I'd do that to if it weren't so much more expensive to make a bunch of water loops.
I think I'll stick with massive heatsinks for high end aircooling lol.
Printable View
I'd do that to if it weren't so much more expensive to make a bunch of water loops.
I think I'll stick with massive heatsinks for high end aircooling lol.
Yeah, or massive air. Either way, the starting clocks aren't too terribly important, as long as it can clock well when modded, IMO.
3870 = 105W TDP, 3870X2 = 190W TDP.Quote:
Originally Posted by Caveman787
4850 = 110W TDP, 4870=160W TDP.
I´m not dumping on ATIs parade. I´m just not jumping the hype race.
Anyway, show me how a real 4870X2 with 4870 clocked GPUs would be doable without a 3 sloot cooler design or a dustbuster. Plus something that the power connectors would allow.
So yes, it is impossible with 4870 speed GPUs.
yet the 4850 can reach 4870 speeds and Gddr5 supposedly takes less powersso it could be possible.
Funny how gtx200 hype was fine lol.
I dont think GT200 hype was fine. But it was small.Quote:
Originally Posted by Caveman787
Do I need to say minicores, shared memory etc?
You do know 75W is supplied on PCIe 1.0 via the slot? Or 150W on PCIe 2.0, but thats irrevelant since it goes for the lowest dominator.
So 4850= 75W+75W=150W possible.
And you type like someone familar ;)
Slot+6pin+8pin matches precisely what 2 4850 would require.
pci-e 2.0 is specified for 300W max. so i dont think we wont see the X2 at 4870 clocks.
What?Quote:
Originally Posted by Shintai
75+75+150=300w...
Since when does two 4850, which should need under 200w, now need 300w?
It is definitely possible for there to be higher binned cores on the R700 and still stay under the 250w.
I asume you would not OC your X2 card then. Because there wouldnt be any headroom at all.Quote:
Originally Posted by LordEC911
4870 is 160W, 2 is 320W. If we look on 3870 vs 3870X2 (105 vs 190) only 20 watts TDP was saved. Or perhaps that wasnt better binned cores...
Maybe I should also remind you of the 3870X2 with connectors?
http://www.sdbear.com/2/lib/200801/29/361/1.jpg
Did it need 300W?
The 3870X2 (225W) ate more than 2x the power of a single 3870 (105W), due to obvious voltage increase for 850 MHz core. It traded blows with the 9800GX2 for the highest power consumption, depending on which games were being played.
I guess the 4870X2 will actually have lower speed cores (lower than 750 MHz) or at least *slightly* lower voltage with carefully selected quality cores in order to meet the 300W TDP limit. Almost guaranteed that it wont be any higher than 750, since ATI wont have any competition for their king-of-the-hill 4870X2 that completely destroys the 9800GX2.
ATI better be using a slightly higher speed GDDR5 memory, at least 3800 or 4000 MHz rather than 3600 MHz for a single 4870, since it should be more important for players who use those cards to game at 2560x1600 resolution. It was a stupid move by ATI to do GDDR3 instead of GDDR4 on 3870x2 cards.
wait aren't you proving my point?
The Power will be shared between the 2 and since somethings wont be requiring power since they wont be there.
Besides I'm sure 300 watts would be enough.
Regardless enough speculation I'll wait and see but either way I know the R700 is gonna be a powerhouse.
Eh just because ATI can lay back and enjoy there moment doesn't mean there gonna do like Nvidia did and start screwing over there customer's. Atleast I hope not.
It wouldn't be smart for them to try and screw their customers as they need to regain their market share. You won't be able to that without having a seriously good price/performance ratio.
BTW, I'm sure power draw for the R700 will be smaller than 225 Watts otherwise they would not use the card with two 6-pin PEG connectors, just check it here:
http://img.hexus.net/v2/graphics_car.../R7001-big.jpg
Yeah I noticed that to but we don't know under what kinda load that card was.
Heck we don't even know if that card was fully functioning lol.
But You very well could be right and a lower tdp would technically be better because it's gonna be fairly hard to cool 2 cores that close together I'd assume.
Remember that the 3870x2 also had higher clocks, meaning they saved 20w will increasing the clocks quite a bit.Quote:
Originally Posted by Shintai
There is obviously quite a bit of power savings that happened in the 3870x2 and to imply that none of that will happen with the R700 is simply ludicrous...
