particular voltage/speed combination.
Have seen exactly this problem, motherboard power regulation, more times than I care to count. Sometimes not even buying a board from a reputedly good motherboard maker will offer any guarantees. Have seen boards from "good" mobo makers with borderline power control that will still "flat line" in terms of CPU overclock regardless of cooling, but then switch to a different board from the same maker, and waterblock performance overclock effects start showing through properly again.
Just one more knurl to consider in the whole "real world + overclock" equation. Need to be absolutely confident that the person's equipment is not "flat lining" the cooling's ability to provide for better overclocks. Unless the tester has also gone through a few different boards and power supplies, and even memory for that matter, to find a good matching set, then quite often the ability to overclock is not actually limited by the CPU + cooling, but rather by something else entirely.
I have had CPU's that will overclock well and respond like a dream to better cooling in one type of motherboard. Move the exact same CPU to a different board with mediocre power regulation and suddenly everything "flat-lines", and typically at a much lower overclock. The issue being that the board can't supply the required juice to the CPU to sustain the CPU's power draw at anything above a particular voltage/speed combination.
Have seen exactly this problem, motherboard power regulation, more times than I care to count. Sometimes not even buying a board from a reputedly good motherboard maker will offer any guarantees. Have seen boards from "good" mobo makers with borderline power control that will still "flat line" in terms of CPU overclock regardless of cooling, but then switch to a different board from the same maker, and waterblock performance overclock effects start showing through properly again.
Just one more knurl to consider in the whole "real world + overclock" equation. Need to be absolutely confident that the person's equipment is not "flat lining" the cooling's ability to provide for better overclocks. Unless the tester has also gone through a few different boards and power supplies, and even memory for that matter, to find a good matching set, then quite often the ability to overclock is not actually limited by the CPU + cooling, but rather by something else entirely.
I would like to take you up on that offer! You also have to take your case of fans and make your rig run completely silent as mine doesnikhsub1 said:Well let's see, you KNOW I can run at 5Ghz with my P4 on my water setup. You live in So Cal as well no? We should get together and put my PSU, mobo, ram and CPU in YOUR system and see how far we can go on your AC stuff. How does that sound? Then, when you are 400 Mhz behind, you call sell your AC stuff and get a real big setup like mine, since as you've stated so many times, only Max OC matters to you
DFI Daishi
2[H]4U
- Joined
- Feb 14, 2005
- Messages
- 2,839
that was not the challange offered.R1ckCa1n said:I would like to take you up on that offer! You also have to take your case of fans and make your rig run completely silent as mine does
he said that he wanted to see how his gear overclocks with your cooling solution, not if his cooling solution can watercool each and every little component in his case.
just by the by: you can have everything well cooled, in a high flow system with only the CPU and GPU watercooled, without making the system loud. it takes some planning and doesn't look pretty, but you can do it.
I have stated over and over on this forum that silence is one of my main goals. If not, I would be running a pool pump with 3/4" tubing with 80 fans to cool all my stuff. I perfer to get the best of both worlds. Maybe that's where AC comes into play?nikhsub1 said:
BTW: Huntington Beach - You?
DFI Daishi
2[H]4U
- Joined
- Feb 14, 2005
- Messages
- 2,839
were it not for the heat load of my pelts, i would be running 2x120mm panaflo L1A fans @ 5V, 2x panaflo 80mm L1A @7V and one low profile fan stripped out of my old zalmna OP1. overall noise level: quieter than my raptors, so there is no need to make it any less noisy than that.R1ckCa1n said:
number of fans doesn't necessarily mean much if the overall noise level is still low.
DFI Daishi
2[H]4U
- Joined
- Feb 14, 2005
- Messages
- 2,839
automotive forum, perhaps? :|Budwise said:dang, lots of California boys around here. Where's all the Texans at?
plywood99 said:
I don't think they will bother because they are only concerned with whether their CPU's will clock at the stated speed and if they will last as long as the warranty they give. If you want more OC juice then you should just buy a more expensive faster processor. The IHS also gives them some other advantages because you don't have ready access to the circuitry around the die to play games with. It would be easily evident to them if you pop the IHS so if you want to do that you are free to do so. I suppose we should thank AMD for at least making that possible.
Seems to me that they have things exactly where they want them.
Top Nurse said:
Yes, but as heat load increases the tim joint / ihs will have to be improved. Think about quad core and even 8 cores cpu's on Intel's roadmap. Eventually something will have to give. Either a better tim joint, or drop the ihs, and I don't see the latter happening.
On A64's, there is very little circuitry to fool with, mostly just capacitor packs for smoothing out voltage. Multipliers and such are programmed through the pins...
This could work as I am redoing my loop yet again this Friday, if Sharka gets their order.nikhsub1 said:Well my office is in Beverly Hills, Palm and Wilshire... bring your sorry gear up here so I can abuse it
I would love to see the overclocking difference with your CPU loop. Maybe we could remove my IHS? lol