Printable View
Well his personal review confirms for me that CF 4870s best the GTX 280 in AoC and that was my only remaining concern. Good to know as AoC is one of the more demanding games out currently and if the 4870s don't choke with high levels of AA in it, thats awesome.
Who is "they"? AMD? Did they really admit that? Since when do these large corporations ever admit to a flaw in their product? I had heard that both Nvidia and AMD are doing very well at the moment by not admitting to the existence of microstuttering or even the lack of multi-GPU driver support.Quote:
Originally Posted by AliG
They did indeed admit that they should have used the PCIE 2.0 bridge chip and not the PCIE 1.1 one, but they didn't say. But they did indeed admit using the wrong chip.
Uh, you'll see that has absolutely nothing to do with microstuttering though.Quote:
Originally Posted by zerazax
Note they never said anything about microstuttering, they just admitted they made a mistake with the type of bridge chip used. As to for what reasons, I don't know. I'm just re-affirming the fact they did indeed state it.
More like with the performance output.Quote:
Originally Posted by zerazax
and the cards can always be faster and better designed ;)
yup and you wont hear them admitting for a longtime. :pQuote:
Originally Posted by gojirasan
Especial when both aim for multi-gpu solutions in the future. :ROTF:
find it strange that Toms Hardware has no 4870 review yet
Strange?No,they are probably trying hard as hell to find a game+resolution+eye candy mode in which 9800GTX+ beats Radeon 4850 and 260 beats Radeon 4870 :rofl:Quote:
Originally Posted by natty
Give them time,i'm sure with their track reocrd they will provide us a "good material" :D :D
Strangely enough, their previous 4850 review showed it crushing so who knows, maybe the world really is ending
They did admit it, I don't have the link but I distinctly remember them admitting it quietly that they should've used a better chip. It was during either Computex or some other convention so it wasn't a public announcment.Quote:
Originally Posted by Hornet331
lovin the 4870!
http://i137.photobucket.com/albums/q...g/DSC_0285.jpg
And i'm loving your case! :up:Quote:
Originally Posted by lyl
That's a sick setup. I'm probably going to have the same loop going soon though for the GPU as well. You going to put a waterblock on that sucker?Quote:
Originally Posted by lyl
BTW i'm thinking about inverting my motherboard on my Stacker as well... I forgot but what were the advantages of inverting the motherboard? I remember BTX did it for a reason but the thoughts escape me
i wasn't referring to the PLX chip.Quote:
Originally Posted by zerazax
Yeah, i plan on putting it in the loop once theres a worthy block :P Looking for a full cover, but some of the new GPU-only blocks look nice.Quote:
Originally Posted by zerazax
First i gotta get some games for this puppy
http://i137.photobucket.com/albums/q...g/DSC_0280.jpg
I know its off topic but what were the cooling advantages to inverting the motherboard? I want to do it with my stacker once I get my triple rad loop put in but I forgot what it was good for besides the obvious fact that its on the other side for my door...
Source: http://www.webopedia.com/DidYouKnow/...rm_factors.aspQuote:
The BTX, or Balanced Technology Extended form factor, unlike its predecessors is not an evolution of a previous form factor but a total break away from the popular and dominating ATX form factor. BTX was developed to take advantage of technologies such as Serial ATA, USB 2.0, and PCI Express. Changes to the layout with the BTX form factor include better component placement for back panel I/O controllers and it is smaller than microATX systems. The BTX form factor provides the industry push to tower size systems with an increased number of system slots.
One of the most talked about features of the BTX form factor is that it uses in-line airflow. In the BTX form factor the memory slots and expansion slots have switched places, allowing the main components (processor, chipset, and graphics controller) to use the same airflow which reduces the number of fans needed in the system; thereby reducing noise. To assist in noise reduction BTX system level acoustics have been improved by a reduced air turbulence within the in-line airflow system.
This is very reminiscent of the R300 ATi Radeon 9500/9700/9800 hey days.
Even though I am very much an nVidiot I raise my glass to ATi....in fact I'm hoping to get a couple of 1GB HD4870 cards and crossfire them.
