Not sure if this has been posted yet.
http://vr-zone.com/forums/418725/h [...] price.html
Theres quite a few people looking forwards to this cards release, so heres the "official" slides
I like that last slide.
But it looks great, might have to grab one if I ever get my 24" monitor.
1.5 a 9800 GT? is that like a 4870 standards?
Whats funny though is it does amazing in every game except grid lol....
I dunno how much of this I can actually believe.
Its a nice find, but 1.5 in company of heroes?
I'll wait until real benchmarks, but if this card really is what the benchmarks so, they pretty much killed off their 4850, and the 4870 lol.
The last slide is jokes though lol, though u have to see the 8800 GTs staying power. 2007-2009 champ of its level.

88-98gt+old news. Old arch, old process, old DX, oh and did I mention it was old? Much higher power usage etc etc. GRID may make sense as these cards may be somewhat BW starved
I thought the company of heroes bench looked a bit weird myself tbh.
Overall they are saying it's about 33% faster than an 8800gt but those 'benchmarks' shown are pretty scant on details. 50% more of 10fps is still only 15fps.
It has the same pixel fillrate as a 4870 and 4/5th's of the 4870's texture fillrate. Memory bandwith looks low but at 800mhz ddr5 you'd have to imagine that will overclock by a huge amount.
Could be interesting to quadfire these. I dont see Nvidia having an answer except for dropping the 250gts to the same price point, at which point ATI will just drop the 4770 lower. I reckon it will be within 5% of a 4850/9800gtx in performance, possibly even faster than a 4850.
Thats just it, is the BW good enough for a game like GRID, even tho it retains its high fill rate?
trust me, if these are true, add 30% to the 9800 GT power, the 4850 is worthless.
Given this table shows the 4850 only 10% on avg @ 1920x1200 faster than the 9800 GT, so this 30% (if true) would kick off the 4850 and the 4830, and send them to the discontinused bin, since these cards won't be able to have their price cut as much as the 128 40 nm 4770.
add 30% to the 4770, you have a card that can tie the 4870 1 gig and beat the 4870 512.
I doubt ATI is stupid enough to have a card that kicks that man cards out with 100$ price tag.
Not only that, but their series numbers will be screwed up, the 4830 is slower than the 4770? Same witht he 4850? and now the 4870 ?
These benches either don't make sense, or ATI lost it
.
I would much rather call BS benchmarks...

Just like they did when they came out with their slides on the 4850, alot of people doubted those as well. And, as I said then, Ill say it now. Take it with a grain of salt, as they are from ATI, but they were spot on the last time.
The 4830 is history, and theres still going to be seen a difference between the 4850 and this card, as Ive said, the BW isnt there, so smaller res for this card, a 5-10% less fps overall, and the new process. OEMs will eat this card alive, unlike the 4830, and the 4850 is for a higher res scenario
considering that ur link shows the tests done at 1920x1200 on avg with 1 @ 2560x1600 and 1 @ 1600x1050, I think my post stands.
If this card is accurate like the 4850 situation like u said, ATI just killed their 4850, 4830 and 4870 (512 megs atleast)

Also, that chart is set for fps at 100% for the 275. Thats not a direct comparison between the 4850 and the GT. Or 10% of 200fps is greater than 10% of 50 fps
So, you think a 4870 is only 25% faster than a GT, when we know its at least that much faster than a 4850
Then you have to look at the apples to apples of these cards are their respective fps, where say 30% isnt seen as much as 30% of a 275.
As a 40nm part you'd have to think it will overclock like an absolute beast, well ahead of a 4850 at any rate.
Tbh I don't think ATI are really bothered about killing off the 4850 anyway. It's already sold millions of units and now the 4870 is pretty much lower priced than the 4850 was on release. I mean, is anyone actually buying a 4850 now? To me it's overpriced and has been for a while.
I still think the 4770 is a bit of a strange card. First 40nm given a 128-bit bus, but it still doesn't quite make the 75w TDP that would really make it incredible. All I can really think of the reason for it is, ATI truly want to bury the ghost of the 8800 once and for all.
Even here it shows 15 instead of 10%, that still justifies my post.
Thats why these are done in %, so it wouldn't matter what card is 100%, only diff is thatnow the 4850 is infront, and are all using very old drivers.

