FEATURED Autumn Equinox AIDA64 Cache and Memory Competition

[storm-chaser]

Rules of Engagement:
-Activated version preferred, please try to use the full version if possible see below
-The ONLY benchmark we will be scoring here is the AIDA64 Cache and Memory Benchmark
-Leaderboard will include champions in READ, WRITE, COPY and LATENCY divisions
-Please include a snip of your result and post it here (that's it, plus the CPUz screenshot so I have all the hardware info)
-We will also include awards for those with the highest overclock percentage for FSB (as revealed by AIDA64)
-We will also have a combativity award which is given to a member who does exceedingly well in response to another submission or posts up a good number or number(s), and is rewarded
-And a combination champion for a member who does exceedingly well across the leaderboards
-Leaderboards will be updated on a near daily basis (will try my best to update daily)
-Competition will run until the end of the year (deadline = December 31st, 2018 11:59pm)
-Please include a brief summary of your system, if you wish. I.E. CPU, Memory, etc, RAM AMOUNT is a must
-Multiple admissions and hardware changes are allowed, no OS requirement
-Admissions will be split into AMD/INTEL sections and DUAL/TRIPLE/QUAD Channel results to keep it competitive
-Run either AIDA64 extreme OR AIDA64 Engineer edition(s) - Again, I will help getting full version set up
-Highest Latency & Red Lantern Award for slowest rigs (vintage machine(s) will likely win out here, NO underclocking
-Red Lantern Competition is limited to 65NM or higher, as we want to encourage the use of vintage technology
-IF adding multiple submissions, please include which category you want me to put you in (i.e latency or bandwidth category)
-no laptops (unless it's something really vintage)

Leaderboards are up! *Competition is CLOSED*



________________________________________________________________________________________________




________________________________________________________________________________________________


________________________________________________________________________________________________


COMBINATION CLASSIFICATION:
1st - Woomack
2nd - PolRoger
3rd - mackerel

COMBATIVITY AWARD:
-custom90gt
-PolRoger
____________________________________________________________________________________

Highest OC %
Woomack 292% (% FSB)
EarthDog 57% (% CPU)
mackerel 5.2GHz (Max CPU Clock)
Woomack 4500Mhz (Max Memory Clock)
____________________________________________________________________________________

Red Lantern Award:
PolRoger

 

[Woomack]

I would split results into memory channels or Intel/AMD or something like that. It's just because I can (or someone else, no difference) drop a quad channel result and 90% of forum members won't be able to beat it on their PC. What I mean is that it would be more competitive if more people had a chance to win. You also can't compare AMD to Intel when one will stuck at DDR4-3600/3733 and the other one will go up to DDR4-4500+. I just wish to see higher interest in the competition like that.
Would be nice to keep it till the end of the year instead of end of the month, so there will be more results.
It's your compeition so you decide

I'm installing Server 2016 on my X299 rig so maybe I will post something later.

It would be nice is mods could move this thread to memory section so it will be more visible.

 

[RollingThunder]

Moved!

 

[Woomack]

Moved!

thanks, now you can post something

 

[Woomack]

TR [email protected] / ASRock X399M Taichi / 4x8GB Patriot Viper 4 DDR4-3733@3733 14-14-14-32 1N / AMD Quad Channel




[email protected] / ASRock Z390 Phantom Gaming-ITX/ac / 2x8GB Kingston HyperX Predator RGB DDR4-4000@4500 17-17-17-37 2N / Intel Dual Channel


This is what I had on my desktop. I guess I have some better results but won't post anything that I was using in previous competitions or anything older.

 

[storm-chaser]

This is what I had on my desktop. I guess I have some better results but won't post anything that I was using in previous competitions or anything older.

This will keep the competition fresh and incentivize people to utilize current hardware... much appreciated. Lets take this approach and go forward. Since I'm out of town at the moment I wont have a chance to bench my rigs until tomorrow. This also means I won't be able to create a custom leader-board until I get back home, for now it will be text only.

EDIT: 53 views and only one submission... come on people lets make this happen!

 

[EarthDog]

Here is the 7960X as it sits daily...tweak free.



I'll spin up the 9900K as well.

 

[Blaylock]

I don't have a DDR4 rig but wouldn't mind running this a few times. Will there be a DDR, DDR2, & DDR3 sections or is that too segmented?

 

[storm-chaser]

I don't have a DDR4 rig but wouldn't mind running this a few times. Will there be a DDR, DDR2, & DDR3 sections or is that too segmented?

Scoring will be done with the average of your read, write and copy speeds. In addition we may have separate classifications for dual triple and quad channel - depending on interest (I don't think we will segment further than that). I'm also away from home right now so I don't have access to my leader-board templates, but we will get something up in the next couple days. Don't want to get to granular because we want to see clear leadership across the board.

For now it will be text only for average memory bandwidth totals and memory latency in a separate division. There will be a combination classification, a combativity award and red lantern award gifted to the slowest rig. So we have to find that fine line between too much data and too little data.

EDIT: Hope this doesn't confuse anyone. Once we get more participants, I think the scoring will become clear.

 

[storm-chaser]

These rigs are mostly intact and all these benchmarks have been run in the past month in the aim the compete in this competition...

