Overclock.net › Forums › Graphics Cards › AMD/ATI › Hawaii Bios Editing ( 290 / 290X / 295X2 / 390 / 390X )
New Posts  All Forums:Forum Nav:

Hawaii Bios Editing ( 290 / 290X / 295X2 / 390 / 390X ) - Page 376

post #3751 of 3856
Here's a comparison...

Using 1,250mhz memory with 1,125mhz timings at 875mv with 1,205mhz core clock, it's only about 2 frames per second slower.

If I turn up the core clock, I might be able to beat the higher clocked score's

Clocking in at 1,225mhz core with 1,725mhz memory

https://www.3dmark.com/compare/spy/2782132/spy/2757125/spy/2757217
post #3752 of 3856
Hi @chris89

I see that you have the same H5GC4H24AJR memory of my 380 on your 390x. I have tested swaping timings on my 380, and I see that swaping the 1125 timings to the 1500 and the 1625 strap, I have better result than only swaping it to the 1500 strap, and I can go up to 1475mhz. I did the tests months ago so I don't remember well, but I think that I have better results with the 1125 timings and 1475 mhz than with the stock ones and ~ 1500 to 1550 mhz, and less memory errors.

I tested months ago, so I don't remember well how to do it, to mod the timings with the very good Vento041's tool and I did some changes to the 1125 timings: 77713320000000006BBD572F40550F0E2892F7060048C50033BB9D084C0D14205A8900A000000120100C20246F1E2912

I remember to get less memory errors with it. I have been using it some months, without problems. But really I don't know if it is ok because I don't know how to mod the timings well. Maybe somebody could help us with more info of how to mod the timings.

With the 1000 timings I couldn't go for more than ~1420mhz and with less score. I remember to test the OP timings without success.
post #3753 of 3856
Quote:
Originally Posted by mynm View Post

Hi @chris89

I see that you have the same H5GC4H24AJR memory of my 380 on your 390x. I have tested swaping timings on my 380, and I see that swaping the 1125 timings to the 1500 and the 1625 strap, I have better result than only swaping it to the 1500 strap, and I can go up to 1475mhz. I did the tests months ago so I don't remember well, but I think that I have better results with the 1125 timings and 1475 mhz than with the stock ones and ~ 1500 to 1550 mhz, and less memory errors.

I tested months ago, so I don't remember well how to do it, to mod the timings with the very good Vento041's tool and I did some changes to the 1125 timings: 77713320000000006BBD572F40550F0E2892F7060048C50033BB9D084C0D14205A8900A000000120100C20246F1E2912

I remember to get less memory errors with it. I have been using it some months, without problems. But really I don't know if it is ok because I don't know how to mod the timings well. Maybe somebody could help us with more info of how to mod the timings.

With the 1000 timings I couldn't go for more than ~1420mhz and with less score. I remember to test the OP timings without success.

I tested, I find even know the stock timings are 1500mhz, 1250mhz runs just as good on stock timings. I replaced them with tighter timings but the man point of running 1250mhz is undervolt vddci to 875mv from 1000mv & I get errors with 1125 strap on 1250 with 875mv, So I would need to use 1000mv for error free. However its best to run stock timings without errors.

Hows the r9 380 doing by the way?

I can help you mod your timings but it doesn't help anything. Stock timings is best for error free running.
post #3754 of 3856
Quote:
Originally Posted by chris89 View Post

I tested, I find even know the stock timings are 1500mhz, 1250mhz runs just as good on stock timings. I replaced them with tighter timings but the man point of running 1250mhz is undervolt vddci to 875mv from 1000mv & I get errors with 1125 strap on 1250 with 875mv, So I would need to use 1000mv for error free. However its best to run stock timings without errors.

Hows the r9 380 doing by the way?

I can help you mod your timings but it doesn't help anything. Stock timings is best for error free running.

I get a 10145 graphics score in firestrike with 17.10.3 drivers, 1125mhz core, 1.137v with LLS1, 1475mhz, 837mv vddci from stock 1.1mv vddci is ~ 5 or 10w less I don't remember. Is ~ 200 points more than with 1500mhz and 1500 timings, and more or less the same errors.
Warning: Spoiler! (Click to show)

In games I get more memory errors:
Warning: Spoiler! (Click to show)

I have pasted the 1125 modded timings ones to the 1500 and 1625 strap, not only to the 1500 strap. If not I get errors, artifacts and less score. Maybe modded 1250 timings could be better but I don't know how to mod them.

