What's the ideal case to use to test AIOs...

[RevGAM]

...on top ,on the side, and in the front, tubes up and down, given that AIOs come with 250-450mm tubes and my long GC get in the way of having the tubes at the bottom of the rad when mounted on the mobo, and the AIO is in the front or on the side, tubes down? Please, don't suggest something that's going to destroy my pitiful savings.

 

I wanted to test AIOs with the rad in the tubes-down position, so I've been using a vertical mount that sits on the shroud (from EZDIY-FAB). This mount is going to give me different results than having the GC in the mobo slot, but at least it allows me to swivel the GC to improve the temps.

 

I want to test AIOs with the mobo in the PCIe slot, but my XFX GC is too long to do tubes-down, and it would be ideal if the case has top, front and side mounts for 360mm AIOs. I realize that this just may not be possible so if I can only mount it on the side and front, that is fine because I can accomplish the testing on top in my current case. Whatever the case (pun intended), it needs to have enough clearance to hold the thickest AIOs, too.

 

My current case is the Corsair iCUE 5000x RGB Mid-tower ATX. I also have the Zalman i4 ATX and the Kingdom Classico Storage Master ATX. The i4 also has both a GC support and a vertical holder, and the side panel is mesh, but there's no riser cable.

 

Parts

Asus TG H670 Pro Wifi D4

12700k

Patriot Viper Steel 2x32 GB 1800 MT/s (3600 MHz, for those who use the marketing measurement)

4 SSDs

1 SATA HDD (not required)

XFX Speedster MERC 319 6800 XT Core (basically the same as the SWFT, but with useless lights)

Corsair RM850x (or EVGA 850GQ Gold)

Noctua A12x25 (exhaust)

TYIA!

[RONOTHAN##]

40 minutes ago, RevGAM said:

Patriot Viper Steel 2x32 GB 1800 MT/s (3600 MHz, for those who use the marketing measurement)

Just because it will annoy me otherwise, the units here are backwards. It would be 1800MHz or 3600MT/s. 

 

AFAIK there aren't any cases that can fit a massive GPU and tubes at the bottom, so you're probably just out of luck here (I could be wrong though). It shouldn't affect performance though having the tubes at the top or the bottom, so I don't see much of a point to worrying about this anyway. 

[thrasher_565]

imo the best test case is the Thermaltake CTE C750 becuse it can have 140mm fans everywhere... thow the build quolaty is crap thow the sides have no screws to hold them on.

[RevGAM]

8 hours ago, RONOTHAN## said:

AFAIK there aren't any cases that can fit a massive GPU and tubes at the bottom, so you're probably just out of luck here (I could be wrong though). It shouldn't affect performance though having the tubes at the top or the bottom, so I don't see much of a point to worrying about this anyway. 

I'm trying to prove whether MSI's article on AIO placement is correct.

 

8 hours ago, RONOTHAN## said:

Just because it will annoy me otherwise, the units here are backwards. It would be 1800MHz or 3600MT/s.

I contacted 3DMark (FutureMark) a year ago about the 3,600 MHz. I was aware of MT/s but didn't understand it all that well. I was looking for info on what frequency I should see displayed by my PC about my RAM. This was the response I got. You're welcome to follow up with Jarno, who is very nice, about this. I could've misunderstood the bolded part, but it certainly seems to be saying that one stick is 1,800 MT/s. I know you know a lot mroe than I do about computers so I'll let you sort that out.

Quote

UI shows similar information. The @ value is "marketing MHz" due to how DDR works - too many places call it MHz while it is technically actually MT/s and the actual memory bus frequency is lower (for example, in this DDR4 case, half of this value) - we used to show actual frequency, but people kept complaining we are showing wrong value and it did not seem like useful use of our time to start educating every PC user about what double data rate means.

 

--

Jarno Kokko

QA & Support Lead

Also, HWInfo's summary of my system shows that each stick has a clock speed of 1800. This stuff is outside of my comfort zone, but I'm happy to learn.

[RONOTHAN##]

1 minute ago, RevGAM said:

I could've misunderstood the bolded part,

Yeah, that's what happened here. Frequency is always measured in terms of Hz, while data rate can use a few different units (MT/s is common for desktop memory, while Mbps is common for GPU memory). What they were saying is that the memory frequency, it's hertz value, will be half of it's data rate, the MT/s value. The "Marketing MHz" statement was more to say that the actual marketing term was wrong, and thus the real unit is something different. It was worded a little confusingly, so a misunderstanding is understandable. 

 

I prefer to go the other method and just say the data rate after the memory spec (I.E. DDR4 3600) as it's generally less confusing to most people reading without having to explain the whole "memory clock is half the advertised data rate" thing while also being a valid method of expressing memory data rates. The data rate (the part that's advertised with the wrong unit) is the most important metric for actually figuring out the memory performance as it's used to calculate theoretical maximum memory performance with the formula data rate * 64 * 2 / 8 where the 64 is the bits per memory channel, the 2 is the number of memory channels, and the 8 converts from bits to bytes, so something like DDR4 3600 will have a maximum theoretical bandwidth of 57,600 MB/s, which if you use something like Aida64's memory bandwidth test is right about the bandwidth number it will tell you. 

[RevGAM]

4 hours ago, RONOTHAN## said:

Yeah, that's what happened here. Frequency is always measured in terms of Hz, while data rate can use a few different units (MT/s is common for desktop memory, while Mbps is common for GPU memory). What they were saying is that the memory frequency, it's hertz value, will be half of it's data rate, the MT/s value. The "Marketing MHz" statement was more to say that the actual marketing term was wrong, and thus the real unit is something different. It was worded a little confusingly, so a misunderstanding is understandable. 

