-
Posts
123 -
Joined
-
Last visited
Content Type
Profiles
Forums
Events
Blogs
Posts posted by exaberries
-
-
Does anyone have the chipset drivers / MEI for xp? All the links I've found are broken or don't work.
-
Finally have 20 posts so I can join this.
Willing to cover international shipping to canada.
Thanks for doing this.
-
So even though I immediately killed the memory by being to happy to turn a pot heres how to volt mod this card and basically any reference ish 560ti since they use the same controllers.
Gpu core power is a ncp5388. Look for an empty pad right below the ncp5388 based on the orientation of the text. The one on the left is what you want wire this to a 100 ohm pot since the feeback resistance is 2 ohms. OCP removal should be as simple as shorting over the resistor connecting ILIM to ROSC. Load line can be controlled with VFB and VDRP. I think disconnecting the resistor between the pins should remove all load line. On the other hand reducing the resistance from VFB to DIFFOUT should reduce load line but shorting is probably a bad idea. the resistance from ROSC to ground is switching frequency. Don't drop this below 10k as that is already 800kHz-1MHz. I wouldn't go below 15k as that should hit around 500-700MHz.
Datasheet link: https://www.digikey.at/htmldatasheets/production/234279/0/0/1/ncp5388.html
Memory is a RT8101/A. Look for the tiny dot on the chip to orient. I soldered directly to the pin. Feedback measured at 600 ohms so I used a 1M pot but 200k is also probably fine. This controller is really simple so theres not really much more to add.
RT8101/A datasheet: https://datasheet.octopart.com/RT8101GSP-Richtek-datasheet-12527331.pdf
Cap modding:
This card is already really good on the capacitance front so I don't think it did anything. Can't say for sure since I killed the card before running any benchmarks checking to see if the mod worked.
Also delidding is a really good idea. Watch this to see. Don't crush your memory chips though.
- 1
-
58 minutes ago, NoMS said:
Oh nevermind, just figured the issue, this benchmark is super picky with RAM settings, it's working now!
y-cruncher scales nicely with memory settings but also requires the most stability of the hwbot benchmarks
-
guessing this is the same mod as in here:
what caps did you use for the input filter?
-
I've been messing with this controller for a while now and have some basic voltage control for it. Full range voltage control and other details should be done by the end of the summer. The monitoring should be able to be applied to a 8266 as well. Most of this is based on the xml files detailing the i2c/pmbus interface in the evc2 software + disassembling/decompiling the raspberry pi voltage control software for the epower classified. I also verified this working with a arduino through a level shifter on a 680 classified through the evbot header.
Since the epower classified has the controller at address 0x40 instead of 0x46 I modded the binary with a hex editor use the different address. Tested working with the asp1212 on a 780 dcu. Just use vset_mod linked. My test program for the raspberry pi is also linked. To compile it make sure you link with the i2c lib.
Heres a pastebin of the arduino code i've been messing with: https://pastebin.com/HwaiEBRU
Hopefully this should be enough info for anyone experienced with i2c to get basic voltage control. Once I iron out all the details i'll post a more complete overview with tools.
-
Here's the results so far with my 2 kits of 4000-16-16 1.4V ripjaws v:
first kit maxes out at around 1.9v struggles more with 3900-12-11-11 190 tRFC
second kit goes all the way to 2.0v does 3900-12-11-11 180/190 tRFC pretty easy
both kits together run 3250 10-10 pretty easily at 1.75-1.8v still working on 3250 10-9
seems to be still quite a bit of lottery for high scaling but good bin for lower voltages. I got each of these kits for 250 CAD on newegg.ca.
- 1
-
Default was 1.35 I think. I was trying to see how quickly the voltage rised when turning the pot to see what the non linear behavior was like.
-
So I finally got a 560 ti modded for both core and mem, cap mods, delided, and on an aio and before running any benchmarks I raised memory voltage above 1.7 volts (I don't know exactly what it hit) and had immediate shutoff followed by artifacting and inability to get into windows outside of safe mode. The Elpida datasheets suggests that the gddr5 should tolerate 2v. Its possible that I passed this point but i'm thinking that its more likely the memory controller. At least I already had ordered another 560ti (hopefully without elpida this time).
I'm hoping that somene with more experience with this can shed some light on safe memory voltages for fermi cards.
I'll post a full rundown of the mods soon.
-
41 minutes ago, TerraRaptor said:
Why not to make a thing that acts like AIDA64 (measures read/write/copy/latency) with 3dmark01 like formula for final score? I was running aida on very different setups (from mendocino to haswell-e, diffferent chipsets (via/sis/intel/nvidia/amd) and it was quite informative. Problem about aida64 is that diffferent versions have different optimisations - still, you can now make a version that will cover all existing memory types from EDO to DDR4 (I think we won't see anything newer than alder lake with ddr4?). So that will be one "basic" part - which won't be updated with newer algorithms (just bugfixes). Call it "legacy mode". And all newer hardware will have different mode that you will be able to update - and changes in that part can be then covered with hwbot rules.
From what I know you are already basically describing the geekbench memory test. I think it would be better to have two seperate benchmarks rather than a legacy mode. Updating is both kinda necessary to deal with changes in architecure but also a massive problem for rules/scoring. My preference would be having algorithms change with major versions and each major version being a seperate benchmark à la geekbench.