I'm using OverdriveNTool to OC: https://forums.guru3d.com/threads/overdriventool-tool-for-amd-gpus.416116/ , is great you can do profiles to to OC or 2D with lower core and memory clocks.

OCing with the bios is working worst for me, I get lower scores.
post #3755 of 3856
Quote:
Originally Posted by mynm View Post

I get a 10145 graphics score in firestrike with 17.10.3 drivers, 1125mhz core, 1.137v with LLS1, 1475mhz, 837mv vddci from stock 1.1mv vddci is ~ 5 or 10w less I don't remember. Is ~ 200 points more than with 1500mhz and 1500 timings, and more or less the same errors.
Warning: Spoiler! (Click to show)

In games I get more memory errors:
Warning: Spoiler! (Click to show)

I have pasted the 1125 modded timings ones to the 1500 and 1625 strap, not only to the 1500 strap. If not I get errors, artifacts and less score. Maybe modded 1250 timings could be better but I don't know how to mod them.

I'm using OverdriveNTool to OC: https://forums.guru3d.com/threads/overdriventool-tool-for-amd-gpus.416116/ , is great you can do profiles to to OC or 2D with lower core and memory clocks.

OCing with the bios is working worst for me, I get lower scores.

Here I found your 1500mhz timings & above it is 1625mhz... You may try replacing the string here replacing 1500mhz timings with 1625mhz timings to reduce errors.

You may try this.. idk looser timings will REDUCE errors.. so that's good.

I'm error free 1500mhz with these timings

F0 49 02

77 71 33 20 00 00 00 00 CE 51 6A 3D 90 55 11 12 30 96 49 09 00 4A E6 00 22 33 9D 08 74 01 14 20 6A 89 00 A0 02 00 01 20 15 0F 29 2F 94 27 31 16

your 1500mhz timings r9 285 here..

F0 49 02

77 71 33 20 00 00 00 00 31 5A 6B 3C A0 55 0F 13 36 19 1B 06 00 6A E4 00 22 AA 1C 08 7C 04 14 20 CA 89 80 A9 02 00 00 C0 17 12 26 2B 61 2B 37 15


Edited by chris89 - 11/23/17 at 8:07am
post #3756 of 3856
Here are some results from my earlier tests, when I was tuning my timings...

Firestrike 1.1 graphics scores | texture filtering set to 2x and quality to performance in driver settings | R9 290 @ 1200 / 1650 MHz (with Hynix AFR memory):

Default 1650 strap: 14 809
Default 1250 strap: 15 072
Custom timings: 15 388
390 bios mod with custom timings: 15 533

*Stilt timings: 15 039 (* could only be run @ 1500 MHz max without crash)
post #3757 of 3856
Quote:
Originally Posted by Tame View Post

Here are some results from my earlier tests, when I was tuning my timings...

Firestrike 1.1 graphics scores | texture filtering set to 2x and quality to performance in driver settings | R9 290 @ 1200 / 1650 MHz (with Hynix AFR memory):

Default 1650 strap: 14 809
Default 1250 strap: 15 072
Custom timings: 15 388
390 bios mod with custom timings: 15 533

*Stilt timings: 15 039 (* could only be run @ 1500 MHz max without crash)

Can you send me your Hynix 8GB custom timings or create some custom 1500mhz timings? thanks

1500mhz timings for my 390x 8gb : I have no errors with this timngs at 1500mhz but I get errors at 1563mhz... can you make timings that are error free at 1563mhz & new faster timings for 1500mhz?

Do you get errors?

F0 49 02

77 71 33 20 00 00 00 00 CE 51 6A 3D 90 55 11 12 30 96 49 09 00 4A E6 00 22 33 9D 08 74 01 14 20 6A 89 00 A0 02 00 01 20 15 0F 29 2F 94 27 31 16
post #3758 of 3856
Quote:
Originally Posted by chris89 View Post

Here I found your 1500mhz timings & above it is 1625mhz... You may try replacing the string here replacing 1500mhz timings with 1625mhz timings to reduce errors.

You may try this.. idk looser timings will REDUCE errors.. so that's good.