960T


QX6850


E8400


E7500


1600T (Unlocked 960T)


2nd 960T (NO UNLOCK)

 

[custom90gt]

Sadly this CPU and or MB is not too happy running 4 sticks in dual channel so only stock memory speeds for this test:
[email protected] / MSI Gaming 7 ACK / 4x8GB G.Skill Trident Z RGB DDR4-3200@3200 14/14/14/34 1N / Intel Dual Channel

 

[PolRoger]

Some triple channel runs:

DDR3-2275C9 Kingston Elpida variant ic?


DDR3-2175C8 GSkill Elpida BBSE ic:


DDR3-2275C9 GSkill Elpida BBSE ic:



Some quad channel runs:

DDR3-2200C8 GSkill with (8 dimms) PSC ic:


DDR3-2333C7 GSkill PSC ic:


Daily dual channel rig (kicked up a notch)…
DDR4-3866C16 GSkill Samsung ic:

 

[mackerel]

In addition we may have separate classifications for dual triple and quad channel - depending on interest (I don't think we will segment further than that).

Random thought: how about expressing bandwidth per channel? That is, dual channel bandwidths divide by two, 3 channel by 3, etc. I don't know if scaling works that well for that to be fair. Otherwise I agree with previous comments that it isn't really fair for dual and quad channel to battle each other on bandwidth as the most common configurations today. Normalising would also be inclusive of 3 channel.

I also assume latency isn't affected channels so that doesn't need to be separated out.

I'm also wondering if DDR3 might have latency advantages... just for something different to DDR4 to play with for a bit.

 

[PolRoger]

Two slow rig submissions...

Socket 370 Celeron 466Mhz single channel:


Socket 370 Pentium III 933Mhz Coppermine single channel:

 

[EarthDog]

Omg... ssds are faster than that RAM.

 

[Woomack]

Random thought: how about expressing bandwidth per channel? That is, dual channel bandwidths divide by two, 3 channel by 3, etc. I don't know if scaling works that well for that to be fair. Otherwise I agree with previous comments that it isn't really fair for dual and quad channel to battle each other on bandwidth as the most common configurations today. Normalising would also be inclusive of 3 channel.

I also assume latency isn't affected channels so that doesn't need to be separated out.

I'm also wondering if DDR3 might have latency advantages... just for something different to DDR4 to play with for a bit.

Bandwidth per channel would always give better results on dual channel Intel setups. Check how dual channel is scaling on new Intels. Can make 60GB/s+ on 2 channels while you can't really make much above 100GB/s in quad channel.
It's something about max bandwidth of memory controller and related to some other things and delays of IMC with more channels. Latency is much worse with more channels. It's not related to the channels but more to available architecture. There is simply no quad channel IMC which can be close to any dual channel Intel in last ~7 years.

DDR3 is pretty good in latency if you push it higher. Many overclockers were using DDR3 for some benchmarks till better DDR4 were out. I mean something like 2600+ C9 on Samsung or 2666+ CL6-8 on PSC is showing nice results.
Actually you can try as my latency result isn't any special. I think that Johan had 32ns in some other competition and I couldn't beat it back then. Latency depends also on cache speed and is highly affected by sub timings. My current RAM doesn't like too tight timings and is acting weird but I will try to work on it some more.



TR [email protected] / ASRock X399M Taichi / 4x8GB Patriot Viper 4 DDR4-3733@3733 14-14-14-28 1N / AMD Quad Channel

 

[mackerel]

DDR3 is pretty good in latency if you push it higher. Many overclockers were using DDR3 for some benchmarks till better DDR4 were out. I mean something like 2600+ C9 on Samsung or 2666+ CL6-8 on PSC is showing nice results.
Actually you can try as my latency result isn't any special. I think that Johan had 32ns in some other competition and I couldn't beat it back then. Latency depends also on cache speed and is highly affected by sub timings. My current RAM doesn't like too tight timings and is acting weird but I will try to work on it some more.

I was thinking, I do have a pair of DDR3 Vengeance something or other 2400 sticks. Never tried overclocking them though. Thought that might be fun.

I had a look at my past B-die dual channel OC result, roughly 38ns and 53GB/s average but I never managed to push the primary timings hard. No particular attempts at quad channel yet so maybe will get something down at stock as baseline and take it from there.

I also have some ideas on the slowest ram part, but I don't think I have the ability to get an OS on it that would run aida. Wait, this might be worth a trip through my work recycling piles. People often bring old systems to get rid of, something might be old enough... I haven't seen anything new enough to get my interest otherwise.

 

[Woomack]

Run any stability test during memory/cache benchmark and it will give you 0GB/s result

Here is one more from last week. Lower clock at tight timings can make lower latency.

 

[PolRoger]

Red Lantern (slow rig) AMD...

Athlon XP 2500+ Barton, Socket A (unlocked) DDR333 single channel:
View attachment Athlon XP 2500+DDR333 single channel Aida.bmpView attachment Athlon XP 2500+ OC to XP3000+ DDR333 single channel CPUID.bmp

 

[storm-chaser]

Okay guys... Thanks for the subs! I have a little time now to work on the leaderboards and I should have them completed and posted up by the end of the day.

In terms of scaling, I was thinking the same thing mackerel, to level the playing field a bit we could find an appropriate divider, however, it looks like Woomack has confirmed we cannot use a set ratio here as Dual channel scales differently than triple or quad.

For the slow rig portion of the competition, lets set a couple ground rules:
1) We are looking for the highest latency and slowest r/w/c speed.
2) Vintage machines will likely win here... ie. no underclocking, just use older tech

EDIT: I have to go sell my car now so the leaderboards will have to wait until this evening...