Problem with that is the starting point, it could be read the 4850 is 16% faster than the GT, or the GT shows 20% less , depending on the starting point.
I understand your concerns, and its a possibility that these cars may be better, and if the rumors run true, ATI is currently running , or ordering alot of 40nm wafers thru TSMC. It could be theyre really going to give nVidia a price war, with this card leading the way, and everything else is to be cleared out, but thats speculation, but off of good rumors as well.
well, lets hope that if this card is stronger than the rest, they rename it accordingly. 4770 will be misleading.
4950 or something.

I still think itll perform close to those earlier leaked benches, and the naming will be apropo. The tranny number shows a change here. Maybe in certain games we will see huge gains, and others where they were already strong, not so much. I havnt seen the details yet, so who knows for sure. If anyone has em, please post. Die shot would be nice
| jennyh wrote : Could be interesting to quadfire these. |
I remember the article about the 4830 Crossfire config, I think that will be the same situation here except even MORE performance. They might as well make a 4770X2.
enough with the X2 cards lol....when did dual GPU become mainstream, what happenend to the whole, add another card once ur card just can't produce factor?

How about we get drivers handled for single cards first before we bother with X2's and GX2's.
| L1qu1d wrote : enough with the X2 cards lol....when did dual GPU become mainstream, what happenend to the whole, add another card once ur card just can't produce factor? |
Well dual-GPU became mainstream once Crossfire and SLI boards started selling for sub-$100.....that's just my guess though. lol
It is all about the price/performance ratio. Why get a single 4870 1GB, when you can have much more performance for just a few dollars more? Personally, I think it is all because of this economy, but, once again, that is just my guess.
| L1qu1d wrote : enough with the X2 cards lol....when did dual GPU become mainstream, what happenend to the whole, add another card once ur card just can't produce factor? |
i think you should replace your card if your card can't produce anymore... screw adding another "bad" card.
| jennyh wrote : As a 40nm part you'd have to think it will overclock like an absolute beast, well ahead of a 4850 at any rate.
|
i would still rather have crossfired two 4850's than the 4770's the benches have been out for awhile and it will slide in between the 4850 and 4830, either way for 200 bucks you can have an amazing gaming experience, and 4 would be great to see also with so many good motherboards around. ddr5 makes up for the bandwidth lack thierof.
HOWEVER
the real rabbit in the whole for ati is a 4770 x2 product at around or under the 200dollar range would just sell like crazy!!!!!!!
due to the low power consumption,
the price,
and for those people with one pci-e 2.0 slot!
and its up to date technology!
i know its cleche but i gotta use it, this could be the death knell to nvidia!
as far as "bad"cards the game doesnt knw what video card is running it really, i mean its a program that needs processing, which requires resources, an extra card of the same type can drastically change the peformance at a given level so. just because it cant handle your desired settings with one, you probably will with two, what do you mean by a "bad" card?
all video cards are gifts from god, like children to women, even if they are retarded you still love them if they arent breaking any records, they try to please, and well all parts have thier threshold
i rocked my 6600gt like a champion until last month, and i got a 4870 and its a whole other life out thier. goodbye old trucker RIP, ...............hello sweet sexy!
k well, let's say the 6600gt could SLI... (not sure if it can)... would you have added a 6600gt instead of getting a 4850?
As far as preformance, thanks to the higher core clock the 4770 really is closer to the 4850 rather than the 4830. As such it seems a little unfair to compare it to the 9800GT, so I think nVidia will drop the price of it pretty soon. Rather than drop the price of the GTS250, I'm sure that yields of the G92b are good enough that it may be best for NVIDIA to release a 9800GT with all the sps enabled and just call it a 9800GTS. Basically a 9800GTX/9800GTX+/GTS250 with lower clock speeds and a simpler PCB.
I've seen more than one person in this forum and online get excited about overclocking, since this card's built on 40nm tech. I don't get it, why? Does the smaller architecture allow for better OCing or something?
| r_manic wrote : I've seen more than one person in this forum and online get excited about overclocking, since this card's built on 40nm tech. I don't get it, why? Does the smaller architecture allow for better OCing or something? |
It produces less heat, AND it consumes less power. Maybe that is why? (I am seriously just throwing a guess out here)