GDDR5+unknown memory setup+smaller PLX+one PCB and probably countless other stuff.
3870 base card was already lower TDP than even a 4850 is. Yet the exact same design as a 3870X2 can now hold 4870s with a 50W higher TDP than a 4850? Or 55W higher than 3870 for each GPU? Someone just made 100W TDP go away?Quote:
Originally Posted by LordEC911
I´m not saying nothing is happening. Even with 4850 GPUs there would be a 10W or more savings. But you suggest something way beyond in the excess of 100W.
Helmore´s picture is also pointing towards it.
No I am not...Quote:
Originally Posted by Shintai
I am suggesting that someone is relaying incorrect numbers.
A 40-50w savings is all that would be needed.
Also Im amazed no ones thrown out the idea of conservative TDP as in the card will most likely be well within the number.
Plus since powerplay isn't working right maybe there wasting power on both end's load and idle. Because the 4870 pulling so much more juice then the 4850 doesn't seem to make much sense.
When is the R700 information and reviews gonna start showing up august?
I'd say you are just speculating here :D
LOL..missed the reviews? Why you think the 4870 uses so much? Why do you think a 3.2Ghz Yorkfield uses so much (100%) more than a 2.66Ghz? Replace those with 4870 and 4850 and you got your answer.Quote:
Originally Posted by Caveman787
Or why do you think a 8800Ultra uses so much more than a GTS?
When you reach the limits power consumption increases drasticly. At a certain point you would simply pour in power without gaining any performance.
Lol yeah well that's all we can do at this moment.
Though we all wish we knew exactly what it was like.
Whats your view guys on whether its worth buying factory oc cards (like the diamond xoc) for the higher binned chips rather than buying reference and oclocking it yourself. is it worth it for higher binned chips or not :shrug:
So far i buy stock but i might pay more if there really is a gain for the extra money.not totally convinced though :shakes:
duh... i would recommend you to check out the 8pin connector -> the pci-e 8pin connector only has 2 more ground wires nothing more... imho the introduction of the 8pin connector was one off the biggest ripoffs :pQuote:
Originally Posted by Helmore
Number of connectors is no indication of what a card can consume.
They only time higher binned cards that are pre overclocked do much better is if the up the voltage for more overclock headroom.
For $50 more i'm not sure it's worth it unless you get alot more headroom.
Yeah it doesn't tell you how much it takes but it gives you the ability to better guess.
so if R700 has two cores @ 625mhz, what do you guys expect the highest core OC to be before we reach temperature limitations, or limitations due to power draw?
I bet we're going to see 900mhz R700 overclocks in this very forum! ;) It will be funny to see if those cards are pulling 299 watts @ load :)
Ive always thought i could do whatever the manufacturers could do for free with a bit of advice,, the oc are usually conservative but theres that element of chance it could go much higher..
I think id pay a little bit more maybe but im still not decided
Quote:
Originally Posted by Helmore
I think that the pc in the photo isn't working and that is only for demonstration. Just look at the cpu fan. It isn't working !!! So, they only had 2x6-pin cables and so they put them, because if it isn't working, it doesn't matter. All of the above of course is pure speculation.
An article on Nordic Hardware says that there is indication out there that the 4870x2 will outperform 4870 crossfire by an average of 15%.:up:
http://www.nordichardware.com/news,7900.html
Or it could be off :rolleyes:. Occam's razor.Quote:
Originally Posted by eddieobscurant
Why on earth would they show a mock up card? If it is a mock up card and is actually just light plastic, it looks very impressive. Its the real deal. What you can argue is that they could change the design from an ES version.
I think you misunderstand me, anyway maybe it's my fault cause my english ain't that good. Of course it's the real deal an not just plastic and of course is a working sample. I meant that maybe, and I'm just guessing here, that wasn't working and wasn't mean to work at the exhibition and maybe their psu had only 6-pin cables and so they put them even if it wasn't enough for the card to work. Anyway i'm saying all these only to mention that maybe 2x6-pin aren't enough for this card to work.Quote:
Originally Posted by ghost101
um, look at the 6pin plug on the right. isnt that an 8pin socket but it has a 6pin plug in it?
would you even need the 8pin with a pci-e 2.0 board?
