Gigabyte Z590 Aorus Tachyon *Memory Multi-Core GeekBench3* Low Score at Gear2

[Mahameru]

Hello everyone,

I don't know if other user also noticed this problem, but before I decide to write a thread here, I did some testing especially at Gear2 which I feel need more tuning from Gigabyte BIOS Team. Btw I have no issue at Gear1, Gear1 on Z590 Tachyon is flawless both on Single-Rank and Dual-Rank, even managed to get +100Mhz uncore compared to the others board I have. Long story short, I have 2 others boards here (Z590 OC Formula & Maximus XIII Apex). I noticed that OCF and Apex scores on par each other, for example 4800MHz C14-13-13 around 11600-ish Memory Multi-Core GB3, but not with Tachyon, Tachyon scores 11200-ish which is roughly about 400 points less than the others. I did some walkaround to the Tachyon's submission on the HWBot, and I can say that some OCers also face same issue with mine and I am aware that Memory Multi-Core pretty much doesn't bring huge impact to the overall GB3 score. Here I did some testing on OCF (sadly my XIII Apex already sold) and try to make comparison what is going wrong with Tachyon.

Note:

I use 2 different kits, F4-4800C17D-16GVK and F4-4000C15D-16GRTG because the GVK not working pretty well on Tachyon

I am not tied to any board vendor, all the stuff here comes from my own and I have no intention to say bad words to any board vendor

 

Z590 OC FORMULA (BIOS 1.27H)

*) 4800MHz 14-13-13 (Base on memory profile each board) 

GeekBench3

 

AIDA64

 

=============================================================================================================

 

Z590 AORUS TACHYON (BIOS X2G)

*) 4800MHz 14-13-13 (Base on memory profile each board) 

GeekBench3

 

AIDA64

 

As we can see, AIDA64 shows that Z590 Tachyon has much lower Copy-Bandwidth, roughly about 5000MB/s less compare to Z590 OC Formula. And in GeekBench3 Memory Multi-Core score breakdown, mostly OC Formula scores better than Tachyon. I did some initial testing also with AIDA64 at 5066MHz 15-15-15, and result is the same, 77K Copy-Bandwidth for OC Formula and 72K Copy-Bandwidth for Tachyon (I forgot to take a screenshot, maybe will test again later). From my personal assumption, I don't know maybe Gigabyte BIOS team choose to loosen some hidden timings for high frequency memory overclocking. Anyway, thanks for reading my thread.

 

 

[Splave]

RTL too tight maybe?

Some sub timings are different also.

I wouldnt use geekbench3 memory multicore without all the cores to be that accurate, as you can see the single thread is higher on gigabyte at those settings. 

Edited October 4, 2021 by Splave

[Mahameru]

20 hours ago, Splave said:

RTL too tight maybe?

Some sub timings are different also.

I wouldnt use geekbench3 memory multicore without all the cores to be that accurate, as you can see the single thread is higher on gigabyte at those settings. 

Hi Allen.. About RTL, the board set 37-38 it self, I didn't touch anything, but when I try to loosen it to 38-38 or 38-39 it wouldn't post.. As you wish @Splave, below are scores with all cores and HT enable with OCF's timings:

====================================================================================================

2C/2T scores with OCF's timings:

[Splave]

looks correct to me now doesnt it? 

[SparkysAdventure]

This is a Gigabyte-Rocket Lake specific problem to my knowledge. You are not the first I've heard report this.

Edited October 5, 2021 by Sparky's__Adventure

[Mahameru]

2 hours ago, Sparky's__Adventure said:

This is a Gigabyte-Rocket Lake specific problem to my knowledge. You are not the first I've heard report this.

Charles, we've talked about it before on FB messenger, doesn't matter who's the first to report about this, for sure Gigabyte has to take care of it correctly..

[SparkysAdventure]

necropost, but does this board allow you to manual set tighter subtimings?