ATI 4850 benchmark

[Yukirin]

Guys, let bygones be bygones. Let's all


Here's the link to the review Mac Daddy mentioned.

Goldfries HD4850 Review.
goldfries Gainward Radeon HD4850 512MB DDR3 Graphic Card Review - HOME OF THE ORIGINAL PC BUYING GUIDE

P.S: anyway, its really a good learning place for me when reading all the debates that has gone through up until now. I thank you all for the discussion

[Max_87]

Originally Posted by Mac Daddy

Bipolar semiconductors have a positive temperature co-efficient while MOSFET's have a negative temperature co-efficient. So in essence by forcing a "FET" based semiconductor to higher temperatures you are basically going against its basic properties in Physics. So it will never be completely stable under those conditions.

Do we understand semiconductors in general now?

Max on artifacts it is HEAT as that is the basis for 99.9 percent of semiconductor failures no matter what parameter caused it. Try googling that

Well besides petty insults and people who need real life experience on more than a few issues lets get back on topic. I do believe Goldfries picked up a 4850 lets see if he records temperatures properly using manual fan settings

You are still saying that? We might not know about electronics as much as you do, but...

MOSFETs (Metal Oxide Semiconductor Field Effect Transistors) have been used in power electronics applications since thee early 80's due to their appreciable current carrying and off-state voltage blocking capability with low on-state voltage drop. They have managed to replace BJTs in many applications due to their simpler gate drive requirements and higher positive temperature coefficient which allows devices to be paralleled for higher current capabilities.

MOSFETs Basics

In general, semiconductors have a negative temperature coefficient, with the one notable exception of the power MOSFET in that its temperature coefficient is positive.

International Rectifier - HEXFET Power MOSFET Product Information

Engineers have long understood that MOSFETs are positive-temperature-coefficient devices. Therefore, as the temperature of the device increases, the resistance increases. In other words, higher temperatures result in lower currents.

Hot-swap MOSFET reliability - 4/1/2004 - EDN

LOL? empire23, we could use some help here Maybe empire23 can explain why all the google results are saying the opposite

Yeah, there's no need to mention that over and over again. All of us here know that when the temperature go out of manufacturer's specification, the cards will produce artifacts.

I wonder who was the one that started making insults first Sometimes these people really amuses me when they act all high and mighty.

[empire23]

I'm not going to say who's right and who's wrong, but i'll try to clear the air regarding MOSFET topology and ideals. The are a few misconceptions flying around so i'll correct them.

1. There's no set definition for negative and positive thermal coefficient. The words themselves don't matter unless they're put in a useful context. Engineers throw around these phrases with charts included, so one has to switch their definition to meet the situation. It's just one of those phrases that changes meaning depending on the person using it.

AMB.org run by Ti-Kan lists it as negative, IOR as positive, Diodes Inc as negative and so on. My advice is to see what the words reflect above all. EE people aren't into semantics, that's why we keep using words depending on like and need.

Even my Modern Electronic Communications textbook by Gary M Miller and Beasly has alot of definitions that my cranky old lecturer Bruce Varnes defines as ;

a) Hogwash
b) Pure Incompetence
c) Miller was educated with a different system than what the world uses.


2. BJTs are the ones that suffer from thermal runaway, as current to current devices as they pass current, it heats the BJT up, and thus the BJT passes even more current which leads to even more heating. This is thermal runaway in the purest sense.

3. MOSFET are the opposite, as voltage to current devices, they take up extremely low levels of gate drive current, in the nano amp region. As MOSFET heat up due to current travelling via the drain and source, their on resistance increases instead, limiting current flow. This acts as a varistor, and like a varistor, it WILL heat up, but it WILL NOT pass more current.

4. Regarding Clocking FETs higher and higher, realize why a boost of say (analogy) 10 percent voltage will net you 10 percent clock gain and as you go higher to 20 percent, you'll only get a say 15 percent improvement instead of 20? It's because MOSFET don't scale linearly after they exceed their efficient operating zones. You get a compressive function of gain versus voltage. There's only so many volts you can push through a FET till that poor depletion layer breaks from the voltage on the gate.

5. Stability in a practical sense is defined by the ability of a device to deliver results within it's specified boundaries. Stability also can be defined as a suitable MTBF. Which we don't know. Of course, stable short term doesn't mean much, as we all know thermal stress can not only stress a part electrically, but also physically, flexing pads, joint stress and other problems manifest themselves in the long term.

An EE/CE/SE's job is to make sure that the device is long in the bin before these effects start to show. Basically, when dealing with high power devices, actually, most devices, heat is the biggest issue of design.

6. Different Processes give different temperature tolerance, different designs also give different power outputs, complimentary logic cells are heat blasters IIRC. Generally we're taught here to make sure MOSFET don't hit the 70+ region for stability, IIRC an LYN forumer Sniper on the Roof who's an industry insider says that too. BUT one can get away with going higher by generally

- Giving long term reliability the ass end.

MOSFETs can run for a gajillion years at lower temps without the threat of high voltage, but people replace their GCs in probably 3 years, so, one makes a compormise in the search for high clock speed and ass beating performance level.

- "Assume" perfect operating spec.

A favourite excuse everyone in this business plays, one assumes everyone lives in the Siberian wastes and then figures that people in hot countries are too poor to buy their stuff. It depends from manufacturer to manufacturer, and it depends on their desperation lol.

ATi during their X1800 days, knowing that Nvidia had em in a pinch tended to stretch their defition of "reliable operating region", Nvidia did the same with their 5XXX series of cards.

Just givin my 2 cents, not supporting anyone. Just here to tell ye what i know.

[PsYkHoTiK]

Very very good post Empire23... +rep

Very good points on the 3 years, MTBF and operating temps and how it affects efficiency.

A really good 'bigger' picture post which coincides with the points Chai and max brought up.

[empire23]

Haha, this thread has reminded me that i want to build a new rig for all my photo editing, gaming and review needs. Going to have stuff shipping to me and hopefully i can get enough work done to sponsor Twin 4870s for this new rig.

Got the 8gb of memory required, got the PSU....now for the rest.

[Unixlord]

A yorkfield and an X48 motherboard, a well ventilated full tower case.

[Lacus]

Twin 4870 O_O..why don't just get twin 4870X2

[Chai]

Maybe he meant twin 4870X2 LOL

[empire23]

Not much money for now, just a twin 4870 seems good. Just wanted to jumpstart my old reviews ASAP and i need a rig to do it.

Not to mention slogging through labview charts, tekscope bmps and other shit takes a lot of effort on this lappie.

[miahman]

i was searching for crossfire compatible GPU. and i came across a graphic card which is DDR3 and DDR5.

i know DDR3 can be used with DDR3 mobo. but in order to use its DDR5 speed do we need to wait for new DDR5 mobo's or is there another way to use it?