15% more than the 512mb 4870's?? or 1gb 4870's??Quote:
Originally Posted by Mr.BSEL
i want to chuck a 4870x2 in my lan and rig and im hoping a 620w is enough to run this.
Well a 4870 draws about 160 watts max I believe. So factor in say 80% more draw for another GPU, and it's nearly 290 watts. Ouch, I hope that is not true that's too damn much for a single card. Unless my numbers are wrong somewhere.Quote:
Originally Posted by disruptfam
Uh, no, for the fan to be a 'blur' it must be moving faster than the camera's exposure time can snap it at. That's why most pictures will have the fan looking like it isn't moving. Even mine turn out that way. You have to lengthen the exposure time for the fan to become a 'blur'Quote:
Originally Posted by eddieobscurant
hmm... i didn't think that, probably you're right
With 4850's clocks, 4870x2 performs %15 better than 4870 Crossfire? I won't believe it until I see it
We don't know the clocks yet so far like everything else it's just speculation.
But considering it's 2 chips on one card it being faster then crossfire by a little bit would make a lot of sense.
Well this PR slide that was leaked for the R700 in CFX shows 750Mhz clock. Not sure if they backed it down for release due to TDP or power draw or whatever, but this is what we had a few weeks ago that was leaked:
http://www.ocxtreme.org/opb/hd4850/r700slide.JPG
WOW, then the cooler must be really awesome
If the score is true, then it looks like 4870x2 in crossfire scale very well since a single 4870 gets around 3600 on the extreme preset so theoretically, 4 of the 4870 would give 14400. The test shows 12515 which means about 87% efficiency, at least on 3d mark vantage.Quote:
Originally Posted by zerazax
Well as amazing as that sound's and wonderful as that would be I feel that is somehow optimistic.
Unless of course since the 2 chips act as one card and therfore the second double card acts as crossfire. It would then have scaleing similar to crossfire but with effectively cards that are double the performance of 1 4870.
Yeah that came out very confusing eh
Let's see
Single 4870: 3,600
CF'ed 4870 (1.7x scaling for extreme preset): Around 6,100
Plus %15 for 4870x2 = 7,100
CF'ed 4870x2 (again 1.7x) = 12,100
Reported score: 12,500
Not too bad, eh :D
That is some damn good scaling! Very nice!
I'll be waiting on one of these by the looks of it as it should save me upgrading my motherboard etc.
Cant wait to see some numbers and reviews, roll on August!:clap::D:eek::shocked:
Settle down there Shirlock!Quote:
Originally Posted by eddieobscurant
In case you havent ever noticed, cameras take pretty quick snaps. I havent seen a photo of cpu fan moving?? ever!
:shocked: :shocked: :shocked: :shocked:Quote:
Originally Posted by Mr.BSEL
I have a bad case of the wants.
TDP is ~160w, realworld max power consumption is ~140w.Quote:
Originally Posted by eleeter
140x2=280w, which is why I brought up that they only need to cut 40-50w.
http://www.tgdaily.com/content/view/38171/135/
x12515 in Vantage x mode.
:D
Does GPU-Z show 1GB memory on the HD3870x2?
Good catch... no it doesn't
I think this stuff only had one backward, the heat problem ....http://www.xtremesystems.org/forums/...lies/frown.gif
:(http://www.xtremesystems.org/forums/...lies/frown.gif
:(
If it can go max 80 C under full load......i will grab it
When you look at it that way the rumored score of 7K makes perfect sense. If this is truly only at 4850 core clocks then that is very impressive.Quote:
Originally Posted by annihilat0r
Don't really care about heat as long as it can handle the heat without throwing a hissy fit.:up:
The R700 boards were clocked at 778 MHz core, while the GDDR5 memory was clocked at 980 MHz QDR (effectively 3.92 GHz).
This thing looks awesome :) Black is sexy :)
Quote:
Originally Posted by zerazax
Same source I assume, how come the clocks are different?Quote:
Originally Posted by tdream
Things might have changed since then? That PR Slide was from about 2 weeks or so ago which is an eternity....
Interesting to see that GPU-Z did detect 1GB while for the 3870X2 it only detects 512MB. Does GPU-Z actually detect memory or is it just a database even for the amount of RAM? I'd think that detecting RAM on a GPU is one of th easier tasks...
