GIGABYTE AORUS PCIe 4.0 7300 SSD Announced with Up to 7,300 MB/s Read Speed

Report
Zarathustra said:
If that is the case, I would consider the 970 Pro a better desktop SSD. Low queue depth reigns supreme on the desktop. High queue depth is mostly enterprise stuff.

I'm surprised Anandtech did this review. They used to write some of the best SSD reviews before Anand sold the site and went to work at Apple but since then every time I've gone there they just have SSD press releases for new launches. Eventually I just stopped even looking at the site.
Click to expand...
Edit
Agree the 970 pro is techincally a better desktop SSD, though its difficult to tell the difference vs the 980 pro. I've got the 970 in my Daughter's PC (2700X) and the 980 in mine (5950) and I can't really tell a difference using the two different PCs. I'm hoping the 990 pro is a mlc model.

Re Anandtech:
I check the site every now and then because I'm part of the Anandtech Distributed Computing team. From my reading of the site over the past few years with no insider knowledge, they have been carrying on using the stuff created by Anand and Ian where they can, even though both have left. Where they no longer have a legacy tool set (such as GPU reviews), there isn't much content.
 
Last edited:
@Zarathustra

I'll try to do run through on my Sabrent 1TB and Crucial P5 this weekend and post those results as well. The Crucial is in the Gen4 slot while I've got the Sabrent in the Gen3.
 
Denpepe said:
My 980 PRO 1TB seems in line with yours

Read
6605.04 - 4190.85 - 980.01 - 86.62
Write
4940.73 - 4221.84 - 783.84 - 250.64

970 EVO 2TB

Read
3572.96 - 3057.86 - 1005.97 - 52.98
Write
3361.61 - 3209.20 - 852.25 - 280.59

960 QVO SATA 6Gb/s

Read
560.00 - 549385 - 385.83 - 38.37
Write
525.27 - 517.25 - 333.62 - 142.31

That's all I cba to test
Click to expand...

I wonder if we are using the same version of Crystal Disk Mark.

Different versions test different things unfortunately. What stands out is the third Read result. It is almost twice as high as mine from Crystal Disk Mark 8. In Version 8, that third result is RND4K Q32T1, in Version 7 it is RND4K Q32T16. This probably explains the drastically different numbers.

Because of this I can't combine these (or any other Version 7 Test results (this goes for @Riccochet as well) into my table.
 

Attachments

  • CrystalDiskMark_20230111201007.png
    CrystalDiskMark_20230111201007.png
    31.6 KB · Views: 4
  • CrystalDiskMark_20230111200439.png
    CrystalDiskMark_20230111200439.png
    32.1 KB · Views: 4
Denpepe said:
I reran 2 of the disks
Click to expand...

Wow. That is very strange. I wonder why I am seeing so much lower high queue depth 4K Random results than you are on the larger version of the same SSD. Your results are almost double mine.
 
Zarathustra said:
Wow. That is very strange. I wonder why I am seeing so much lower high queue depth 4K Random results than you are on the larger version of the same SSD. Your results are almost double mine.
Click to expand...
Bad joke but still had to say it: Did you check to see if DMA is turned on? How about making sure you're not using 40-pin cables instead of 80?
 
Zarathustra said:
Wow. That is very strange. I wonder why I am seeing so much lower high queue depth 4K Random results than you are on the larger version of the same SSD. Your results are almost double mine.
Click to expand...
no idea, maybe intel vs AMD?
 
Zarathustra said:
Wow. That is very strange. I wonder why I am seeing so much lower high queue depth 4K Random results than you are on the larger version of the same SSD. Your results are almost double mine.
Click to expand...
What slot are you having the drives in to test. It's possible the faster ones are using the native controller as opposed to the board controller for the M.2 NVME drive. (meaning direct to CPU or through a shared bus.)
 
Here are my results from the 2tb 980 Pro with heatsync I'm running. Don't forget it's also PCIE 4.0 x 4.

1673467188594.png

I wonder if the Pro is what the difference is.
 
I just upgraded my T-Force Cardea Z340 4TB to a Western Digital SN850X 4TB and the speed difference is insane. My games load so much faster now.
 
Denpepe said:
no idea, maybe intel vs AMD?
Click to expand...

Could be. Maybe I need to check for BIOS/AGESA updates. Ia haven't done that in a while...

Grimlakin said:
What slot are you having the drives in to test. It's possible the faster ones are using the native controller as opposed to the board controller for the M.2 NVME drive. (meaning direct to CPU or through a shared bus.)
Click to expand...

Hmm. My manual doesn't say anything about that. It is a Threadripper board though, so I have 64 total Gen4 lanes to deal with.

Presumably 4 are used by the chipset, just like with the regular Ryzens, which leaves 60 lanes for everything else.

I have four 16x PCIe slots.

Max they can run at is 16x/8x/16x/8x

The last slot will be in 4x mode if M2_2 is populated though (which it is)

So with all of the PCIe slots and one of the M_2 slots out of the way, I'm up to 48 lanes.

There are four more 4x m.2 slots though (for 5 in total), so that is 16 more lanes, which brings us to a grand total of 64.

So, yeah, at least one of them has to be shared with the chipset, but I am not sure which one...

More research:

iu


I am just confused about the PCIe lanes. They are saying 72 in total, but I could have sworn Threadripper always had 64...

Unless they are counting the ones that come off the chipset, but even so, those are Gen4, so how could they multiply?

Also, looks like the TRX40 chipset uses 8 lanes.

Either way, there is no indication in the manual which come straight from the CPU and which come off the chipset. For all of them to be viable, I'm guessing at least two need to come off the chipset. If I designed this board, I would make the two m_2 slots in the weirdo DIMM.2 slot be the chipset slots, but I am not sure if that is the case.

The 980 Pro that I tested above is in the primary slot though. One would expect the primary slot to be all CPU.

My quick googling raised more questions than it answered...



Grimlakin said:
Here are my results from the 2tb 980 Pro with heatsync I'm running. Don't forget it's also PCIE 4.0 x 4.

View attachment 2196

I wonder if the Pro is what the difference is.
Click to expand...

All three of ours are Pro's though.

I'm betting its either an AMD thing or an AGESA update I need.

That sounds the most plausible to me.
 
Last edited:
Zarathustra said:
I'm betting its either an AMD thing or an AGESA update I need.

That sounds the most plausible to me.
Click to expand...

Looks like I am one BIOS update behind. I'm on 1502, but 1603 is out. And 1603 does update the AGESA (to 5.15_1AVUB_PI 1.0.0.7_017) but what that means, who the hell knows.
 
According to hwinfo, my 980 pro (the 1 Tb drive) runs at NVMe 4x 16.0 GT/s and the 970 EVO at NVMe 4x 8.0 GT/s
 
You must log in or register to reply here.
Become a Patron!
Back
Top