To those who are lucky enough to have their hands on one of these lovely cards, what are the dimensions?
How "thick are they?" (as in do they take more than 2 slots or just 2 slots like my G92 GTS)?
Due to the ill thought out layout of my motherboard (P5E3 vanilla) I will not be able to use any PCI-E devices if I get 2 Radoens, luckily my X-Fi can sit in the bottom PCI slot (where it is currently).
would be nice if the P5E3 was layed out like the other P5E boards with say a PCI-E slot for audio devices above the primary PCI-2.0x16 Graphics slot.
Anyway just thought I would say that ATi's marketing team are exceedingly clever. nVidia were boasting and parading around terms such as "Whoopass" and "wide gulf in performance over previous generation", yet ATi were "playing the price to performance game" and instead it appears they accidentally produced a high performing product.
Kudos to ATi, now is the time for Intel to acquire them...
John
Ah thanks, makes senseQuote:
Originally Posted by Unbornchild
They're 9.5" long and are just a normal 2 slotterQuote:
Originally Posted by JohnZS
guys here is what i found will add more if i find more
http://nwgat.wordpress.com/2008/06/2...-4850-roundup/
If someone has found a test comparing HD4870 in crossfire compared to a HD4850 in crossfire running on a AMD phenom then please write the link here. These setups seems to perform fairly similar on Intel computers for some games where I have seen that those games run better on AMD in high resolutions compared to Intel, but Intel run them better on low resolutions.
The reason why those cards perform similar could be that the FSB isn’t fast enough.
These cards are so fast so they need very fast I/O, intel may have problem to deliver that I/O performance. AMD is very strong at I/O because it has it's hypertransport.
Well at higher resolutions, the GPU is increasingly the limiting factor so in truth at higher resolutions, CPU becomes more irrelevant...
It depends, but I/O will also be something that can limit performance, especially when it comes to very fast cards.Quote:
Originally Posted by zerazax
Check this review: http://www.overclockersclub.com/reviews/intel_q9450/
AMD Phenom that is at lower clock and smaller cache pulls ahead for some games as the resolution goes up. That can only be explained by bad I/O performance on the Intel computers.
hmm 4870 single card looks tempting.
which drivers are people using? I checked ati stie, they dont list any for R700 cards yet
NOTE TO ANYONE READING THIS THREAD:
DO NOT BUY ANY 4870's at this time
over in the general discussion forum we are finding that the numbers of dead cards has reached the double digits (read: more than 10) and as an owner of 2 dead HIS 4870's at this time i can safely say they are not safe to buy
10 people out of how many? Tonnes of people have bought 4870s and there are bound to be a few lemons. Yes a lemon isn't the nicest thing to happen but until it hits 40% or so I would still buy one.
is adi more than 10 cards because i dont know how many its probably actually more than 15 and that is just on XS other communitys aregetting similar figures for dead cards, if i didnt know what i was talking about i wouldnt have posted itQuote:
Originally Posted by MomijiTMO
Overclocking kills them. BOTH CCC and BIOS overclocking.
Is there alot of HIS cards out there, or are the HIS cards the problematic ones?
i know of at least 8 dead HIS, 2 of which are mine, one of mine died at stock clocks
Gddr5 ???
Mine hasn't died on me yet. >.>
Are they dieing regardless of brand? I have a Gecube HD4870 on order.
Sounds like there was a bad batch.
It's notable, but I wouldn't worry too much. It seems HIS is being affected the most, which means it's a problem with their OEM.
better cooling
?
Everyone who's cards failed seemed to have flashed the bios of them. It looks like the actual flash programs are causing the issues somehow.Quote:
Originally Posted by SNiiPE_DoGG
They are ALWAYS safe to buy because if one dies but you didn't do anything to it (no cooler change, no voltmod, no bios mod, no overclock attemps other than CCC Overdrive, etc) you can always safely bring it back to the shop for immediate exchanging a new one or rma to the etailer.Quote:
Originally Posted by SNiiPE_DoGG
The number of members at this and similar fora that can't wait overclocking all new stuff they buy is in the four digits.