You do quite a bit of double posting Zerazax :p:, you first ask thing over at Beyond3d and then here :p:.
But to answer your question; I don't know :p:
I'm also doubting the clocks, R700 at 750MHz. just seems way to high to me for the power figures that have been stated in some earlier articles and it's not like they have to outperform 4870 in CF. They only need the single card performance crown....
Well neither has produced an answer so I'm going to keep asking :p:
I think they will lose some market share to 4870CF if it's slower. The people who buy these sorts of things are looking to have the fastest solution possible. Only some of those people actually care about how many of their PCIe slots are taken. In my own case I am using a uATX board with only one PCIe slot. So it is more compelling. But most people with an X38, X48, or X58 should be able to run CF without much of a downside. That is unless the 4870x2 really does have some new tech in it's CF implementation.Quote:
Originally Posted by Helmore
It probably will have some new tech in it's implementation and if they only price it at around 150% of the price of a single 4870 then it is justified to not perform faster/same as a 4870 CF set-up.
would be very impressive if these worked well in xfire.
churnem out baby.
and work on them drivers :hehe:
power seems pretty good for such a monster (gulp)
Oh man is this thing ever going to run hot :p
'support global warming(or cooling/ wtf), upgrade your graphics solution now!'
:hehe:
i can only game for 1 third of the time due to tdp, but oo it looks good:D
im hoping some1 comes out with a single slot air solution to cool both gpu'sQuote:
Originally Posted by halo112358
15000rpm fan? :rofl:Quote:
Originally Posted by disruptfam
that is actually doable.. but not with normal copper or aluminum coolers (resulting in a much too high cost)
im running out of vram at 19x12 4xAA with my 8800gt (still m2tw) - rivatuner tells me so:
http://i181.photobucket.com/albums/x.../19x12-4AA.jpg
@504MB vram-> big stutter/lag and fps drop.
i think im low on bandwidth aswell.... cos similar lag effect occurs when vram not maxed.
My point was its the exact same score? Higher clock, what it got slower :shrug:Quote:
Originally Posted by zerazax
Quote:
Originally Posted by tgdaily
Just take the motor out of a 15k raptor... :DQuote:
Originally Posted by Shintai
nVidia's GPUs hog more RAM in general.
*want*
Maybe not in general, but their G80 architecture and everything derived from it do compared to ATI's HD2xxx and higher series (R600 and derivatives). At least that's what testing with their HD3870 cards has so far revealed, it could be different with these cards....Quote:
Originally Posted by Macadamia
Although I doubt that....
That's very true, and from what it seems, the 4850 doesn't lose much ground to the 4870 due to memory bandwidth either, if you factor in the clock differences the bandwidth really only gives an extra 2-3% and that's only with max details. Thus I am extremely happy with my 4850 since I only paid $150 W00TZ!!!!!!
Just wanna comment on the color as so many are "black PCB omgwtfbbq". Just cuz this sample is black doesn't have to mean retail cards will be. Same concerned early NVIDIA samples in the past, was colored black but retails appeared the usual "green". That's the first thought that popped up in my mind seeing the black PCB at least. I wouldn't count on it that it will be black that's all, but what does the color really matter anyway? :p:
Well diamond's OCX Black edition has to be black for sure, but other vendors could go with the traditional red. But black just looks so much cooler with that red and black heatsink blower setup, I don't think ati would change that especially with how awesome the gtx 280 and 9800gx2 shrouds are. If anything, nvidia is great at designing shrouds and to the least ati can use a black pcb
Where? In the SS it shows 750 Core, 900 QDR.Quote:
Originally Posted by tdream
Thats what the Nordic Hardware little preview blurb said
Ehm, isn't this a little misconception? The command clock runs at that speed while the read/write clock runs at twice that speed but delivers data in a (more ordinary) dual data rate fashion. So it's basically the chips true command clock, while data transfers truly happen at twice the clock. No QDR anywhere in there, only DDR....Quote:
Originally Posted by annihilat0r
bingo. :up:Quote:
Originally Posted by Helmore
Quote:
Two R700s Churn-out X12515 in 3DMark Vantage
AMD, Austin have managed a benchmark score of X12515 in the 3DMark Vantage benchmark using two Radeon HD4870 X2 cards in CrossfireX, a feat that takes three GeForce GTX 280 cards in 3-way SLI to achieve. The R700 boards were clocked at 778 MHz core, while the GDDR5 memory was clocked at 980 MHz QDR (effectively 3.92 GHz). This brings the total on-board video bandwidth to a stellar 250.8 GBps.
