4850 tops 3100 points in 3DMark Vantage Extreme
Second INQpressions Edging 8000 in Performance mode
WE HAVE TO ADMIT that having fooled around with the HD4850, that idle temp of 70+ degrees (Celsius) and 83 at load put me at Defcon 1. Sure, GPUs can take those temps and keep on working (the 8800GT did that handsomely, although it was quite a whiner), but why did DAAMIT put out a cool HD3850/3870 and come out with such a scorcher? It looked like the thermal management was on the fritz as the fan never kicked into high gear to cool things off. I rang up Captain Hook of the HMCS DAAMIT who proceeded to explain that he believed the drivers weren’t up to spec, which was the disadvantage of using pre-release samples to produce benchies before launch date. True as that thought may be, however, I don’t have a sample… I have a retail card, with what I would consider to be retail drivers. In that context, Cap’n Hook gave us some top secret post-8.6 Catalyst drivers (which, we presume, integrates the hotfix into the Catalyst) and re-ran the benchies, we also got ‘round to benchmarking under 3DMark Vantage.
When questioned about the choice of cooler, Cap’n Hook said that “customers requested single slot coolers” and the channel considered it was “more important to have a slightly warmer single-slot design than having a dual-slot design”. Uhm… sure. Why not? ATI partners - on the other hand - have already told us they’ll launch dual-slot cooling solutions in a bit, but only *after* ATI dumps the surplus of stock coolers.
Most of you will be asking what’s the importance of whether or not you have a heater inside your PC? Most will say: “who cares, as long as it works”? Well, it does matter, it matters to the HTPC crowd who are air-flow and real-estate constrained, it matters to the overclockers and, personally, it matters because it gives you a good idea of how far you can overclock it and how robust the marchitecture can be. We don’t even want to consider what dual-, triple- and quad-fire setups would do to your case.
Being the unsatisfied customer that I am, and thinking there might be more to the matter than meets the eye, the next step was a simple matter of replacing the stock cooler on the HD4850 with one I had at hand: a Zalman VF-1000.
The HD4850 redone, the Zalman cooler (Top) and the bare electronics (Bottom)
Removing the stock cooler from the HD4850 revealed a lot more about the card than I initially had imagined. It’s quite simple though, unscrew the necessary bits, pry the PCB from the cooler ever so carefully and you get an exposed RV770 core, memory ICs and remaining electronics (do this after the card has cooled down).
The stock cooler unplugged, the new ATI driver did nothing to improve fan control
The VF-1000 is compatible with the mounting holes on the 4850 as we’d guessed by a couple of measurements. Once screwed on, the VF-1000 kept the 4850 GPU at a wonderfully low 35-Celsius. The new cooler doesn’t have RAM sinks, but we guess you can get some RAM sinks and put them on the 8 modules without interfering with the VF-1000 and maybe improve OC-ing performance.
We moved on to overclocking via Overdrive (Windows Vista DRM stopped us from installing ATITool…) Overdrive is conservative about overclocking, limiting clocks to a short window (600-700MHz for clock and 750-1200MHz for memory). Overdrive’s Auto-tune sent the card to 690MHz, but during this second testing session we guessed the card could put out some more. We maxed-out the clock at 700MHz. The memory overclocked to 1168MHz on nothing but air.
3DMark Vantage Stock Speeds vs. OC (700MHz/1188MHz)
We hadn't run Vantage benchmarks last week, so Monday we spent some time with the card and with the OC. No, we don’t have a 9800GTX+ in the lab, but you shouldn’t have any problems comparing the numbers with the multitude of 9800GTX+ benchmarks out there. We think the results speak for themselves…
Stock speeds - Performance Mode - 7069 Marks
OC 700MHz/1188MHz - Performance Mode - 7944
Stock Speeds - Extreme Mode 2799 marks
OC 700MHz/1188MHz - Extreme Mode - 3189
Wethinks things are looking up for DAAMIT, while this isn't even the highest-performing card in their stable. Use a little bit of elbow grease, overclock and the performance is all there. If you're ok with the heat inside your case, $199 seems a good price to go out and buy now. If not, we're sure partners will be launching more cards with better coolers in no time.
If only Nvidia would be so nice to send a 9800GTX+ our way so they can make their point... µ
Comments
Heat inside?
Sounds like u are pretending that the card produces less heat when a better cooler is placed on it. Thtas wrong. Actually the heat inbside your case is the same with the stock cooler and aftermarket cooler.Only that the air in ur case air heats up faster with the aftermarket cooler, than with the stock, cause it removes the heat already at a lower max coretemp. the actual coretemp may be lower with the aftermarketcooler, but that does not affect the heatdump.
