Oh the irony, you say don't judge something by its cover (ati) yet say nvidia's card will be faster? Do you have any proof for that?
kthnx
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
5870 - $299
- Thread starter bigdogchris
- Start date
Oh the irony, you say don't judge something by its cover (ati) yet say nvidia's card will be faster? Do you have any proof for that?
kthnx
He didn't say will, he said "should" and "if," neither are declarative statements. Considering NVIDIA is slated to release it's part at least a month and a half after AMD's, it SHOULD be faster if they want to make any kind of impression on the market.
mastercheeze
2[H]4U
- Joined
- Jul 23, 2004
- Messages
- 2,799
Same here. Especially at the res I game at (1680x1050), my 4870 keeps chugging along. I might just spend my 5870 monies on this newfangled SSD thing I keep on hearing about.
Yeah really,I already beat Crysis/warhead, yes it would be interesting to play it maxed out but after a week I will only get bored with it again.Crysis 2 will probably be a console port with minimal increases in graphics for pc.
The only game I have heard about that I'm getting pumped up for is Battlefield 3.Anyone know any other PC FPS games that will be DX11 that will push the boundaries further than Crysis?
Other than that the 4870 I'm using plays every game smooth enough.
Zorachus
[H]F Junkie
- Joined
- Dec 17, 2006
- Messages
- 12,402
Awesome games to upgrade for;
Operation Flashpoint 2, due Oct
http://www.youtube.com/watch?v=J7jdtSszBT4
Borderlands also due in Oct;
http://www.youtube.com/watch?v=wG7lSTFtZ9Q&feature=related
Left 4 Dead-2 due in Nov;
http://www.youtube.com/watch?v=T71rp_IBzOY
2TFLOPS? $299?
I don't know how true this is....
http://www.product-reviews.net/2009/08/28/2tflops-for-radeon-hd-5800-series/
I don't know how true this is....
http://www.product-reviews.net/2009/08/28/2tflops-for-radeon-hd-5800-series/
- Joined
- Feb 11, 2008
- Messages
- 5,295
So any word on when a 5870 X2 version will be released? Hoping they won't wait three months after the release of the single-GPU cards... If that's the case, likely I'll be going back to nv since the 300 series should be faster than the single GPU 5870 and going 5870 crossfire is looking to be a step back/co-equal with my 4870 X2s.
FYI, FLOPS capacity does not equal better gaming performance.
Rumors point to November.
lolSinglePrecision.
Actually, 58xx itself will trigger cuts in 48xx (and 46xx) pricing, simply to get the darn things out of the warehouses.
And unless GTX3xx is markedly faster than 5870, it would be an unmitigated disaster.
Currently, only HD4890 remains above $200 retail (and that is just barely); when 58xx arrives in the warehouses, the discounting already taking place will only grow.
Well, with the latest rumors pointing the 5870 to being 60% faster than the 4870 there is room for the GTX3xx to make a serious move.
- Joined
- Dec 9, 2000
- Messages
- 16,278
seeing as Nvidia claims that GT300 will have 512 SP, there is a good chance that it's going to be faster then the 58xx series cards if they are only 1.6x as fast as the 4870
Excuse my technological ignorance, but isn't the NVIDIA part MIMD? So is that 512MIMD or an actual fewer number of units that give a maximum theoretical performance of 512 SIMD units? The reason I say this is 9 months ago they couldn't even shrink a GTX280 enough to fit two separate physical GPUs on a card (enter the GTX 295 with two GTX 275 GPUs), but now they somehow managed to fit more than twice the number of MIMD units on reasonably sized die (the GTX series, even at 55nm, was still huge), within a decent TDP? I dunno, seems too good to be true.
The 260 and the 280 share the same die. When a 280 comes out with bad SPs, they turn it into a 260.
They were looking at keeping cost down, not couldn't.
Frankly, the TDP isn't as big of an issue as getting rid of that much heat. Which may have been what you ment, I'm not sure. There are still some tricks for them to pull with coolers, but there is really only so much you can do when you are trying to dispate 300W+ in a 10.5x.75"x4" volume without bursting ear drums.
Going from 55nm to 40nm will help a lot. It is almost a 2x space required reduction. Don't forget about the voltage drop that comes as well. Really, we'll just have to wait and see what they decided to do because in the end, they could easily make a graphics card that was 100 times as powerful if they were willing to have a die area of around 500,000mm^2 and only have 1 in 1 billion yield.
What video card has a TDP of 300W?
The numbers on [H] reviews of power consumption of beefy systems is the complete system power dissipation. The video card is probably consuming less than half that.
http://www.nvidia.com/object/product_geforce_gtx_295_us.html
Specifications >> bottom of page >> Maximum Graphics Card Power (W) 289 W.
You got me there, but the caveat is that's a dual GPU card.
Also, practically speaking the power consumption is more like 210W:
You got me there, but the caveat is that's a dual GPU card.
Also, practically speaking the power consumption is more like 210W:
"Common" power consumption is an odd beast. Is Prime95 a good power consumption test or is Intel Burn Test a good power consumption test for a CPU? We know intel burn test will produce much higher temps (and draw more power) than Prime95, but we rarely ever see that while doing any number of "CPU limited" tasks.
Actually, the reason they quote such a high figure for power consumption is probably because they can't count on your average consumer making a good decision when buying a power supply, so they inflate that number slightly. Also, if the GPU is burning 210W measured on the rails, it's pulling anywhere from 10 to 20% more than that from the wall, due to the loss in the PSU.
You still can't show me a single GPU card that's remotely close to 300W. I guess that means you're absolutely right about troubles trying to dissipate 300W in a video-card sized package...