-
-
-
-
If anyones is looking for boardviews for other stuff:
asus: http://schematic-x.blogspot.com/2018/04/asus-pack-198-files.html
also has asrock including oc formulas: http://schematic-x.blogspot.com/2019/09/asrock-schematic.html
gigabyte: http://schematic-x.blogspot.com/2018/04/gigabyte-schematics-pack.html
and gpus: http://schematic-x.blogspot.com/2018/04/blog-post.html
- 2
- 1
-
I've never personally had problems just changing without resetting with msconfig on windows 7. I think its the removememory/truncate memory that you have to reset before use but i've only ever tried that once on windows 10.
-
To add to what leeghoofd said with 3d stacking new cpus might start to have enough cache for entire benchmarks to run in. Though autodetecting the amount of cache and changing the benchmark is a good idea you have to be careful, with new releases it might just start scoring better diminishing the effect actual ram oc. I generally think that having a benchmark which significantly changes the workload it runs based on what hardware its running on can lead to wierd results for global ranking not really a problem for hardware ranking. IMO ideally that platform that the benchmark is run on should have little effect on the score but I know that's not really possible due to the effects of architecture like infinity fabric, etc...
I don't think a ram focused benchmark that uses a lot of ram (20GB+) / requires a lot of stability would be interesting since it would limit pushing things the edge. Maybe something that uses about 1GB so maxmem is still a thing but cache hopefully matters less.
-
(wanted maxmem in GB + 0.8) * 1024 is what I put into msconfig for windows 7 usually as a quick and dirty way to get close. I'm sure someone can give a more accurate way to calculate what the exact value you should enter
-
10 hours ago, Leeghoofd said:
Voltages shouldn't be a problem. Multies we use the trick to push one higher
Core voltage is where I have the most problems just stop setting at some point randomly, oc panel also can't change vcore anymore. I have more problems when a benchmark such as superpi is open. CPU ratio stops working on the oc panel but continues to work in turbov core sometimes. BCLK continues to work in both. Same issues even when oc panel is not plugged in.
I also have been running into a wierd bug where trying to assign per core oc when core are disabled in bios instantly locks the system.
-
-
Does anyone have the turbo vcore version/MEI driver. I've been using turbo vcore 1.10.07, 1.01.15 and 1.00.3 with MEI 10001204 and it takes about 3-4 tries for voltages / clock speeds to actually set.
-
Here's a bunch of info on rebranded asus controllers. I've been collecting info for a while from forums, techpowerup, manufacturers listed, i2c, etc... I can't guarantee that a lot of this is correct, but hopefully it's a good starting point. Question marks indicate a bigger guess.
asp0907
likely: up6218
gpus: 580 matrix core?; 580 dcu core; 570 dcu core
asp1000(c)
likely: chil 8326/8
motherboards: 970 pro gaming aura core / nb; r4e core; m4g core
asp1101
likely: ir 3541? / chil 8325A/B?
motherboards: r4e vccsa, vdimm
asp1103
motherboards: m9a dimm
asp1106
likely: rt 8894A
motherboards: tuf x570 pro core; prime x570 pro core; strix b550-f core
gpus: rx 470 strix core
asp1203
asp1209
likely: upi thing
asp1211
likely: chil 8228g
gpus: 7970 dcu core; 270x dcu core; 280x dcu core
asp1212
is: chil8318
780 (ti) dcu core; 780ti matrix; 770 dcu core; 660 dcu core; 670 dcu core; 680 dcu core; 650ti core?
datasheet for asp part: https://www.infineon.com/dgdl/pb-asp1212.pdf?fileId=5546d462533600a40153567f4b2128a4
datasheet for chil 8318: https://www.infineon.com/dgdl/pb-chl8318.pdf?fileId=5546d462533600a40153567f951a28bd
asp1213
likely: upi thing
asp1215
likely: upi thing?
gpus: 270 dcu core
asp1251
likely: ir thing
asp1252
likely: onsemi thing
asp1253
asp1257
likely: ir 3583
motherboards: r5e core
asp1300
likely: ir 3567b
gpus: rx 480 strix core
ir 3567b datasheet: https://www.infineon.com/dgdl/pb-ir3567b.pdf?fileId=5546d462533600a4015356803a7228ef
asp1400a
likely: onsemi 6 phase thing
motherboards: m11h core
asp1401b
likely: onsemi thing
asp1405(i)
likely: ir 35201
motherboards: m12e core + 2nd for vccsa; m12h core + 2nd for vccsa; prime x299 ed 30 core; c8i core; c8h core; z2e core; c6h core; m9a core
ir25201 datasheet: https://forum-ru.msi.com/index.php?action=dlattach;topic=104392.0;attach=38179
asp1500u
likely: upi thing
gpus: 980 ti strix core; 980 ti matrix core
asp1900b
motherboards: tuf z590 plus core
changelog:
21-07-26: add m9a voltage controllers
21-08-04: add full ir35201 datasheet link
21-08-05: add 770 dcu and 780 ti matrix to asp1212 list
21-08-09: add 570 dcu to asp0907 list
21-09-20: asp1211 is probably chil8228g
-
finally got my first powercard working, cut from a 560 ti with a dead core/memory
memory vrm is still shorted to ground but the controller is reciving power so there might not be any more work to get it working if the short didn't exist
the vrm doesn't require 3.3v and runs off of 12v for power
+12v for controller is from pcie so you have to connect one of the 6 pins to the power plane on the front (all the way at the bottom)
+5v is needed to enable the controller
simple feeback mod for voltage control and an led for +12v power indication
unfortunately there is little ground but since vmem is shorted to ground all of vmem works as a ground plane
[FS] sold out
in Sold
Posted
pm for gemini