ATI 4850 benchmark

1. 4870 is using a bigger, dual slot cooler.
2. Fan speed is controlled by 4870 and not constant. Fan might be running at very low speed at idle.
3. "temperatures never rose above 82C since the fan would immediately spin up."
4. The readings are not comparable in the first place, we don't even know the delta T between ambient and GPU.
5. Different cards, different coolers, there are too many variables to compare anything.
6. In other words, it's pointless.

LOL

 

NP, I am eager to learn anyway, since this is also part of my hobby. (Even though my work has no relevant to it >.<)

Very true. But I really wonder about point number 5 you stated. "Different cards, different coolers"

I've been wondering ever since my friend ask me this question, does each manufacturer of GC (Be it ATi or nVidia card) stock cooler have any difference in cooling the card ? Like some Manufacturer uses "Silent Fan", "Big Fan<ASUS>" or whatever what they call, when comparing with each other, how significant can it be ? I have no experience in this side of field so I can't really say much to him.


Btw, I think someone should change the Thread Title to HD4870/HD4850 Benchmarking
What do you guys think ?

 

Exactly my point where is the information on fan speeds .. auto .. manual 50% .. 100%??? When I run benchmarks I always have CPUID monitor running as well as EVGA precision/Rivatuner so all settings and temps are recorded. Otherwise whats the point?

If this 4870 running stock speeds can't stay under 82C with a dual slot cooler under any fan conditions I don't want it in my machine.

Also interesting your points on this Max considering you were quick to post unsupported benchmarks from other forums while trying to make a point. Where are the above parameters in those links?

Anyway I do agree with your points and Delta T will be used in any future reviews I conduct involving temps perhaps even a revision of the Hyper212 review

 

Quoted for teh lulz.

and yet you were questioning why 4870 has lower load temps than 4850?

LULZ

Me posting unsupported benchmarks? Come onnnnn. I can say every benchmarks are "unsupported", even YOUR benchmarks. I don't know why you are being offensive here, but that post was directed at that table ALONE.

Let me make myself clear AGAIN. Only that table is pointless because the testing methodology doesn't provide accurate/meaningful results.

I trust results that came from proper testing methodology. Unlike someone who instantly discredit anything they don't want to see (and support the ones they want to see) regardless of the testing methodology.

Oh, and what parameters? LOL

 

We are a very meticulous bunch here. Nothing but the best in stringent methodologies (lab test equipment, testing). We only refer to a select few and we can almost detect any fallacies in testing. At the same time, we applaud and value quality testing. Unfortunately, anything short of that is half assed and misleading... We would rather not review something if we even remotely feel that we cannot deliver the very best in our methodologies. Very high standards.

Chai and Max are the two people I highly respect in terms of methodolgies. The three of us regularly chat about stuff and discuss points back and fourth. We reach a consensus on things after going back and fourth on things. True teamwork right there. But one thing you will never see present is bias and ignorance. Opinions? Always (and educated ones at that - not just childish banter).

lol temperatures are relative to ambient. Temperatures on their own are irrelevant. Delta Ts are relevant.
One cannot compare load temperatures mearly on load temperatures. Heck even my cooling setup would have a high load temperature if my ambient was 40C.

Load/Idle temperatures and ambient temperatures are two halves of the equation. Delta T is the result of the equation.

So what are the temps of it is a good question, but only if followed up with what are the idle temps. Incomplete information is as good as wrong information. Anyone telling you they have a constant temperature short of a proper lab environment is full of it (delta T testing with LAB equipment is the next best thing).

Anyone can write a newegg-like review/advice with minimal (or no) testing.

 

Whoever that still couldn't understand Chai's post, I suggest:

1. Copy down "If manufacturer allows the card to run at 82C, why do you care? They know more than any of us. They create specs, temp tolerance. We only need to adhere to them." on a paper.
2. Repeat 1000000x times.

 

Read my post and view the results again and your response. IF the 4870 has a dual slot cooler and the 4850 doesn't why is the idle temp on the 4870 higher? And then read carefully again and tell me why the 4870 temp is lower under load?

Math doesn't lie and those results make absolutely no sense so I suggest you read a bit more closely before LULZ

Palit Radeon HD4870 512MB Graphics Card Review - Page 20 - Hardware Canucks

 

I disagree completely temperature especially involving semiconductor junctions is the most important parameter of them all. Define stability?