This is not the first time a HIS video card has failed in larger than supposed batches.
HIS IceQ3 (? or stock) 3870 also did that. Funny when it's BBA. Really, I think people should have stayed out after the last burnout.
According to user reports, that's not entirely the case. Some say their cards failed while unmoded. You may choose to believe their statements or not, but personally I will take these fellow XS members at their word until I have reason otherwise. Really they are still gathering more information about what the cause is. If it is defective parts, I am interested in seeing how it is handled by the companies involved. If it is user error, it would be helpful to know what sort of actions to avoid. But either way, more information is needed.Quote:
Originally Posted by Motiv
As for not buying the cards, I don't know about that either. What sort of failure rate are we talking about here? With more information about that trend I am certain informed users can make decisions for themselves about whether they want to take the obvious risk behind being an early adopter. Are we seeing an abnormally high failure rate, or was market demand simply so great that reports of potential issues came to light more rapidly?
I think it has to do with the GDDR5. I'm seeing loads of people with 4850's without any problems, and they oc'ed and vmoded the crap out of them (Example: gurusan). I didn't hear of a 4850 dyeing so far and the only difference i can think of between 4850 and 4870 is the GDDR5. Maybe that + excessive heat because of the crap stock fan speed is enough to randomly kill some. (altough users fixed that with the bios editor, got good temps and cards still died).
your joking? from what i have seen there mostly cards released in the US that are exhibiting this problem , and allot of them have been flashed , cards do "just" die , it happens , maybe the gddr5 is extrasensitive?Quote:
Originally Posted by SNiiPE_DoGG
anywayThese are the first ATi cards with 1Mb ROMs. As such, old tools might not work all that well for flashing the bios, in spite of their claiming RV770 support. So a wee bit of patience in your extreme overclocking endeavors might be very very healthy.
could be a problem?
Oh no man, sorry to hear you had 2 cards die on you...sounds like ATi have inherited AMD's problems :down:Quote:
Originally Posted by SNiiPE_DoGG
Thanks Zerazax looks like I can just about squeeze 2 1GB HD4870 puppies into my PC.
John
Does removal of the stock cooler void the warranty on these AMD cards? If there are any reliability issues at all I want to choose the company with the best warranty and customer service when I buy my HD4870x2.
For what its worth my card is a Sapphire and it has been running fine. :D
Here 4850's sell for $180 and 4870's for $270. Awesome price, ATI!
This thread has now taken a hilarious turn. There is a REASON BIOS flashing isn't covered by the warranty and we are seeing exactly why.
Show me where large batches of non flashed, non-overclocked cards are dying (taking into account that people who experience a problem are more likely to whine than those who don't) and maybe we will have something.
Until then it is just a matter of people wading into BIOS flashing without thinking of the ramifications.
hehehe true...but since we're in XS, nothing is left to stock LOL :ROTF:Quote:
Originally Posted by SKYMTL
I agree.To add to your post,i think most of people forget this is XS,so users are often taking their cards to the limits(BIOS flashing,volt modding etc.) so it's not uncommon we will see more problems at this forum than at any other out there.Most users are not OCing their systems,and with the latest "fix" for fan speed that Mascaras posted recently,Radeons are now idling 25 degrees lower than ususally(with 65% fan duty cycle).Quote:
Originally Posted by SKYMTL
Informal, it's actually healthy to criticize ATI now and then. Supporting them 100% of the time doesn't help your credibility :stick:Quote:
Originally Posted by informal
This is a QC (quality control) issue for the very reason that it is not an issue at Nvidia. GTX 280s were clocking to 700Mhz on air on stock cooling with sufficient fan speeds from day One. The dead cards are a small percentage and thus a small cause for concern, but the lack of availability plus the issues with the cards that AMD has yet to officially address on its own or any AIB for that matter, collectively means they are not giving a frack about the enthusiast users right now.