I'm error free 1500mhz with these timings

F0 49 02

77 71 33 20 00 00 00 00 CE 51 6A 3D 90 55 11 12 30 96 49 09 00 4A E6 00 22 33 9D 08 74 01 14 20 6A 89 00 A0 02 00 01 20 15 0F 29 2F 94 27 31 16

your 1500mhz timings r9 285 here..

F0 49 02

77 71 33 20 00 00 00 00 31 5A 6B 3C A0 55 0F 13 36 19 1B 06 00 6A E4 00 22 AA 1C 08 7C 04 14 20 CA 89 80 A9 02 00 00 C0 17 12 26 2B 61 2B 37 15
Warning: Spoiler! (Click to show)


I tested it in may, so if I remember correctly with 1125 modded timings to the 1500 strap and 1500 timings to the 1625 strap or the stock timings, I was getting lower scores than with 1125 modded timings to 1500 and 1625 straps. With the 1125 timings to 1425 and 1500 strap I was getting errors and artifacts, not with the 1125 modded timings to the 1500 strap and 1625 strap to stock timings like I said befores.

I see errors, but is been stable for months.

I have the same memory as you so I have the same timings for the 1500 strap, not the ones for the 285.
post #3759 of 3856
Quote:
Originally Posted by mynm View Post

[/SPOILER]

I tested it in may, so if I remember correctly with 1125 modded timings to the 1500 strap and 1500 timings to the 1625 strap or the stock timings, I was getting lower scores than with 1125 modded timings to 1500 and 1625 straps. With the 1125 timings to 1425 and 1500 strap I was getting errors and artifacts, not with the 1125 modded timings to the 1500 strap and 1625 strap to stock timings like I said befores.

I see errors, but is been stable for months.

I have the same memory as you so I have the same timings for the 1500 strap, not the ones for the 285.

That timings is your 380 bios you sent me, it has different timings. Unless your using a different rom?

Mynm_stock.zip 42k .zip file

Tighter timings means errors but maybe faster but errors actually reduces fps, by as much as 1-5 fps depending on how many errors. 222 could be possibly 1-2 fps.

By using my timings, it should be error free.

Looser timings means less or no errors at all.

Tight means running a lower clock timings on a higher clock & Loose means running a higher clock timings use on a lower actual frequency.

Tight = errors : less stable
Loose = no errors : more stable

By the way Its not possible to use 1125mhz timings on the 1500mhz strap, far too tight timings for that high a clock. I tried it, one 1 strap below possible for tight timings testing. However loose can hop like 7 straps below.

Only 1 strap below is possible... ie 1125 timings for 1250mhz or 1375mhz timings for 1500mhz...

However for loose timings, you can use 2,000mhz timings all the way down to 1400mhz, zero errors extra stable extra consistent... No errors means completing each read & write flawlessly.
post #3760 of 3856
Quote:
Originally Posted by chris89 View Post

Can you send me your Hynix 8GB custom timings or create some custom 1500mhz timings? thanks

1500mhz timings for my 390x 8gb : I have no errors with this timngs at 1500mhz but I get errors at 1563mhz... can you make timings that are error free at 1563mhz & new faster timings for 1500mhz?

Do you get errors?

F0 49 02

77 71 33 20 00 00 00 00 CE 51 6A 3D 90 55 11 12 30 96 49 09 00 4A E6 00 22 33 9D 08 74 01 14 20 6A 89 00 A0 02 00 01 20 15 0F 29 2F 94 27 31 16

Sorry my timings are only made for HYNIX AFR 4 GB, because that's what I have. I have already but up my TV1 bios that uses these earlier in this thread.
Anyway, here is the timings hex string I use for my custom timings: FFF13320000000006BC5582A5055090C200FC6040044820022BB1C005C0B1420008900A000000120120D20254C1F2613
Here are decoded timing values for R9 290 4GB (Hynix AFR), that compare 1625 - 1250 - my custom - 1000 MHz timings values:


I designed my custom timings for the best 3D performance, which means high memory clocks with as tight timings as possible. There are alot of memory errors, but no visual artifacts. I have also increased the memory "refresh factor" in the bios for a bit better performance.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD/ATI
Overclock.net › Forums › Graphics Cards › AMD/ATI › Hawaii Bios Editing ( 290 / 290X / 295X2 / 390 / 390X )