Only thing that helps is a 9800gtx like cooler that extracts the air to the outside. But unless there are airvents for that purpose, the casefans have to cope with it. since 4850 actually has pretty significant lower power consumption thatn 9800gtx+ and even 8800gts 512mb, the heatdump from the Powersupply shoudl also be lower. so if u plan to use a cooler which does not transport the heat outside anyways, the heatdump of the 4850 is still pretty low.
More concerning maybe the expected lifetime of the rv770 at such temperatures.
good start
Cooling is certainly still to be improved, not only for the RAM, but especially something should be done about the voltage regulators.An idea would be to saw off the corresponding part of the stock cooler - it will fit perfectly and doesn't cost a thing, and with the downwash from the Zalman (or any similar cooler), it should actually work quite well.
Please could you add the photos!
Where are the pix!What a useless article
So your card is hot. So you mount a 3rd party cooler. So you overclock it to an artificial limit, possibly way below the actual one. Then you post the benchmark results.What you fail to mention is:
Did the card get cooler?
Did your PC case get cooler? (see David)
How far can you REALLY overclock the card now compared to before?
Eh David......
The heatsink, while it absorbs the same amount of heat energy, it is spread over a MUCH larger surface area, so therefore leaves the sink at a slower rate, so at any given time is releasing less heat to the surrounding area than the stock heat sink would, keeping the temperatures inside the case lower.And as its spread over the larger surface area, the heat can be removed easier by the airflow within the case.
Please note;
The HD4850 seems to be almost inaudible! Extremly rare for such a high-end card with standard cooler.So why complaining about high core-temperatures?
Look what Dave wrote, he has a great point.
Charlie Is Collecting Pics.
It isn't well known, yet Charles D. actually has been collecting Info & Pics to make intelligent choice in Buying Computer, When That Time finally arrives.Charles hordes them in ?Windsor Castle or was it in deep freeze of White Castle on University SE & Central NE, we'll never know.
Now Charles has so many pics & specs hidden away for day of reckoning that just few more won't Hurt.
Well Come On: Paul is Now collecting Pics too. Where are they Paul, Hidden in Your Envelope.?
Actually I feel Bashing theINQ Journalist is Poor Mans Cry & in every case very valid unknown issues appear in those articles.
Next: Game Card with Slide In Toast Holder Pics.
Why does 4850 Not run its fan wide open constantly? What makes Extreme score extreme? When will Pauls new found habit of snatching Pics come undone?
drashek
amount of heat is the same
When testing cards for temps, the heat produced by the GPU is constant (or at least should be ... bad test otherwise).The max temperature of the chip might vary from one cooler to another, but the amount of heat generated remains constant.
So if a cooler is capable of keeping the chip at a lower temp, that simply means that the cooler allows for a smaller difference in temp between the surrounding air and the GPU to maintain a constant temp of the GPU. This is usually done with larger amounts of air passing over the heatsinks.
Remember: The same amount of heat is generated over the same period of time.
So it doesn't matter the efficiency of your cooler, it is always spewing the same amount of HEAT into your case.
That said, keeping electronics cooler is always a good thing. And more efficient coolers allow for more overclocking and what not. So it's definitely something to strive for.
Is heat the problem?
If the concern is to take the heat out from the case, why not to try a 4850 with a VANTEC CCB-A6C?IMO, that would address the heatdumping problem.
part 2
isn't this a second part to the first article a few days back? (duhhh)I think the point here is showing that there's hidden performance in the 4850, that might be unleashed by using a different cooler. Most reviews state that even the PCB on the 4850 gets unusually hot, and that limits overclocking performance, doesn't it?
ur base
ATI to Nvidia: Im at ur base, ste@ling your customersit is hot because the stockfan only runs at 25%
it is that hot because the stock fan only runs at 25%. Just wait for the RivaTurner 2.10 and you can kick the fan to 50% or 60% and the temp will drop 16-20 degree easily, while keeping the card inaudible.thermal coeffs
I didn't actually read the article just the comments.Heatsink performance is measured in degC per watt. A good passive heatsink might have a thermal coeff of 0.9DegC/Watt. What this actaully measures is the thermal gradient across the heatsink from the device to the air temperature around it. So this means that a better heatsink is actually cooler to touch than a poorer heatsink with the same energy input. This is because the more efficient heatsink has a lower temperature rise for an equivalent amount of power dissipated.