That may change in the future, but this is a serious issue. Thankfully, W1zzard has helped us isolate the problem and avoid it until a permanent fix arrives.
Perkam
Perkam ,i personally think this is AIB problem,not AMD's problem.But yes,the issue is there and needs to be addressed.How it will be done,i don't care,as long as it is fixed.The problems is the fan speed and it's rather low(yeah for the sake of quiet card,but it's still low).
I still can't see how this is a "serious issue" when the majority of people reporting problems are people ones who flashed their BIOSes.Quote:
Originally Posted by perkam
BIOS flashing is not an exact science and it is funny that you bring up the GTX 280 cards since no one was flashing the BIOS on those cards to get to that 700Mhz clock speed. I fail to see how parallels can be drawn between the two in the first place based on the methods people have used for overclocking in each instance.
On way or another, people have to realize that a "serious problem" to the overclocking community means squat in the marketplace as a whole. I have said it once and I will say it again: there is a reason overclocking and BIOS flashing isn't covered by the warranty.
:eek: How dare you criticize another Canuck, eh ?!! :eek: Thou hast broken thy code !!Quote:
Originally Posted by SKYMTL
Et tu SKYMTL :( , then fall Perkam :(
:p:
Some users are finding problems from CCC Ocing and from using them in Xfire at stock. Check the video card section :)
Perkam
i still think your betting your 300$ on a card thats got a high chance of dying within a few days. if you all knew anything about the issue you would know that quite a few cards on this forum have died without flashing any BIOS at all so i wouldn't be so quick to discount it as the sole reason. (this includes one of mine and 6 (4 his and 2 powercolor) others that i can specifically think of at this time tho there are more that i wont risk mentioning for fear of exageration,)
Jimmyz has said that other forums are reporting similar deaths.
so it comes down to this: do you want to buy a card that has a high chance of having a defect and even if it works your not gonna be able to OC it because the quimonda ram is holding you back? or do you want to wait, get a little price drop, and get a card you can work with? it seems pretty logical to me not to buy a 4870 until the problem has a fix or is fixed by AMD
Well we can't help it if you start giving invites to people whose card has failed :rofl: :rofl:Quote:
Originally Posted by SNiiPE_DoGG
In all seriousness, it's still a very small number and WAY to early to make comments like yours. In the worst case scenario, you could have a card die on you, be denied an RMA, be forced to buy another one, and still save money vs the GTX 280 :p:
Seriously, this stuff happens from time to time. For the owners, it is a serious issue, but as long as they are patient, they will get their due :)
I'm personally waiting for an official response from AIBs.
Perkam
well whatever, cant say i didnt warn em :rolleyes:
I am just playing Devil's Advocate here. :up:Quote:
Originally Posted by perkam
Another aspect to remember is that everyone out there is basically using beta (or what ATI calls "hotfix") drivers which can cause any number of things to go bump in the night. I have even had some issues with the CCC of the Series_5 betas failing to recognize the hardware after uninstalling the 8.6 drivers from my HD3870.
Personally, I think there is alot more going on here than buggy cards that go butt-up when having their BIOS flashed. Alot of it has to do with underlying driver issues IMO. This is why I never (EVER) use an overclock on a day-to-day basis with a hotfix driver from ATI. Not to mention that there is a REASON the CCC Overdrive only goes to a certain speed.
One way or another, if there are widespread reports of non-overclocked cards biting the dust I will be the first to jump on the lynching bandwagon. Until then though, I have my pitchfork, tourches and white bedcovering safely tucked away. ;)
i have a HIS 4580 which i have abused and its going strong still right now... flashed maybe 6-7 times trying different clocks after voltmodding. am sitting at 725/1050 which seems to be its sweet spot. i do have it on water tho... still if it does pop i wont be blaming ati... also my card was pretty weak on stock volts anyway, was only syable at 650 on the core so i needed to abuse it anyways..
4850's are not the issue, they are all fine, its the 4870's and the gddr5 that has problemsQuote:
Originally Posted by nunchucka
Quote:
Originally Posted by SNiiPE_DoGG
It's not. It's a few people trying to flash them and failed.