Something you might note. The graph you linked showed a 285 consuming 150W and the 295 consuming 200W. Fankly, that doesn't make much sense to me.
In any case, heat load imho is a much bigger problem then TDP as really to give it more power you can just keep adding 6 or 8 pin power plugs to your hearts content.
- Joined
- Dec 9, 2000
- Messages
- 16,278
I'm a bit of a newb with this but
SIMD = single instruction multiple data
MIMD= multiple instructions multiple data
each has pros and cons, but there is a pretty big discussion on which is better for graphics and HPC (cuda / quadro)
currently NV is SIMD, intel is going MIMD with larabee and NV is supposed to be going MIMD with GT300
I don't see MIMD or SIMD units being smaller / larger usually SIMD is "easier" and less "expensive" to implement though.
I don't see a problem with the GTX285 taking 150 where the 295 takes 200w
the GTX285 is clocked higher, has some "active" units etc. and possibly the GTX295 was held back by the CPU in that testIt makes perfect sense since the 295 is 2x 275s, and nvidia would probably select better binned chips for the 295.
Secondly, you can't add power with adding heat. 99% of the power juicing through your computer is dissipated as heat. It takes a minuscule amount of power to actually flip bits in a chip, the rest is just wasted.
Hrm... it seems Nvidia claimes TDP of 204W on the 285 and 219W on the 275. Strange.
I agree completly. I was stating that there is no problem with supplying more power to a card, there is a problem with removing the heat that power causes from the card.
Binning can have very significant changes in power consumption for the same reason you get higher overclocking: higher clocks, less voltage. The less voltage you use, the less hard the chip crunches with every clock cycle.
I doubt the pixel shaders have much to do with the power consumption, but the buses probably have a lot to do with it.
And the 295 is clocked slower than a 285.
I trust xbit labs power measurements as they use some very sensitive equipment to make their measurements, that take into account things like power factor and what not.
If you don't trust xbit how about this graph from [H]:
Quad SLI, system power consumption = 290 W total system power. If anything, that [H] graph is screwy because look at this one measuring the 275, I find it hard to believe that a single 275 draws more power than a 4870 x2 crossfired:
Basically, I trust xbit because they're my favorite for measuring power. They do it most consistently and compare against broad range of the competition, but really it's hard to trust anyone's power consumption measurements. One thing that holds as a general rule, however, is that actual power consumption is usually AT LEAST 10% less than the TDP. Pick your favorite site, though, and they're probably wrong. Everyone and their cat always posts wildly varying numbers for power consumption.
---
As far as your second statement, I'm sorry I misunderstood.
Yeah that's why I'd rather refer to TDP. It may be high, and it may allways be 30% high, but at least among a single brand it should be somewhat consistant.
Saying "heat load is a bigger problem than TDP" does not make much sense because TDP is a measure of heat dissipation requirements.
http://en.wikipedia.org/wiki/Thermal_design_power
Edit. Other than this nitpick, I agree with your point. Power is like food: whatever goes in must come out. And coming out is usually the messier part.
Last edited:
You are right.
No, they couldn't, otherwise they would have. If something is possible but not feasible due to budget, physical, or some other kind of restraint, they still failed to do it - i.e. couldn't. Shrinkage brings pros and cons - I don't know if it's directly related, but everyone remembers intel's failures in the Prescott shrink to 65nm. My point is that what is currently being reported about the GT300 is pretty impressive, and I'm skeptical.
Well, to clarify, what I mean is this 512 number that's being thrown around is that:
A) Actually 512 MIMD units (which I find hard to believe)
or B) "256"MIMD units that can give the equivalent theoretical performance of 512 SIMD units under certain situations (the 256 is arbitrary, it can be any number between 2-512, I just used it to make a point).
Finally, this Xbit lab numbers are about right, but these graphics cards are power hungry beasts. When I had my GTX 295 heavily overclocked (but at stock volts), my system was drawing 515-520W from the wall. That isn't to say it isn't a great piece of hardware, because it is. No other piece on the market matches it's performance and it still maintains a very respectable performance/watt ratio (you'd have to look at 4770 CF to get better numbers). For now, I've actually reduced it to stock clocks again and undervolted it at the same time, and power draw from the wall has dropped to the 405-420W range. This number actually impresses me more, considering I can get GTX 260 SLI performance with about 2/3 of the power consumption foot print.
Wow, one of the only other people on [H] forums that thinks that performance per watt is a [H]ard ratio.
I'm going to hypothesize that it's easy to have this viewpoint when you live with your parents and don't have to pay for most of what you use.
I believe he was agreeing with you.
Oh, and frankly, 20$ will buy you 750 Watts for 8 hours a day 7 days a week for a month. And durring the winter, it's free as you would otherwise spend the same money on your electric heater to heat your house. For some of us, 20$ a month just doesn't matter.
AboveBeyond
Weaksauce
- Joined
- Mar 1, 2007
- Messages
- 79
Damn $399 for the 5870? If true, this pushes it out of my budget. So I guess I may be looking at the 5850 then...
You don't know how much $20 will buy you because
a.) energy costs are different in different parts of the country.
b.) energy costs are different during different hours of the day.
Secondly, it's always worth it to conserve. $20 x 12 = $240, a free plane ticket every year. What if someone has more than one computer? That's $480 in savings. What if the machine is on 24/7?
Saving $20 a month, at 4% interest compounded monthly will net you $14,000 after thirty years, even though you only deposited $7200. That would make a nice down payment on a Corvette. Or a second Corvette.
Not to mention the environmental side of things..how much less coal would we burn if everyone reduced their electric bill by $20 per month?