http://www.tgdaily.com/images/storie...nphenom_02.jpg
With inputs from TG Daily
http://forums.techpowerup.com/showpo...77&postcount=1
regardsQuote:
Radeon HD4870 X2 Readies for Release
The ATI Radeon HD4870 X2 codenamed R700 is slated for a week 4, July release. This dual-GPU card which features two of the RV770 graphics processors which have been an overnight success for AMD/ATI in their single-GPU avatars, the HD4850 and HD4870, will be priced at US $499. AMD claims that this card will overwhelm the GeForce GTX 280. AMD will be in a position to supply samples for internal testing and quality evaluation by 15th of July and July 24th is slated for completion of testing following which, production of a 'non-public' version could commence.
HKEPC
http://www.techpowerup.com/64378/Rad...r_Release.html
I got to hold one of these yesterday, just like a 3870x2 with a few components moved around. Can't wait for them to hit the shelves, it'll do well on stock cooler :up:
4th week of July, that's exactly in line with what the slides have said. Those slides were from the first week or so of June and the 4th week of July is around 8 weeks later....
I'm eager to know what the card will offer, especially in what way it will differ from the 3870X2. BTW, have they decided upon a name, or is the jury still out on that one?
I just sold my old 3870x2 this weekend to make room for this card. 12.5k in vantage at those clocks is pretty amazing granted there were 2 in crossfire.
Can't wait!
At first I wanted a 4870, but now seeing this... :eek:
Would go well with my new q6600! no comments about this AMD guy buying an intel.... :para:
Pardon me if I missed it and it was said already, but approximately how much is this going to cost?
Around $500
Well although my other hobby (Home Theater and Speakers) is getting most of my money leftovers, I think it is time to have a serious look at this.
What I want to see is how it performs at 2560X1600 with at least 2xAA &16AF since this is the only resolution I am interested in! I want to believe that using two 4870X2 on water and overclocked to death will be enough or else I will be seriously disappointed at this time and age.
I wonder when they will have 1GB versions out.
Whoa! Card looks super fast!
Can't wait to buy one and slap a water block on it.
So better have PCIe 2.0 slot for this?
Was hoping I could get 100% from this with P35 board (just 1 card).
I'm sure you will, I see no problem with using PCIe gen1 when you only intend to run it as a single card.
i wonder if this card will have problems on the dfi x38 like the 3870x2 did? or vice versa
P35 has 1 PCIe 2.0 slot. (top one). The other is PCIe 1.0 electrical x8Quote:
Originally Posted by Red Evil
Uh, what? P35 has no PCIE2.0 slots.
The P35 has one 16x connector PCIe 1.1 and a 4x (electrical 16x) PCIe 1.1.Quote:
Originally Posted by Zytek_Fan
There's no PCIe 2.0 on P35.
Yup and that second physical 16x PCIe slot is connected to the southbridge through a 4x electrical connection as you just said, but this ads another bottleneck as it has to transfer data between the north and southbridge when using CF and this connection between these 2 bridges is PCIe gen1 4x in bandwidth. So there is a bit more latency and then there is the fact that this connection has to be shared with other components like HDD, USB and ethernet. All in all, it's quite a significant bottleneck.Quote:
Originally Posted by ToTTenTranz
whats about the NDA? :rolleyes:
4 more weeks to go.Quote:
Originally Posted by Taoof
so this is probably a waste if i'm still on a p35 board? grrrr
+1Quote:
Originally Posted by LordEC911
:bounces:
No, why?Quote:
Originally Posted by skindogg
Card use a lot of bandwitch = PCIe 1.0 maybe not fast enough?Quote:
Originally Posted by annihilat0r
Ofc it will work but not sure if 100% perfomance.
But I dont really know.
I thought the change from PCIe 1.0 to 2.0 only has to do with the power provided through the slot, and not really about bandwidth?