Review: Geil Polaris RGB DDR5 8000 CL38 2x16 GB

#6131386

Krizby

2023-05-12 12:36

Rocket Lake Refreshed with 8000MT DDR5 is gonna be interesting...

#6131413

386SX

2023-05-12 13:39

"Geil" may be a perfect German description of their kit. 😀

#6131421

wavetrex

2023-05-12 14:06

I am stunned at the pace with which DDR5 is gaining faster and faster models, much faster than DDR4. It took over 8 years for DDR4 to advance from basic 2133 kits to the fastest and most rare 4600+ DDR5 is already hitting the number "10000" at the top-end, with acceptable priced kits easily at 7200+ (which is equivalent in real clocks to DDR4 3600) I'm pretty sure the best binned overclocker memory will reach 12000 or higher before DDR5 is replaced by something else.

#6131425

barbacot

2023-05-12 14:19

wavetrex:

I am stunned at the pace with which DDR5 is gaining faster and faster models, much faster than DDR4. It took over 8 years for DDR4 to advance from basic 2133 kits to the fastest and most rare 4600+ DDR5 is already hitting the number "10000" at the top-end, with acceptable priced kits easily at 7200+ (which is equivalent in real clocks to DDR4 3600) I'm pretty sure the best binned overclocker memory will reach 12000 or higher before DDR5 is replaced by something else.

Great speed but these DDR5 latencies are still to high, maybe in a year or two this will also be fixed.

#6131429

wavetrex

2023-05-12 14:28

barbacot:

Great speed but these DDR5 latencies are still too high, maybe in a year or two this will also be fixed.

Actually, no. CL38 at 8000 means same latency as CL19 for DDR4-4000, which is almost exactly to what high-end modules have (usually CL18 or 19 at that speed, with some very expensive ones doing CL16)

#6131521

bobnewels

2023-05-12 19:19

Thanks for the review,I am surprised there is an actual difference in FPS with even 1440p,I can see it at 1080p. I am actually so surprised I ordered a Ram kit CL32 7000mt after I read the review a couple hours ago and the RAM will be at my house in a few hours. I have to check now for myself if there is actually a FPS difference at 1440p/2160p for the hell of it.

#6131525

XenthorX

2023-05-12 19:24

Really like when you include those memory timing formula. Maybe one day I'll remember it 😱. Must have read what all those timings are over 5 times and always end up forgetting.

#6131526

Bender82

2023-05-12 19:29

wauw ... from 6000 mhz to 8000 mhz 8 fps 😛

#6131573

schmidtbag

2023-05-12 22:21

Bender82:

wauw ... from 6000 mhz to 8000 mhz 8 fps 😛

Honestly, considering modern games aren't really CPU bottlenecked, 8FPS is more than I'd expect.

#6131639

bobnewels

2023-05-13 02:03

So just to follow up. Got my DDR5 CL32 7000Mt Ram. Tested at 1440p/2160p and get the same performance as DDR5 CL36 5600/6000/6400. Guess I got to try to overclock to DDR5 CL38 8000. Probably not a chance to get that high on 13700KF and cheapest Z690 Motherboard with 4 Dimms slots but I suspect nothing will change are far as FPS. So out of the box performance no change from 5600Mt-7000Mt for PC Gaming FPS.After overclocking Ram if no change ,I may grab 13900K for fun and do up another AMD system.

#6131673

Carfax

2023-05-13 10:04

bobnewels:

So just to follow up. Got my DDR5 CL32 7000Mt Ram. Tested at 1440p/2160p and get the same performance as DDR5 CL36 5600/6000/6400. Guess I got to try to overclock to DDR5 CL38 8000. Probably not a chance to get that high on 13700KF and cheapest Z690 Motherboard with 4 Dimms slots but I suspect nothing will change are far as FPS. So out of the box performance no change from 5600Mt-7000Mt for PC Gaming FPS.After overclocking Ram if no change ,I may grab 13900K for fun and do up another AMD system.

