Neo Forza MARS DDR4-4000 2x 32 GB

[ir_cow]

Neo Forza MARS takes this high-capacity memory kit to the limits. Configured as 2x 32 GB, with a 4000 MT/s memory profile and modest primary timings of 19-23-23-46, Neo Forza is looking to grab the attention of content creators. Follow along as we dive into this review and see what Neo Forza has to offer!

Show full review

 

[tabascosauz]

@ir_cow 16Gb CJR. I wouldn't just call it CJR and leave it at that, it's a new breed altogether from 8Gb CJR. Doesn't seem to be a lot of information on it, it's not a top dog but in comparison to 8Gb CJR it doesn't seem to be much worse. I don't think 8Gb CJR scales at all past about 1.4-1.45V so I guess the 16Gb is already a bit better?? So it's somewhere between CJR and DJR, but no idea if 1.5V is actually sustainable, 1.45V is already a little dangerous for 8Gb CJR (piece of cake for DJR).

CJR evolved but not strong, 16Gb new Hynix CJR overclocking test - iMedia (min.news)
translated/ripped off of
CJR进化了但不强，16Gb新海力士CJR超频测试 - 哔哩哔哩 (bilibili.com)

Also makes a lot more sense and standardization in OC community to call it a 16Gb IC than a 4GB part (isn't 16Gb = 2GB???).

As usual thanks for peeling off the heatspreaders and showing us the chip.

 

[ThrashZone]

Hi,
1.4v for 32gb sticks isn't bad
My royal 4000c16 2x16gb sticks need that much

 

[ir_cow]

@tabascosauz It was suggested to me not to raise the voltage or encourage it. I couldn't find the datasheet on these 4GB ICs either. 1.45V is probably still safe. Admittedly I didn't spend much time with the overclocking and just lowered the primaries until it wouldn't boot with 1.4v. These 4GB ICs are not the same as the 1GB for sure.

 

[InVasMani]

I'd say in regard to pushing above 4000MT/s DDR4 a lot is tied to the MB quality as well as CPU quality and memory kit quality. It's probably more a combination of the three than one in particular and also cooling. I'd also consider the Mushkin 2x32GB kits of DDR4 if looking at these the 3200 CL 14 and the 3600 CL16 kits in particular are similar, but higher true latency and better fallback on official MB/CPU support. Depending on usage the higher MT/s with worse true latency isn't going to be a worth while trade off, but there are cases favorable to each at the same time.

It's not merely bandwidth vs latency, but also periodic refresh intervals that can skew benchmarks due to synchronization though if you test enough times can favor either kit especially without averaging test runs well which can make things a bit misleading. It's refresh rate intervals aspect also that makes tightening CL a bit less important than raising MT/s conversely though you can shift it towards CL latency tightening at lower MT/s to improve a bit of stability.

I think regard to voltage the high density kits are a little more tolerant to slightly higher voltage than lower density kits as a rule of thumb if I'm not mistaken due to a little higher capacitance within the cells from more density. It's not usually good or better necessarily though heat is a challenge at higher density. They are the perfect candidate for a bit of active memory cooling especially in DIMM configurations.

 

[ir_cow]

@InVasMani the games are 3 runs average together. Can't get much better without just wasting time.

 

[Raven Rampkin]

@tabascosauz It was suggested to me not to raise the voltage or encourage it. I couldn't find the datasheet on these 4GB ICs either. 1.45V is probably still safe. Admittedly I didn't spend much time with the overclocking and just lowered the primaries until it wouldn't boot with 1.4v. These 4GB ICs are not the same as the 1GB for sure.

G.skill rates 16Gbit CJR up to 1.55V themselves,

 

[InVasMani]

@InVasMani the games are 3 runs average together. Can't get much better without just wasting time.

I wasn't questioning your test methods or reasons behind them. That was more of a remark in regard to a lot of reviews you sometimes see with memory comparisons of CL vs MT/s that can leave out of context refresh rate intervals and some of the abnormalities that they can incur in the results from synchronization that's periodic and you can bump up refresh intervals, but it's worse on stability slightly it's normally worth doing for the added bandwidth, but complicated at times and exacerbated in higher density kits on the stability since those are more prone to current leakage with more IC's.

G.skill rates 16Gbit CJR up to 1.55V themselves,

G. Skill rates some of it's binned DIMM's up to like 1.6V/1.65V actually. You're still dependent upon silicone lottery of the CPU's IMC however and at the same time MB support and/or design capabilities.

 

[ir_cow]

G.skill rates 16Gbit CJR up to 1.55V themselves,

Good find. Proabably means they can do 1.6V like DRJ ICs. But I'm not going to try.

Edit: Looking closely at that picture, those are 16GB per DIMM. Which is 2GB ICs not 4GB like the Neo Forza kit. I do see Gskill sells 4600 (2x 32GB) with 1.55V.

I'm running these in my personal computer right now. 3600 CL16 so I don't have to play with the FLCK/SoC voltage. I would have bought the correct speed in the first place if I didn't have these.

 

[tabascosauz]

Good find. Proabably means they can do 1.6V like DRJ ICs. But I'm not going to try.

Edit: Looking closely at that picture, those are 16GB per DIMM. Which is 2GB ICs not 4GB like the Neo Forza kit. I do see Gskill sells 4600 (2x 32GB) with 1.55V.

I'm running these in my personal computer right now. 3600 CL16 so I don't have to play with the FLCK/SoC voltage. I would have bought the correct speed in the first place if I didn't have these.

No, he's right, the G.skill kit is indeed 16Gb CJR. It's single rank, but the S820C in the 042 code tells all. An 8Gb CJR would be 8820C/8821C.

 

[The King]

@ir_cow Zentiming for AMD system would be a nice addition.

Was FCLK/MCLK/UCLK 1:1:1? No Whea 19 erorrs @ 2000 FCLK?

This Gigabyte motherboard was used for the benchmarks. For AMD, voltage tuning was required to keep a FCLK ratio of 2000 MHz, and therefore retain 1:1 synchronisation for optimal performance.

 

[ir_cow]

@ir_cow Zentiming for AMD system would be a nice addition.

Was FCLK/MCLK/UCLK 1:1:1? No Whea 19 erorrs @ 2000 FCLK?

No errors. ran TM5 and Karhu software before benchmarks to make sure things are stable. I should do this all the time, but usually its only when I have to tinker with the CPU voltages for XMP. Overclocking, every time, otherwise how do you know its stable right??

As for Zentimings and the OC section. It has been lost. Same with the next DDR4 review Lesson learned, do not over multitask. When I get time, I'll throw the memory back in and grab zentimings for you.

 

[tomcug]

Edit: Looking closely at that picture, those are 16GB per DIMM. Which is 2GB ICs not 4GB like the Neo Forza kit. I do see Gskill sells 4600 (2x 32GB) with 1.55V.

That Neo Forza kit is using 2 GB ICs (aka 16 Gb), not 4 GB ones. 2 GB * 16 (it's double-sided) = 32 GB for a single module.

 

[ir_cow]

That Neo Forza kit is using 2 GB ICs (aka 16 Gb), not 4 GB ones. 2 GB * 16 (it's double-sided) = 32 GB for a single module.

Your right. Must fix review. I got things mixed up. Math is hard