- last edited on by
ROGBot
Purchased from Aliexpress, on visual inspection seems like was worth the wait coming from far east 😄 . This should be an improvement over the EK-Supremacy EVO Threadripper Edition.
RGB is not my cup of time but the store that had the best price sold the product with Bykski RGB/AURA . So loose in parcel was Bykski RGB unit, it has 6 connectors, 3 per long side of unit, each with 6 pins, labelled as fan 1 to 6. One short side has 2 connectors with 4 pins labelled as RGB. Other short side has DC 12V 2 pin to molex with pass through, 5 pin labelled as IBM, which has 3 wires to a LED in metal casing :confused: . No manual in box and no idea on usage, nor was there a plan to use it TBH 😛 .
Bykski A-Ryzen-ThV2-X has a nice weight and good finish IMO :cool:. This V2 edition also supports AM3/AM4 as well as Threadripper, read and weep EK 😛 . 3 sides are clear and one long length has an integrated LED unit, this has 4 pin connector and 50cm of cable. There seems to be an 4 pin extension cable included or maybe part of Bykski RGB unit :confused: . Cold plate has a fine machined finish, I may lap it at some point, doesn't seem too shabby at all. Looking forward to plumbing this in 😄 and comparing to EK :rolleyes: .
RGB is not my cup of time but the store that had the best price sold the product with Bykski RGB/AURA . So loose in parcel was Bykski RGB unit, it has 6 connectors, 3 per long side of unit, each with 6 pins, labelled as fan 1 to 6. One short side has 2 connectors with 4 pins labelled as RGB. Other short side has DC 12V 2 pin to molex with pass through, 5 pin labelled as IBM, which has 3 wires to a LED in metal casing :confused: . No manual in box and no idea on usage, nor was there a plan to use it TBH 😛 .
Bykski A-Ryzen-ThV2-X has a nice weight and good finish IMO :cool:. This V2 edition also supports AM3/AM4 as well as Threadripper, read and weep EK 😛 . 3 sides are clear and one long length has an integrated LED unit, this has 4 pin connector and 50cm of cable. There seems to be an 4 pin extension cable included or maybe part of Bykski RGB unit :confused: . Cold plate has a fine machined finish, I may lap it at some point, doesn't seem too shabby at all. Looking forward to plumbing this in 😄 and comparing to EK :rolleyes: .
I opened a ticket with EKWB support and reply is no cold plate update for std block. They seem to have discontinued the nickel/plexi block I have from their webstore, also other iterations are reduced. With the ticket I have been sent a €25 store coupon.
The profile I use on TR/ZE for 24/7 use is only RAM OC, as this profile had under gone testing when I had the EK block I had not rerun Y-Cruncher with Bykski. Now the temps were not differing here enough, what was differing was the average MHz of CPU 😄 .
This one of the 1st screenies I have with this RAM/profile whilst on EK block, when stress testing with Y-Cruncher.
Now above is using a differing version of Y-Cruncher and HWINFO, so here another EK one and 5.5hrs, using exact same versions of apps in use currently. OS has had updates paused when I went quad channel as I wanted a fair comparison with the dual channel kit that I had been using before (above/below are all quad channel/same kit/UEFI, etc, only block swapped, coolant reused).
We see ~3.5GHz average. On Bykski it's closer to 3.7GHz 😄 .
I thought "WHAT!?!", gotta be some kinda fluke, so rerun on another repost of rig.
I still think come on, this can't be right. Perhaps that day the sun/moon was aligned correctly in Threadripper land. So below is another rerun on following day 🙂 .
Now someones gonna notice and say "gup this can't be true the temps are higher average on Bykski", when I look at CPU Core Power (SVI2 TFN), the 3x Bykski runs use ~40W more core power. So the CPU was sustaining more MHz and using more power. The VRM temps are also higher on Bykski runs.
I have no explanation for this, other than Bykski on Y-Cruncher equals moar MHz :cool: .
The profile I use on TR/ZE for 24/7 use is only RAM OC, as this profile had under gone testing when I had the EK block I had not rerun Y-Cruncher with Bykski. Now the temps were not differing here enough, what was differing was the average MHz of CPU 😄 .
This one of the 1st screenies I have with this RAM/profile whilst on EK block, when stress testing with Y-Cruncher.
Now above is using a differing version of Y-Cruncher and HWINFO, so here another EK one and 5.5hrs, using exact same versions of apps in use currently. OS has had updates paused when I went quad channel as I wanted a fair comparison with the dual channel kit that I had been using before (above/below are all quad channel/same kit/UEFI, etc, only block swapped, coolant reused).
We see ~3.5GHz average. On Bykski it's closer to 3.7GHz 😄 .
I thought "WHAT!?!", gotta be some kinda fluke, so rerun on another repost of rig.
I still think come on, this can't be right. Perhaps that day the sun/moon was aligned correctly in Threadripper land. So below is another rerun on following day 🙂 .