Memory speed will only impact performance when a game is CPU bottlenecked. So at 1440p and 2160p, you likely won't see any improvement in averages. You might see an improvement in 1% and 0.1% lows however and the frametime graph should be smoother with less variance. Hitting DDR5 8000 while fully stable is a pain in the ass. I can run at DDR5 8000, but not stable unless I pump a ton of voltage into the memory controller and VDIMM voltages and use relaxed timings. It's not worth it though. Manually tuning the sub timings at lower frequencies will typically yield better performance than higher frequencies with XMP timings. Case in point, my memory scores are comparable to what was achieved in this review despite running at DDR5 7600. The DDR5 8000 gets higher write and copy scores due to the Apex motherboard with 2 DIMM slots I believe. My motherboard is an Asus Z790-E. https://www.guru3d.com/index.php?ct=articles&action=file&id=87191&admin=a6c913ce369e94812aca8a11cd929a99f84ab8ed https://imagizer.imageshack.com/v2/545x525q90/923/LMen17.png

#6131703

alanm

2023-05-13 12:58

Carfax:

Memory speed will only impact performance when a game is CPU bottlenecked. So at 1440p and 2160p, you likely won't see any improvement in averages. You might see an improvement in 1% and 0.1% lows however and the frametime graph should be smoother with less variance. Hitting DDR5 8000 while fully stable is a pain in the ass. I can run at DDR5 8000, but not stable unless I pump a ton of voltage into the memory controller and VDIMM voltages and use relaxed timings. It's not worth it though. Manually tuning the sub timings at lower frequencies will typically yield better performance than higher frequencies with XMP timings. Case in point, my memory scores are comparable to what was achieved in this review despite running at DDR5 7600. The DDR5 8000 gets higher write and copy scores due to the Apex motherboard with 2 DIMM slots I believe. My motherboard is an Asus Z790-E. https://www.guru3d.com/index.php?ct=articles&action=file&id=87191&admin=a6c913ce369e94812aca8a11cd929a99f84ab8ed https://imagizer.imageshack.com/v2/545x525q90/923/LMen17.png

DDR5 + Raptor Lake dynamite combo. And agree re sub-timings, they are more important than the primaries with Intel/DDR5.

#6131767

Mda400

2023-05-13 16:54

Carfax:

The DDR5 8000 gets higher write and copy scores due to the Apex motherboard with 2 DIMM slots I believe. My motherboard is an Asus Z790-E. https://www.guru3d.com/index.php?ct=articles&action=file&id=87191&admin=a6c913ce369e94812aca8a11cd929a99f84ab8ed https://imagizer.imageshack.com/v2/545x525q90/923/LMen17.png

You should be able to get read/write/copy bandwidth closer to each other if you lower the write-to-read/read-to-write tertiary timings. The 8000 mhz kit probably has it low enough in it's XMP profile already.

#6131907

Carfax

2023-05-14 06:49

Mda400:

You should be able to get read/write/copy bandwidth closer to each other if you lower the write-to-read/read-to-write tertiary timings.

You wouldn't by chance happen to know which timings those are would you, or maybe a guide that could help me? The only secondary timings I've modified so far are: 1) tREFI (this one had a big impact) 2) tRFC (minimal impact) 3) tWL ( improved write performance)

#6132112

Mda400

2023-05-15 04:15

Carfax:

You wouldn't by chance happen to know which timings those are would you, or maybe a guide that could help me? The only secondary timings I've modified so far are: 1) tREFI (this one had a big impact) 2) tRFC (minimal impact) 3) tWL ( improved write performance)

tWRRD_sg tWRRD_dg tRDWR_sg and tRDWR_dg are more sensitive to small changes and may not go below 17-18 at best.

#6132127

Carfax

2023-05-15 07:54

Mda400:

tWRRD_sg tWRRD_dg tRDWR_sg and tRDWR_dg are more sensitive to small changes and may not go below 17-18 at best.

Major props to you man, you were absolutely right! It's amazing what a few timings will do for memory performance! 😱 I just plugged in some arbitrary numbers so I haven't finished tweaking or testing for stability yet, but so far my write and copy have jumped by healthy amounts. And this whole time I was thinking it was the motherboard itself that was doing this and not the timings 😕 https://imagizer.imageshack.com/v2/534x515q70/924/G6Rz6x.png

#6133946

wavetrex

2023-05-21 00:32

wavetrex:

I'm pretty sure the best binned overclocker memory will reach 12000 or higher before DDR5 is replaced by something else.

https://videocardz.com/newz/ddr5-memory-has-been-overclocked-to-11202-mt-s Already at 11000+ ... damn ! I think I was severely underestimating in my previous post how fast DDR5 can go. Stunning !