I admire the knowledge and expertise that you, Max and Psyk have involving computer software and hardware but I am an electronics guy and have been since before I started college in 1985.

Under no circumstances with any technology using semiconductors especially MOSFET technology will perform efficiently at temperatures like 80C. You will lose throughput and conductivity. Maybe I am old school but that is how I feel.

Just a FYI I have been running tests including the Crysis benchmark on this 9600GT before I start to mod it. If I decrease temps to the GPU and support components by changing the fan from 35 to 50 to 100, I see improvements on all benchmarks without changing any settings. Temperature affects many aspects of an electronic component of any type including impedance .. sigh

 

It wasn't the temperature I was generalizing on it was the scope of your post. Correct me if I am wrong again but temperature is the most important aspect of any electrical component.

Incorrect

 

LOL since you are such a well versed electronics guy, I am sure you know TDPs right? Different parts have different TDPs? TDPs are used so that the manufacturer can design cooling to keep things within spec (aka not exceeding junction temperature) right?

I'm sure I don't have to bore you with that and the fact that the 4870's cooling has to dissipate ~50w more heat (notice I am not using temperature - but you should already know that) and hence needs a more robust cooling to keep it within operating temperature (however high temperature it is intended to be that is categorized as normal operating temperature).

I mean after all, we as consumers, we do trust the manufacturer has appropriately spec'd things out nice and proper, that the parts can survive through the rigorous ordeals known as normal operating conditions (whatever is deemed normal plus a usual buffer). A bad example of one not properly setting levels is NVIDIA with some of their offerings.

I mean it is just the main reason why cooling enthusiasts spend so much money to run things cooler (aka dissipating the heat) in hopes of increasing headroom by exceeding the cooling capacity of the 'stock' cooler.

But you already know this. You actually already knew why the 4850 has a single slot cooler and runs cooler than a dual slot cooled 4870 that has a higher TDP. Psh. No worries.

Oh yeah, as for your MOSFET example, the junction temperature is the HIGHEST temperature of a semiconductor on any particular piece of hardware (usually the piece of silicon known as a GPU/VPU). If you looked at the high end stock coolers (such as the G80 and G92 - list goes on!), there are power circuitry parts that are actually (surprise!) cooled by the heatsink as well. But you already knew that right?

 

Did you know a MOSFET uses a negative temperature co-efficient?

 

I understand your point but I am going with the basic electronics I was taught and still use. Running a smaller and cooler "die" will always result in a better performing and longer life component.

Then why are you?

 

QFT.

You might want to hit the books again there guy....

Am I the only one that sees the disconnect there?

There is a reason why they are cooled down on graphics and mobos you know (more cooler = more efficient to a certain extent?). Keep them in (again) normal operating conditions. EEs think these things through pretty well. We as the consumer-enthusiasts abide by them (hey I don't water cool my mosfets or use DI on them), but I keep them cool to a certain degree so it does not fizzle out.

But I am no electrical / electronics major (nor do I pretend to be one). I will leave that to Empire23 (resident EE) to explain better or I could ask Azim (hasn't been on here in a while but a good friend IRL - who is a minted EE). Heck I could be wrong as well on some parts (fully expect to on some un-cited parts).

And nice to reply my foot note and not the body.

 

LMAO you don't get it do you?

1. 4870 has a higher TDP than 4850, so naturally it runs hotter, and the fan might be running at very low speed.
2. but the cooler on 4870 is designed to keep the temperature at maximum 82C, as quoted from the article "temperatures never rose above 82C since the fan would immediately spin up."
3. 4850's cooler is different and it doesn't have to keep the chip running at 82C or below.
4. Why? because the 4850 has different temperature tolerance, when operating at different speed/voltage, despite both 4850/4870 are using the same RV770 core.

Is no.2 and 3 sooo difficult to understand? LMAO

This quote suits you perfectly:

 

yadda yadda yadda. You are over generalizing things. Even different MOSFETs have different tolerance/operating temperature range. Most of them operates just FINE at 80C.

Power MOSFET Drivers - Power Management - Analog (Thermal&Power Management)

wow! you increased the performance by decreasing the temperature? that's amazing. omg i'm gonna do this on my E8400 and get some 1337 superpi resultz lmao.