Now someones gonna notice and say "gup this can't be true the temps are higher average on Bykski", when I look at CPU Core Power (SVI2 TFN), the 3x Bykski runs use ~40W more core power. So the CPU was sustaining more MHz and using more power. The VRM temps are also higher on Bykski runs.
I have no explanation for this, other than Bykski on Y-Cruncher equals moar MHz :cool: .
Mystery solved :D.
The UEFI has a bug 😄 . Well I believe a SMU quirk :eek: . In each screenie with Bykski Y-Cruncher posted before, the SMU power reading for CPU is never below ~180W. This should not be so when CPU had been at lower states prior to being loaded. Here is Bykski on same profile as EK and at ~3.5GHz as SMU bug is not present in this run.
I do not believe it is a setting that creates the quirk. I also believe it is not the app loading SMU that does this as I have some CB15 runs which on same profile have excessive scaling IMO.
For example usual for my 1950X at stock but RAM OC'd is ~3000 points in CB15 with no performance bias in UEFI/OS tweaks, I have several instances of ~3150 points. As usually when I bench I don't use monitoring I had not known why this was happening. I will check and I believe the same SMU quirk is occurring resulting in increased MHz for the higher scoring runs.
The UEFI has a bug 😄 . Well I believe a SMU quirk :eek: . In each screenie with Bykski Y-Cruncher posted before, the SMU power reading for CPU is never below ~180W. This should not be so when CPU had been at lower states prior to being loaded. Here is Bykski on same profile as EK and at ~3.5GHz as SMU bug is not present in this run.
I do not believe it is a setting that creates the quirk. I also believe it is not the app loading SMU that does this as I have some CB15 runs which on same profile have excessive scaling IMO.
For example usual for my 1950X at stock but RAM OC'd is ~3000 points in CB15 with no performance bias in UEFI/OS tweaks, I have several instances of ~3150 points. As usually when I bench I don't use monitoring I had not known why this was happening. I will check and I believe the same SMU quirk is occurring resulting in increased MHz for the higher scoring runs.
LOL 🙂 , indeed where is Elmor holidaying and who funded it 😛 .
Any how here is CB15, left correct, right quirked. Scores are lower than usual as the monitoring is set to lowest interval it allows, creating some overhead.
I now know what to look for.
I have tried the saved profiles I use, ie 3200Q and 3466SQ and I can not pin point a change within txt log or the settings changes log we see in UEFI upon saving that would highlight why this happen.
I have also loaded 3200Q then manually made as 3466SQ and then loaded 3466SQ saved profile and seen no change between them in logs.
Checking each profile I can not also say when it will trip to quirked SMU. I will keep this info under my hat for now. Wait to see if later UEFI exhibit issue, if so then report. Shame I just don't know what is tripping it.
Any how here is CB15, left correct, right quirked. Scores are lower than usual as the monitoring is set to lowest interval it allows, creating some overhead.
I now know what to look for.
I have tried the saved profiles I use, ie 3200Q and 3466SQ and I can not pin point a change within txt log or the settings changes log we see in UEFI upon saving that would highlight why this happen.
I have also loaded 3200Q then manually made as 3466SQ and then loaded 3466SQ saved profile and seen no change between them in logs.
Checking each profile I can not also say when it will trip to quirked SMU. I will keep this info under my hat for now. Wait to see if later UEFI exhibit issue, if so then report. Shame I just don't know what is tripping it.
All good chap IMO 🙂 . I reckon it's lacking some speedier development as it's not like mainstream like AM4. On say that though I could clearly see TR was much better UEFI at launch than AM4.
Anyhow back to topic 🙂 . Here is 7.5hrs Y-Crunching on Bykski, without the SMU quirk, on UEFI 0902, CPU stock, RAM 3466MHz The Stilt.
To me the Bykski fin area seemed to small, HardOCP at the time of the review hadn't measured area. So from the review images of other TR blocks we have ~32mm x ~53mm.
So made my own template 🙂 .
Few mm overhang on width, near the top/bottom the angled cuts/chunky fin not so good. All in all happy with this block, recently I assessed ~27% cheaper than EK TR block (without their recent coupon) and ~10% more performance plus feature AM3/AM4 support and RGB 😉 .
Anyhow back to topic 🙂 . Here is 7.5hrs Y-Crunching on Bykski, without the SMU quirk, on UEFI 0902, CPU stock, RAM 3466MHz The Stilt.
To me the Bykski fin area seemed to small, HardOCP at the time of the review hadn't measured area. So from the review images of other TR blocks we have ~32mm x ~53mm.
So made my own template 🙂 .
Few mm overhang on width, near the top/bottom the angled cuts/chunky fin not so good. All in all happy with this block, recently I assessed ~27% cheaper than EK TR block (without their recent coupon) and ~10% more performance plus feature AM3/AM4 support and RGB 😉 .
