• 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.

ATI check

K

Kendrak

[H]ard|DCer of the Year 2009
Joined
Aug 29, 2001
Messages
21,141
What are people currently getting on their 4850/4870 cards.

 
hmm yes i was wondering the same thing because i was looking at putting together a new comp and the ati x2 cards are alot less then the nvidia solution.
 
On a 4850 stock clocks

4730-2906ppd
4742-4081ppd
4743/4743-4046ppd
4752/4755-3492ppd
5732/5734/5736/5738/5739 etc...- 2373ppd

I just sold my 4850 i'm all nvidia now. The 4850 was slower, ran hotter with more noise, consumed more power, consumed more cpu utilization, and crashed my computer while folding.

Not a good card to fold on stick with nvidia.
 
4870 @ 850 core

Vista x64:
~2800 on the 511 pointers
~3300 on the 384 pointers
~4400 on the 477 pointers
~4900 on the 548 pointers

WinXP either 32 or 64bit:
~3200 on the 511 pointers
~3800 on the 384 pointers
~4900 on the 477 pointers
~5400 on the 548 pointers

Vista x64 takes a big hit, not sure about 32 bit. Ever since core 1.22 came out it also uses 3-8% extra CPU time in Vista under the System process but uses 17-25% of a quad core. In XP it just uses 23-25%



 
There should be some new posted in a day or so in the ATI section of the FF.

Until then, I can't say much due to the beta testing rules.

I will say that things are looking good and those numbers posted won't be correct for very much longer.
 
EvilAlchemist said:
There should be some new posted in a day or so in the ATI section of the FF.

Until then, I can't say much due to the beta testing rules.

I will say that things are looking good and those numbers posted won't be correct for very much longer.
Click to expand...
will cpu utilization be going down too?
 
[H]ugh_Freak;1033756931 said:
Vista x64 takes a big hit, not sure about 32 bit. Ever since core 1.22 came out it also uses 3-8% extra CPU time in Vista under the System process but uses 17-25% of a quad core. In XP it just uses 23-25%
Click to expand...

Yes, That is the DirectX bug.

For some reason, the ATI client is using a DirectX subsystem resource.

The reason for this is really not clear yet, but it is a documented issue.

Hopefully a solution will be located here in the near future.
 
Kind of weird ATi is the default hardware tested on but they are lagging on taking full advantage of the GPU client software?

Hope they can produce. Tons of cards are sold for DC use only. ;)
 
Kowan said:
Kind of weird ATi is the default hardware tested on but they are lagging on taking full advantage of the GPU client software?
Click to expand...

See, that is not really the full story .....

The cards are being used 100% ... the PPD difference is due to benchmark process ..

Every time the nVidia side is given a large work unit, they bitch and moan saying "MY PPD" .. no fair and so they pull the WU and give them to ATI folders since the cards can handle them very well ...

Only the top end nVidia cards can handle the larger WU, but since they are in the minority, it just compounds the issue
 
My 280s and 285 avg 6300-6400 PPD on the 511 wu Vista 64bit.
If AMD was equal or better my second box would be running Red instead of GREEN. ;)

All it does is fold 24/7. I'll gladly build a third box once the power shifts. :)
 
I get around 4k with 477-pointers (have one right now), 2800 with 511-pointers, 4800-4900 with 548-pointers, and around 3000 with 348-pointers. I've been getting a ton of 511-pointers lately though.
 
Zero82z said:
I get around 4k with 477-pointers (have one right now), 2800 with 511-pointers, 4800-4900 with 548-pointers, and around 3000 with 348-pointers. I've been getting a ton of 511-pointers lately though.
Click to expand...

on the nvidia side we've been getting a ton of 511 pointers too. I don't mind ATI it's just that the slow performance and the high cpu utilization left me with a bad taste. If the performance and support improves i'll build another box with ati instead of nvidia and have a box of each.
 
EvilAlchemist said:
He is prob running 1 client on each gpu on that card ....
Click to expand...
I know, but 12k is still high for two 4870 GPUs. I top out at around 5k with my single client, so the fact that he's getting 12k means he must be averaging 6k per client for those particular units, which is much higher than I've ever seen.
 
Zero82z said:
I know, but 12k is still high for two 4870 GPUs. I top out at around 5k with my single client, so the fact that he's getting 12k means he must be averaging 6k per client for those particular units, which is much higher than I've ever seen.
Click to expand...

Your are using Vista .. which incures a penalty due to the DirectX bug. At least I think thats why mine is about 500-700PPD lower in Vista then in XP.
 
Every time the nVidia side is given a large work unit, they bitch and moan saying "MY PPD" .. no fair and so they pull the WU and give them to ATI folders since the cards can handle them very well ...
Click to expand...

So is ATI actually running better science but the point values are deflated compared to nvidia?
 
flogge said:
So is ATI actually running better science but the point values are deflated compared to nvidia?
Click to expand...

Reverse part the statement.
nVidia points values are inflated compared to ATI due to current benchmark setup / cores used to provide those numbers
 
EvilAlchemist said:
Reverse part the statement.
nVidia points values are inflated compared to ATI due to current benchmark setup / cores used to provide those numbers
Click to expand...


aka nvidia's paying them off to make them look better then ATI... lol jk :p
 
EvilAlchemist said:
He is prob running 1 client on each gpu on that card ....
Click to expand...

Yep, one on each core, and with each core clocked to 800Mhz. It depends on the units, sometimes I only get 2000PPD on each core...but now and again I get 6000PPD or higher :). Running vista.

 
Actually, the numbers got inflated because we were very much CPU limited on the smaller proteins, especially on 48XX, and they were not. We have a core that drops CPU load, but we have run into issues with triggering VPU recovers on XP, which we believe will be fixed in the next driver, and then this updated core can go out. For large proteins, the performance gains will be small, but on small proteins, performance should increase more noticeably. We still do proportionally better the larger the proteins get compared to Nvidia.

We are working to get this core out soon with a driver check to enable/disable code paths, but when that code is change, we need to go through another QA round. We do our best to avoid causing headaches and system instabilities, which is why we do so much testing before things go out. Unfortunately, we still have issues like driver changes that cause VPU recovers on some XP configs starting at Cat 9.1 and hopefully fixed in Cat 9.3.
 
You must log in or register to reply here.
Back
Top