This means that the heatsink is a lower temperature, so the device stays cooler. This can be construed as "producing less heat" because the temperature at the surface is lower because the sink removed it. The device "dissipated the same power" which is not "produced the same heat" as that depends on the heatsink. So in fact the author is correct, except on a point of grammar.
All this depends of course on adequate air flow in the case, otherwise the thermal efficiency of the sink is reduced.
So in short the temperature will be lower, but the power dissipated will be the same - unless as David correctly pointed out the airflow in the case is very poor, in which case it heats the case up quicker.
Riding the Thermals (not a ^%%Drashek)
Actually efficiency decreases with increasing temperature due to increased leakage currents, shifts in operating area characteristics, etc. Without limits or cooling the eventual result can be thermal runaway and self-destruction... So running the circuitry at a lower temperature does result in a marginally lower power consumption (i.e. heat production), and more importantly to OCers, faster operation.p.s. verification code: zjamdog... I like it
Where are the referenced pictures for this article, and more importantly the pictures of the CMMA girls mentioned in that bit?!?!?
to limit overclocks?
I'm betting my money on ATI not going with a great cooler and limiting overclocks to 'only' 700mhz to prevent sales of the 4850 killing the 4870 stone dead...um helllo?
the 4850 presents itself as a fantastic new gen card,the REASON the card is so HOT is that the fan controller does not speed the fan above 10% try cooling a 2900xt or 9800gtx at 10% or even may i say, a gtx 280 at 10% fan speed.
I had that same problem with my 3870, it burnt my hand wen i went to see how hot it was, reason, 10% fan.
at least, you definately wont hear the card/
anyway, once rivatuner supports 48xx series
putting the card's fan to a reasonable speed (50%) will warrant plenty more overclocking headroom.
you do not need a custom cooler,
the overclocks are not hard to get
any novice who knows how to turn the fan speed can do it. hell, it was overclocked in CCC
pfft hard to get overclocks my ASS!
Thermal awareness day.
Heat is bad because it increases electronic resistance, moreso in narrow trace chips, increases noise and likelihood of errors like artifacts and lock up under stress, decreases life of components even at specified operating frequencies.80'C is hot. So logic suggests we should watch out for a high rate of early product failure with this card.
SInce the Zalman reportedly reduces temps to 35'C, the conclusion is the cooler is inadequate and the fan runs too slowly at the default setting.
Whether ATI care is another matter. Maybe they are avoiding certain identifiable issues such as double slot cooling and also noise reputation because they see these interfering with uptake. Presumably they are prepared to lose a few RMAs to avoid the obstacles especially in the early days where first impressions are formed. But then they would gain a reputation for unreliability, unless their cards are magic and defy the (old) laws of physics and can run that hot. Alternatively they are simply doing an AMD (being incompetant and then hiding from the facts) and are selling cards with slow running fans because they cant do any better.
I think the magic card scenario is the least likely somehow but if true there is one gripe that the case heat issue does apply despite an external venting cooler particularly in marginally ventilated cases because if the core is at 80'C the thermal gradient is large and heat will emerge through all conductive pathways, including the back of the card and even the PCI bus interface, thus heating the mainboard which may not be made of magic.
Most overclockers aim for under 50'C. Mind you I did run a FCPGA at 60'C for a number of years until I discovered the HS was only making contact down one edge. But 80'C is a justifiable cause for concern judging by previous generation's standards.
ASUS software can control fan speed
I purchased the ASUS 4850 card -- the included software (ASUS Smartdoctor) allows you to change the fan speed - Manual, Smart, or Auto -- my idle temps never go over 50 degrees C. Even playing games, it never goes over 80 degrees C.I Wonder..hmm
I wonder how the temps would be on the 4850/4870 using the single-slot cooler that Sapphire uses on its Atomic and Toxic line of 3870's?had the same problem
I went from an X1900XT to an HD4850 and my case temps jump 20 degrees F... I'm going to have to get a PCI slot fan or a new cooler for this thing asap.Get it right
First I've been watching heat in systems since we put the first computer in Canberra's Defense Buildings. Excessive component heat = component failure. Note the word component.So the zalman does two things, one reduces build up of component (there's that word again) temp build up and as it is more efficient allows less rpm on the fan and makes it quieter. As for the build up of heat in the case, most cases are rubbish. I replaced the sides of mine with alloy mesh and threw out the noisy case fans years ago.
In places where the summer temp can get up to 40c and no air conditioning you better watch your temp or shut it down.