Process shrinks usually consume less power, thereby letting off less heat, allowing for higher overclocks (usually). Companies also tend to throw in a few optimizations/revise a few things when they shrink a chip (Like how Penryn has SSE 4.1)
As we get to smaller processes tho, the gate and the channel become smaller also, which can cause some problems. Electro migration, which is simply, the electrons are no longer contained within where theyre designed to flow, but flow outside of those designed areas, cause eroding of the silicon itself, as it slowly burns it up thru friction. This leakage not only causes wear and tear, but more power is needed to do the same things, as alot of electrons are lost in this way. As you crank up the speeds, the electrons are flowing tighter, causing more electro migration, and thus heat and wear and tear. CPUs have more or less hit a wall, going as fast as they can, before they burn up.
Shortening the lengths from 1 transistor to the next, channel length, etc allows for faster speeds just due to distance alone. At these process sizes, it doesnt sound like much, but multiply workload availability by say 1Ghz per second, and those distances add up, using 2 differing processes.
As for comparing it to the 9800GT, its obvious looking at the slides, theyre talking about a pricing segment, and what the competition has in that segment. Its simply another bang for the buck scenario
| L1qu1d wrote : trust me, if these are true, add 30% to the 9800 GT power, the 4850 is worthless. http://i5.techpowerup.com/reviews/ [...] l_1920.gif Given this table shows the 4850 only 10% on avg @ 1920x1200 faster than the 9800 GT, so this 30% (if true) would kick off the 4850 and the 4830, and send them to the discontinused bin, since these cards won't be able to have their price cut as much as the 128 40 nm 4770. add 30% to the 4770, you have a card that can tie the 4870 1 gig and beat the 4870 512. I doubt ATI is stupid enough to have a card that kicks that man cards out with 100$ price tag. Not only that, but their series numbers will be screwed up, the 4830 is slower than the 4770? Same witht he 4850? and now the 4870 ? These benches either don't make sense, or ATI lost it I would much rather call BS benchmarks... |
You sir, seem to be in need of a bit of a math lesson. Not meant to be cruel or as an insult. I'm just trying to help.
These graphs make sense when you are comparing ALL the items listed against one median, in this case the GTX275. The entire chart is every other card's relation to that specific median, represented as a percentage of said median. It is important to understand that percent changes do not add in the usual way, if/when applied sequentially. The percentage difference between items in this chart cannot be directly applied to the percentage differences between two sub-items on this chart because their comparison would involve a completely different median. A percentage rate attained using one item on a list as a median, applied to percentage rate attained used against an entirely different median is NOT cumulative, as it seems you may have been initially suggesting.
Now, that said, and I hope clearly enough, basically, if you want an honest point of reference for the 4770, find some recent stock 8800/9800GT benchmarks that include the 4830, 4850, and 4870 that use the same test hardware (which was probably AMD/ATI, and no one seems to use an AMD/ATI testbed, anywhere, other than AMD/ATI themselves), the same drivers (there's a major point you completely overlooked), operating system, DirectX version (AMD probably used 10.1 for their card since they actually support it) and the same versions of the game in question. Once you've done all that, then we'll see where things end up.
Since the above won't happen, just wait and see and stop speculating, cause it's hurting my brain to think this hard so late at night. ![]()
P.S. This applies to all such % graphs.
Thanks for the explain guys!
Hmmmm... I think I should work the card into this proposed Q2 2009 2-year build. May 4 is the projected release after all.
6600gt never came sli that i know of, and i used that card way past its expiration date, so with my previous statement about adding another card i meant, most current up to date ie dx 10 or 10.1 cards show decent improvement in sli crossfire scaling to warrant the upgrade. the 4770 will be no different i suspect, and "should" scale the best yet.
k well.. i dunno, where do you draw the line for whether to add another card, or just to replace your existing card?
i'm planning to stay away from sli/crossfire until many/all games support it.
if somehow one could convince me that sli/crossfire is a good choice when building a computer from scratch...
There are 734 identified and unidentified users. To see the list of identified users, Click here.
You are about to answer a thread that has been inactive for more than 6 months.
If you still wish to proceed, please ensure that your posting is original and does not duplicate or overlap any prior responses to this thread.