 

I prefer to go the other method and just say the data rate after the memory spec (I.E. DDR4 3600) as it's generally less confusing to most people reading without having to explain the whole "memory clock is half the advertised data rate" thing while also being a valid method of expressing memory data rates. The data rate (the part that's advertised with the wrong unit) is the most important metric for actually figuring out the memory performance as it's used to calculate theoretical maximum memory performance with the formula data rate * 64 * 2 / 8 where the 64 is the bits per memory channel, the 2 is the number of memory channels, and the 8 converts from bits to bytes, so something like DDR4 3600 will have a maximum theoretical bandwidth of 57,600 MB/s, which if you use something like Aida64's memory bandwidth test is right about the bandwidth number it will tell you. 

So what exactly was the doubled number he mentioned?

[RONOTHAN##]

3 hours ago, RevGAM said:

So what exactly was the doubled number he mentioned?

The memory frequency (the actual memory clock reported by software like CPU-Z and HWInfo, measured in MHz) is doubled to get the memory data rate (the number advertised by memory manufacturers, should be measured in MT/s)

[RevGAM]

25 minutes ago, RONOTHAN## said:

The memory frequency (the actual memory clock reported by software like CPU-Z and HWInfo, measured in MHz) is doubled to get the memory data rate (the number advertised by memory manufacturers, should be measured in MT/s)

So, it seems like you're saying that the 3600 MHz is derived from the 1800 clock? Is it correct that the specs should show 3600 MT/s?

 

I'm finding this a bit confusing so, if you can give me a real example with numbers so that I can see how the various measurements relate to each other, I think that would be helpful. 

[RONOTHAN##]

1 minute ago, RevGAM said:

So, it seems like you're saying that the 3600 MHz is derived from the 1800 clock? Is it correct that the specs should show 3600 MT/s?

Yes. 

 

1 minute ago, RevGAM said:

I'm finding this a bit confusing so, if you can give me a real example with numbers so that I can see how the various measurements relate to each other, I think that would be helpful. 

Sure. 1800MHz = 3600MT/s, 1600MHz = 3200MT/s, 3200MHz = 6400MT/s, etc. I can get screenshots of my own system later today if you want them, but HWInfo and CPU-Z will report that the frequency is 3600MHz, while the data rate reported in tools like Aida64 will be 7200MT/s or just DDR5 7200. 

 

If it makes more sense to think of it this way, 1800MHz, or 1,800,000,000 oscillations per second, is the frequency of the sine wave on the memory bus, whereas because memory data transfers take place on both the rising edge and the falling edge of that sine wave, there is 2 transfers for every oscillation and therefore 2x1,800,000,000 transfers per second, or 3,600,000,000 transfers per second, or 3600 mega-transfers per second (MT/s). 

[RevGAM]

1 hour ago, RONOTHAN## said:

Yes. 

 

Sure. 1800MHz = 3600MT/s, 1600MHz = 3200MT/s, 3200MHz = 6400MT/s, etc. I can get screenshots of my own system later today if you want them, but HWInfo and CPU-Z will report that the frequency is 3600MHz, while the data rate reported in tools like Aida64 will be 7200MT/s or just DDR5 7200. 

 

If it makes more sense to think of it this way, 1800MHz, or 1,800,000,000 oscillations per second, is the frequency of the sine wave on the memory bus, whereas because memory data transfers take place on both the rising edge and the falling edge of that sine wave, there is 2 transfers for every oscillation and therefore 2x1,800,000,000 transfers per second, or 3,600,000,000 transfers per second, or 3600 mega-transfers per second (MT/s). 

Thanks so much!

 

 

[thrasher_565]

7 hours ago, RevGAM said:

Thanks so much!

 

 

cpuz shows half of the ramspeed for w/e reason. and if you buy say 3200mhz ram dose not mean it runs stock a thows settings just might oc to that number.

 

also ram has 2 names like ddr3 is also called pc3 why i dont no...

then there also rank witch is how many chips and is it on both sides.

 

then there descktop ram and server ram and was amd ram too...

Edited March 14 by thrasher_565

[NorKris]

Room temp as to be measured for the tests. 

fan speeds need to be fixed (same speed all the time) 

max and avg watt during cpu loads needs to be measured 

[RevGAM]

48 minutes ago, NorKris said:

Room temp as to be measured for the tests. 

fan speeds need to be fixed (same speed all the time) 

max and avg watt during cpu loads needs to be measured 

Yes, sir! For this test, though, I'm just trying to verify MSI's findings.

 

Max and avg watts of what? Each core? 

[NorKris]

1 minute ago, RevGAM said:

Yes, sir! For this test, though, I'm just trying to verify MSI's findings.

 

thats not a just xD

 

1 minute ago, RevGAM said:

Max and avg watts of what? Each core? 

package and the most difficult core  should be ok  

[RevGAM]

4 minutes ago, NorKris said:

thats not a just xD

Does that mean it is/I am unjust?  

4 minutes ago, NorKris said:

package and the most difficult core  should be ok

Hmmm... I'll have to ask Santa which core is the most difficult. Generally, I would say an apple's core.  

 

On second thought, though, a pineapple's!

[RevGAM]

16 minutes ago, NorKris said:

thats not a just xD

 

package and the most difficult core  should be ok  

I changed my mind.  A pine cone's!! :  

[RevGAM]

1 hour ago, NorKris said:

thats not a just xD

 

package and the most difficult core  should be ok  

The softest core is in a frozen custard cone!!!!

[NorKris]

this is getting hardcore

[RevGAM]

32 minutes ago, NorKris said:

this is getting hardcore

XXX?

[NorKris]

1 minute ago, RevGAM said:

XXX?