intel 13/14th gen fast RAM experiences: >7,000 MT/s is XMP stable!

[ir_cow]

I was getting a weird consistent frametime spike every 1 second in metro exodus, even in the intro video. Changing to XMP1 from XMP2 solved it. It also doesn't spike when using XMP Tweak.
I didn't try any other games but I can if anybody wants to know if it's just metro or if it's global.

@Upgrayedd have your checked out memtest5 to make sure your memory is stable yet?

 

[Upgrayedd]

@Upgrayedd have your checked out memtest5 to make sure your memory is stable yet?

I can. I've just been doing the stock 4passes in memtest86 that take forever.
I've got a couple profiles saved. I'll try 7800 XMP Tweak cache49 and then 8000 XMP1 in memtest5. I've also got 7600Tweak manually set to 7800 that passes I can try.

Could you point me where to get TestMem5?

 

[ir_cow]

memtest86 is a waste of time. 3 rounds of memtest5 with Extreme1@anta777.cfg and you'll quickly find out

Memory Testing with TestMem5 TM5 with custom configs

www.overclock.net

 

[fusseli]

memtest86 is a waste of time. 3 rounds of memtest5 with Extreme1@anta777.cfg and you'll quickly find out

Memory Testing with TestMem5 TM5 with custom configs

www.overclock.net

I'm still both surprised and disappointed by this. My stubborn mind still wants memtest86 to be the standard lol. Although this is perhaps another knocn on Asus, they provide memtest86 with bios in ROG boards, but I guess it's not worth that much eh

 

[PurSpyk!!]

My biggest issue, is with 4 dimm slots used you cannot get the rated speeds, my memory will only go to 5600 any faster and Windows has serious issues. Using an Asus z790 ProCreator Wifi MB, with 64GB Corsair Vengeance ram

 

[ir_cow]

I'm still both surprised and disappointed by this. My stubborn mind still wants memtest86 to be the standard lol. Although this is perhaps another knocn on Asus, they provide memtest86 with bios in ROG boards, but I guess it's not worth that much eh

It was good for DDR3 and mostly DDR4. Sure isn't enough for DDR5. When DDR5-9000 passes with flying colors and doesn't even get into windows before BSOD, we got a problem.

My standard is y-cruncher 2.5b + 3pass memtest5 just to get started. That is good for the most part if your lazy and don't want to do 25 loops.


Alright I guess it didn't immediately BSOD. Sure wasn't stable enough to do much of anything.

The only good tests for memest86 now is 5-6 , which will error out if one of your SA/TX/VDD2 is too low. Usually its SA for me. Test 13 is like half-ass hammer test, but it beats nothing.

 

[fusseli]

My biggest issue, is with 4 dimm slots used you cannot get the rated speeds, my memory will only go to 5600 any faster and Windows has serious issues. