How high can you fly? AIDA64 memory bandwidth competition

[storm-chaser]

Last year we had a similar AIDA64 competition entitled "How low can you go?" which focused purely on memory latency. This time around we are focusing on bandwidth, so I wanted the catch phrase to follow in a similar vein.

I know there is a thread for AIDA64 scores, but I thought generating some local competition here would be a little more fun and entertaining than just submitting a score for the database. We are looking at bandwidth only, so it's not the same game.

Rules of Engagement:
-Latency is out the window - The ONLY benchmark(s) we will be looking at here will be your memory read, write and copy speeds
-Emphasis is on dual channel mode. Overall champion will be determined by averaging your three results (r/w/c). So the rig with the highest memory bandwidth average in a dual channel configuration wins the overall title.
-Recommend you make a good run with dual channel, and then add memory if you want to compete in the other categories with more channels
-The leaderboard ranking Results will be listed from highest to lowest in terms of bandwidth, but each user will have a data point for placings in their class.
-Run the AIDA64 Cache and Memory Benchmark. It takes but a minute! Please include a snip of your result and post it here
-Competition will run until noon on January 1st, 2021
-Please include a CPU z screenshot as well (see below for layout)
-Multiple admissions and hardware changes are allowed
-Multiple rigs are allowed
-This is an unlimited competition, meaning there are no restrictions on hardware
-Full version of AIDA64 preferred. Please PM me if you are using the AIDA64 demo version for specifics on how to get the proper data points

I will get started with my 9600KF rig as an example to follow in terms of layout:
Please use this layout with your submission.



I have color coded the results for easy comparisons.
Red = Team AMD
Blue = Team Intel

Anything in bold is notable...

Leaderboard Updated 12-29
6:30PM

 

[Dinnercore]

What about dual, quad, hexa, octa etc. channel setups?

Dual channel will have no chance vs quad.

 

[storm-chaser]

What about dual, quad, hexa, octa etc. channel setups?

Dual channel will have no chance vs quad.

Okay, good point.

Not sure the best way around this. Perhaps we can have winners in all those categories, but that might make the competition too fragmented if we are keeping track of too much data.

Any suggestions here?

I still think the rig with the most bandwidth should win the competition but we want to make it competitive for ALL participants if at all possible...

 

[Deleted member 194470]

So you have to buy aida64 to do this test? It does not seem like a good test then. I downloaded the trial version but it doesn't show all the results, some say "trial version"

 

[storm-chaser]

So you have to buy aida64 to do this test? It does not seem like a good test then. I downloaded the trial version but it doesn't show all the results, some say "trial version"

It sucks. But if you run it enough times you might get all the data that I will need. Try running it a couple times and see if the other stats come through. Alternatively, you could try running the memory read / write and Copy tests manually by going to the benchmark menu: In this case, you may have do the same and run them a few times each to get all the data for the leaderboard. I have another fix and I will PM you later about that.

 

[Deleted member 194470]

Ok nice loophole it worked. I know I am not going to win anyway lol.

 

[xman2007]

I still think the rig with the most bandwidth should win the competition but we want to make it competitive for ALL participants if at all possible...

That will make it the least competitive IMO as when you start getting quad/octa channel results, normal everyday dual channel results will be obsolete in comparison, I mean if it's just the fastest then it just depends who has the most money to throw at a rig/memory etc

 

[storm-chaser]

That will make it the least competitive IMO as when you start getting quad/octa channel results, normal everyday dual channel results will be obsolete in comparison, I mean if it's just the fastest then it just depends who has the most money to throw at a rig/memory etc

Right, I get that. Still trying to figure out a viable way for everyone to compete and have fun.

I think we will have leaders in:

dual channel
quad channel
hexa channel
Octa channel

Maybe this will be a good compromise. All that being said, it's not uncommon for the drag car with the most money behind it to win the race. so I see no problem with having the overall win go to the highest bandwidth rig, regardless of how expensive or how many memory channels it has.

Results will still be listed from highest to lowest bandwidth, but each user will have a data point for placings in their class.

Sound good to everyone?

 

[Dinnercore]

Making a class for each is a good idea.

What I would suggest is: Put the focus on dual channel results.

As in, the price will be given to the best dual channel average. Everyone can compete in that class, if you run quad channel just take some RAM out. And if quad channel architectures suffers from going dual, the better. Means you can't just buy the first place with an expensive rig you happen to have sit somewhere.

 

[storm-chaser]

Making a class for each is a good idea.

What I would suggest is: Put the focus on dual channel results.

As in, the price will be given to the best dual channel average. Everyone can compete in that class, if you run quad channel just take some RAM out. And if quad channel architectures suffers from going dual, the better. Means you can't just buy the first place with an expensive rig you happen to have sit somewhere.

Okay, sounds good to me --- I see what you are saying. We will keep the focus on dual channel, and the best prize, the DDR4 Trident Z memory kit will go to the first place winner in that category. The other categories will get prizes as well, but the competition as a whole, will put emphasis on the dual channel ranking as your primary placing.

Okay guys, we are ready to begin.
Should be a fun and healthy COVID activity
Lets just hope we get enough interest here to make this viable.

 

[mrthanhnguyen]

ddr4 4700 and above will beat me. Can't get this dual rank kit more than 4600.

 

[storm-chaser]

Quick Update. I added the three submissions we have so far and created this leaderboard below. It's just beta right now, so let me know if you think I should add/change anything before proceeding. FYI this leaderboard will be posted up at the top, I just have to get a mod to unlock the first post so I can edit it.

 

[jesdals]

You should make sure you run the same version of the software, did a trial run of the latest edition:

The test still seems to favour boost clocks so you would proberly get best scores after fresh boot with low temps on cpu

Btw nice that you can see most of the settings

You can get the values as shown on the trial version by running the individual menues

If you want your best score on a Ryzen 3000 cpu try to set a fixed cpu OC and you will proberly see the highst score due to Aida64 favors cpu speed

 

[Zyll Goliat]

I am sure I am not going to win but for the sake of good sportsmanship and to fill this info-data base I will post my results:

Xeon 2650 V2 8c/16t (OC)3,4Ghz(all cores)-turbo boost 3,85Ghz
Memory Kingston Hyper blue (OC)1510Mhz(1333M)hz quad Channel/timings 9-9-9-30 CR2

 

[mrthanhnguyen]

Come on guys. Aida64 is only $1.5 on ebay.

 

[jesdals]

Come on guys. Aida64 is only $1.5 on ebay.

I am dirt cheap and saveing the money for a 5950x and a 6900xt come december...

 

[Nero1024]

I think you should limit RAM frequency for this competition or else, whoever has most expensive rig will win by default

 

[Dinnercore]

I think you should limit RAM frequency for this competition or else, whoever has most expensive rig will win by default

How would that change a thing about expensive RAM? You can still buy better timings then. It would change nothing. I think just limiting to dual channel is fine. And after all, this is an OC oriented competition I think. So you need some factor to be tweaked for competitions sake.

You can find cheap B-Die and OC it, like my Viper Steel:
This is at just below 1.50V for daily. A 32GB 4000 CL15 kit was 600$ by the time I bought these for 200$.

Just as a comparison for quad-channel, I don't want to participate.

 

[mb194dc]

Wont' do better with dual 2667 ram I think... Can't quite make it to 40 multiplier and complete the bench. Crash every time no matter what SOC or core voltage I'm using. Memory bandwidth in Aida scales with clock speed when using version of ryzen master I have. Never figured out why, OC in the bios doesn't produce the same results.

CPU Queen:

 

[Nero1024]

How would that change a thing about expensive RAM? You can still buy better timings then. It would change nothing. I think just limiting to dual channel is fine. And after all, this is an OC oriented competition I think. So you need some factor to be tweaked for competitions sake.

You can find cheap B-Die and OC it, like my Viper Steel:
This is at just below 1.50V for daily. A 32GB 4000 CL15 kit was 600$ by the time I bought these for 200$.
View attachment 173754

Just as a comparison for quad-channel, I don't want to participate.

Anyway. I am pretty sure this competition is not fair. If you know you can't beat the top results with your rig, it's pretty much pointless to even do the test.

I have the leaderboard template ready to go, so if we get enough interest, I will post up a leaderboard. I know there is a thread for AIDA64 scores, but I thought generating some local competition here would be a little more fun and entertaining than just submitting a score for the database. We are looking at bandwidth only, so it's not the same game.

Rules of Engagement:
-Latency is out the window - The ONLY benchmark(s) we will be looking at here will be your memory read, write and copy speeds
-Emphasis is on dual channel mode. Overall champion will be determined by averaging your three results (r/w/c). So the rig with the highest memory bandwidth average in a dual channel configuration wins the overall title.
-Recommend you make a good run with dual channel, and then add memory if you want to compete in the other categories with more channels
-The leaderboard ranking Results will be listed from highest to lowest in terms of bandwidth, but each user will have a data point for placings in their class. We could also have a single channel comp for kicks, if anyone is interested in that.
-Run the AIDA64 Cache and Memory Benchmark. It takes but a minute! Please include a snip of your result and post it here
-Competition will run until noon on January 1st, 2021
-Please include a CPU z screenshot as well (see below for layout)
-Multiple admissions and hardware changes are allowed
-Multiple rigs are allowed
-This is an unlimited competition, meaning there are no restrictions on hardware
-Full version of AIDA64 preferred. Please PM me if you are using the AIDA64 demo version for specifics on how to get the proper data points
-I have a massive surplus of computer parts so instead of selling them I figured would offer them here to amplify the competitive nature of this comp and pass them on to fellow tech enthusiasts who will know better how to use them to their full potential. Or just keep them as a trophy of sorts.

Overall winner in dual channel mode will receive a like new 16GB G.Skill Trident Z 4000MHz CL18 memory kit (B-die):

View attachment 173572

Winner in octal channel will get this like new SoundBlaster SB1350:



Winner in hexa channel will receive this like new MSI GeForce 710 1GB GPU (also have standard height rail for it):


Winner in quad channel will receive this Core 2 extreme unlocked X9000 processor - I got this thing up to 4200MHz in my Lenovo T61p, and I think there is even more headroom with a better cooling solution. I'd recommend using throttlestop to overclock:


I will get started with my 9600KF rig as an example to follow in terms of layout:
Please mimic this layout with your submission.

If you want to make it fair, it's better to just use some kind of random name picker in the internet containing everyone who leaves a message on this topic

 

[sneekypeet]

Anyway. I am pretty sure this competition is not fair. If you know you can't beat the top results with your rig, it's pretty much pointless to even do the test.



If you want to make it fair, it's better to just use some kind of random name picker in the internet containing everyone who leaves a message on this topic

No OC competition is fair if you do not have good hardware to run in it. Sorry to be this blunt, but I do not see a need to keep complaining. Either enter with your best results and hope, or move on.

 

[storm-chaser]

Here is my z820 dual processor rig to throw into the mix. AIDA64 incorrectly identifies my CPU as the E5 2667 v2, which is a retail chip. In reality I have the OEM equivalent, the nearly identical E5-2673 v2. Same except the OEM processor is slightly more voltage efficient. And comes with a substantially lower TDP than it's retail counterpart.

 

[the54thvoid]

Any reason my posts were deleted?

Same for my message. Power abuse. I guess reasoning is not allowed on these forums. Mod decides what is right and what is wrong. Kinda dictator mindset

This is a thread posted by a member to record scores for which they're giving out a prize. It's not a thread for bitching about the OC scene, the unfairness of cost, or anything else in that domain. Threads are created for purposes and those purposes are not met by off-topic comments. Posting in a forum is always subject to mod scrutiny and where mods see errant posts, or posts that do not contribute to the purpose of the thread, they can and often will, be removed.

 

[LiquidTrance]

HI,
I would like to participate, however, I don't expect to win. But If i do win the dual channel contest, please pass the prize along to the runner up since I don't have a need for the ramkit since i'm on 4 sticks. Just want to participate for fun ^^.
With that said, I share with everyone the highest frequency I could stabilize on 4x8GB.
Overclocking the GSKILL F4-4000C17Q-32GTZR ramkit(4x8gb) to 4257MHz with the same xmp primary timings.

 

[Arctucas]

This is a thread posted by a member to record scores for which they're giving out a prize. It's not a thread for bitching about the OC scene, the unfairness of cost, or anything else in that domain. Threads are created for purposes and those purposes are not met by off-topic comments. Posting in a forum is always subject to mod scrutiny and where mods see errant posts, or posts that do not contribute to the purpose of the thread, they can and often will, be removed.

PMed.