Stop spreading false rumour. :down:
OK guys it seems most of deaths were actually caused by truncated bios.
http://www.xtremesystems.org/forums/...=1#post3101668
I guess warning still stands. DO NOT FLASH THE CARDS.
sigh. this seems to go back and forth without result. basically nobody has a clue, yet.Quote:
Originally Posted by Unbornchild
agree, someone should setup a pollQuote:
Originally Posted by SKYMTL
there will allways be a bad batch, just look at RAM, might well be some chips that are fried, it happends
I have a pair of HD4850 and HD4870. When I run Vantage, 4850 CF doesn't show any artifacts and I get p10k+. However, I get a bunch of artifacts on 4870cf but I'm able to finish the benchmark and get p12k+. I'm installing games now. We'll see what happens.
your talking to a guys whos 4870 have faild without ocing or flashing. :p:Quote:
Originally Posted by Unbornchild
You believe that, read this again. ;)
everyone with a 800 shader paperweight aka broken no-boot 4870
http://www.techpowerup.com/articles/...ng/vidcard/152
fix for you! from teh w1zz :)
I see one problem with that...does everyone have a replacement video card to boot from?
Also which review is the best?
and has anyone found out what speed cpu unlocks the full power of the 4870?
Not always of course, but a PCI or PCI-Express cheapie card can be picked up from a local shop for $5-20 depending on your luck, a small hassle but worth it and cheaper/quicker than buying a whole new card of the same type :). Alternatively, if you have a friend with a computer (I think most of us do!) you could ask if you can borrow their videocard for a few minutes :D .Quote:
Originally Posted by Caveman787
Warn us of what...? That you can buy a video card and have it not work or frazzle within a week..?Quote:
Originally Posted by SNiiPE_DoGG
Sorry, thats been the norm for 10 years! Your not out anything but time and a possible few $ for shipping. Though I think we all thank you for the obvious heads-up!
.
http://www.xtremesystems.org/forums/...&postcount=529 :rolleyes:Quote:
Originally Posted by Unbornchild
It doesn't matter. If you changed the fan speed even, the original BIOS dump was half the size of what it should have been so it didn't matter if you OC'd or not - all you had to do was change the BIOS and back to the wrong file.
Either way, it seems like a lot of those that were bricked were fixed as the correct size (1Mbit) is now being used. Obviously the ones that truly weren't touched and died have to be investigated or RMA'd, but it looks like those who played with the BIOSes early no longer have to worry as the fix has been put up.
so is the stock bios is a problem bios or it is ok? I mean the stock bios is it the full size bios?Quote:
Originally Posted by zerazax
The stock BIOS is fine. What was happening supposedly was that GPU-Z and RBE were truncating the BIOS to 61kb files or so rather than the full 128kb which caused cards to die. Dave Baumann notified W1z about it and they found a fix with the help of some XS members and the correct BIOS is now on the site if your card died after a flash. A lot of members have revived their cards already..
now thats over
any pictures of that watercooled diamond 4870 oc
In the past all (consumer) graphics card BIOSes were smaller than 64 KBytes. With the introduction of GDDR5 additional space was required for the memory training code. Since ATI could not fit that piece of code into the 64 KB available, they increased the BIOS size to 128 KB (1 MBit).Quote:
Originally Posted by zerazax
For an unknown reason ATI's flashing software Winflash/Atiflash does not correctly detect those extra 64 KB and forgets to save them when you dump your BIOS to a file. The same happens to GPU-Z (fixed in 0.2.5 and newer). So everybody who extracted their HD 4870 BIOS ended up with no GDDR5 microcode.
from here, http://www.techpowerup.com/articles/...ng/vidcard/152
QftQuote:
Originally Posted by Unbornchild
not at all, if you use the new fixed bios from techpowerup you'll be fine, the issue was due to not all of the bios's cache being detected and half of it not being saved, but that's fixed now so its no big deal if you have the right bios flashing tool (that detects the whole bios) and right biosQuote:
Originally Posted by SimBy
My Sapphire 4870 user review.
