well if it's "Internal Performance Counter" then it could be linked to "in-house testing"
AMD RV770 refresh -> RV790
- Thread starter w0mbat
- Start date
well if it's "Internal Performance Counter" then it could be linked to "in-house testing"
I am revising down my guess for the number of wafers needed say absolute max 500(ie 100k chips) for launch and less per month afterwards. Quite possibly alot less. Don't think is going to have the volume that the RV770 enjoyed.
Just saw the power figures for 4890, is getting close to 200W peak. Also apparently close to 250W for the overclocked part(surely not going to be called the 4895???). Think this is too much for a card to sell into the mainstream, is really more an enthusiast thing. Compare to the 4850 which maxxed out round 150W.
Re a 4870 level part, yes there seems will be a gap there, but how to avoid it becoming a repeat of the current situation with the 4830? (ie have to disable 4850/4870 chips to sell as lower price 4830s to meet volume) What percent of RV770 reach 4870 levels? Surely would be too much waste 4850 and 4830 chips left over to continue producing RV770.
Re the picture skew - has anyone considered comparing the memory chips sizes on both boards to account for the skew? Should be same chip in both photos.
That's what I was wondering about, too.
You cannot quote sources, interpreting things into it, which aren't there.
So, IMHLO, either VR-Zone got it wrong (+6% seems way too little) or...?
How much close?
The GTX 260 216 has a TDP of 182W, so it's close to 200W, too. And I don't think that the upcoming GTX275 will have a lower consumption... But they're going to be mainstream too.
The HD4850 has a TDP of 110W, actually, and the HD4870 of 150W. How in the world the same chip could draw 50 or 100W more, only due to some minor changes?
No, it shouldn't, but it does apparently.
TDP, or rather something like "maximum power draw" for HD 4870 was 160 Watts. With PCBs now sporting provisions for 6- and 8-Pin-Connectors simultaneously (though apparently only for vendor overclocked versions), one should be prepared to see this figure rise quite a bit. After all, two 6-pin connectors would be sufficient for up to 225 watts theoretically.
But then, if a 30 percent perf increase is achieved, the card would definitely play in the ballpark of GTX 280 and above, so it should be allowed the same power requirements.
But then, if a 30 percent perf increase is achieved, the card would definitely play in the ballpark of GTX 280 and above, so it should be allowed the same power requirements.
Jawed
Legend
TDP is meaningless
Posts around here should be informative:
http://forum.beyond3d.com/showpost.php?p=1266102&postcount=275
These posts measuring/stating the power of HD4890 are meaningless. Bear in mind that we can't even get an agreement over the maximum power draw of HD4870
Jawed
Posts around here should be informative:
http://forum.beyond3d.com/showpost.php?p=1266102&postcount=275
These posts measuring/stating the power of HD4890 are meaningless. Bear in mind that we can't even get an agreement over the maximum power draw of HD4870
Jawed
How much close?
The GTX 260 216 has a TDP of 182W, so it's close to 200W, too. And I don't think that the upcoming GTX275 will have a lower consumption... But they're going to be mainstream too.
The HD4850 has a TDP of 110W, actually, and the HD4870 of 150W. How in the world the same chip could draw 50 or 100W more, only due to some minor changes?
Awhile back HT4U came up with numbers showing a higher consumption than listed TDP for a lot of cards(see a way back in this thread where it was talked about). Anyway on chiphell(see post #246) ia google translate:
1) 790XT core is indeed a 5-phase power supply 3 each 1195-chip-chip 40A 1165 Total 2 each 30A peak current of 180A can provide more than R600 core power.
2) 790XT fan power consumption 6W
3) 790XT core peak power than the whole card 770XT high-power
4) 790XTX the entire card from the 250-power very close
From the HT4U the 4870 can get up to 187W running vantage, if 3) above is true 200W shouldnt be much of an obstacle in the same circumstances for the 4890.
Those lower power 50nm samsung GDDR5 parts would really come in handy for this part. Nothing but qimonda on the leaked cards so far as far as i can tell though.
Actually in all the pictures of HD4890 that I've seen so far there are only two 6pin connectors...
Posts around here should be informative:
http://forum.beyond3d.com/showpost.php?p=1266102&postcount=275
These posts measuring/stating the power of HD4890 are meaningless. Bear in mind that we can't even get an agreement over the maximum power draw of HD4870
Jawed
Agree.
Here an HD4870 1GB shows a maximum power draw of 187W, while a GTX285 (55nm, remember) 214W.
So considering that the HD4890 is just a more powerful RV770, and the GTX275 is just a GTX285 with a smaller bus and somewhat lower clocks, they should draw more or less the same power.
http://ht4u.net/reviews/2009/power_consumption_graphics/index18.php
"provision" - i.e. the board layout is ready for 6- plus 8-pin connector.Actually in all the pictures of HD4890 that I've seen so far there are only two 6pin connectors...
"provision" - i.e. the board layout is ready for 6- plus 8-pin connector.
The fact that the board layout is ready for 6+8 pin doesn't mean that the card will actually have 6+8 pin connectors.
For this reason speculations about 250W of power draw in a hypotethical OC version (OC at which point? It is already almost sure that a stock RV790 will reach 950MHz in OC with the standard cooler, given the max figure in CCC) are at this stage completely pointless...
With the price wars at this point that they have reached now, i highly doubt the possibility, that any manufacturer would not avoid unnecessary costs (such as unused power routing).The fact that the board layout is ready for 6+8 pin doesn't mean that the card will actually have 6+8 pin connectors.
But you're right: The provision itself does not mean that the cards are bound to have an 8 pin connector.
edit:
So, please don't tell me I didn't say so: http://forum.beyond3d.com/showpost.php?p=1279413&postcount=930
Last edited by a moderator:
leoneazzurro
Regular
It is possible that the same PCB will be used for standard and overclocked RV790 parts, i.e. standard parts use 6+6 pin and OC parts use 6+8 pins.
IMHO;
no it's not. Because the Crysis Benchmark from [OC-Heaven] showed a weird result and so it seems the test was not valid. At least thats the conclusion from PCGameshardware about the test at oc heaven. Link: http://www.ocheaven.com/news_view.asp?gid=8362
no it's not. Because the Crysis Benchmark from [OC-Heaven] showed a weird result and so it seems the test was not valid. At least thats the conclusion from PCGameshardware about the test at oc heaven. Link: http://www.ocheaven.com/news_view.asp?gid=8362
Crysis benchmarks do not make sense.
They will have to be very overclocked, then... Maybe liquid cooled too...
Because usually oc parts sport only 50-100MHz more on the core, and that seems to be completely manageable by a simple HD4890 (judging from the safe oc figures of CCC).