with my old 9800gtx OC vs 4870 performance
http://forum.donanimhaber.com/m_24357058/tm.htm
Yeah true use the correct bios and ull be fine.Quote:
Originally Posted by AliG
Can someone clarify if my motherboard (P5W DH) will be a bottleneck for a single HD4870? If so how much of a performance loss would that mean?
why would it be a bottleneck? Right now we're looking at maybe 10x-12x Pci_e 1.1 is used, not even the entire x16, you won't lose any performance by not going to pci_e 2.0
If that is true why do P45's bottleneck them in crossfire compared to x48 or has that now been proven to be false?Quote:
Originally Posted by AliG
one card is fine, but crossfire uses more bandwidth and is best on a 16x/16x
Any ideas when the 1GB HD4870 is going to hit the streets?
Why is that then?Quote:
Originally Posted by bill_d
is it not because its different chipsets and mobos. i still think it is no bigge to crossfire with pci-e 1 compared to 2Quote:
Originally Posted by Mumid
From what I have seen, I would say around July 8th.Quote:
Originally Posted by JohnZS
What is the ETA for the official driver for the 4870?
Finally, Techpowerup: Diamond HD 4870 512 MB GDDR5
http://www.techpowerup.com/reviews/Diamond/HD_4870/
Thanks, just wanted to be sure.Quote:
Originally Posted by AliG
They don't, that was probably a driver issue.Quote:
Originally Posted by Mumid
This is correct - the problem was that RBE and GPU-Z were only pulling 64k of the 128k bios out of the card, flashing a truncated bios back in caused the cards to not post. No one knew this because all prior ATI cards (including the 4850) have a 64k bios.Quote:
Originally Posted by AliG
I recovered my card from a bad flash (it's rockin' by the way) and then used RBE to mod the 128k bios to include ramped fan control and flashed the newly modded bios back in - it works fine, just like you'd expect.
You can use me as the canary in the mineshaft for the next couple of weeks if you're feeling cautious.
Thats what i was hoping (as i have just bought a P45!) but have you got any evidence of this please?Quote:
Originally Posted by ToTTenTranz
Cheers
Then I'll ask you two things if you don't mind :p:Quote:
Originally Posted by halo112358
1 - Have you tried the 160MHz clocks at idle?
2 - When UVD is active, the card uses 750MHz? If so, I'd try to lower it to 160MHz too. The 3870 uses 700MHz when UVD is in use, but it works just fine with the standard 2D clocks of 300MHz, so I lowered them via BIOS.
I'd do this myself but I don't have my card yet :(
I haven't messed with bios clock changes, I'm still playing with the card through CCC. I have a feeling it's going to take water to really push the core (lotsa heat) so I'm just using it normally for now.Quote:
Originally Posted by STaRGaZeR
Here are the clock state profiles (UVD is using 4/5/6 for low/med/high)
http://img397.imageshack.us/img397/8...ocksym4.th.jpg
From this you can see that the card doesn't ever clock down below 500 on the core and the ram doesn't clock below 900 ever. I kinda wonder if they had issues with the ram during power state changes. I also heard rumors about one of the manufacturers having a fixed bios out that enabled better powerplay functionality, so I'll see if I can find it and try it out.
it's just powerplay not working yet read someware it will be fixed with driversQuote:
Originally Posted by halo112358
Looking at that pic the only thing you have to do to have proper PowerPlay is change clock 1 to 160MHz and 1.04v, and about UVD it only has on/off (using/not using), only two power states, so you can discard clock 5. Off is 2D clocks (500 in this BIOS), so I guess it will use 750MHz when decoding.Quote:
Originally Posted by halo112358
Does voltage adjustment in RBE even work? I was under the impression it did nothing?
Check it with a DMM, but in the 4850's it works up to 1,20v
What's the default voltage? 1.15v?
maybe 4870s can go to 2.3 bit idk my girlfriends dad is letting me borrow his 6decimal accurate multimeter so I guess I'll test my cards voltages out in a few days ;) :D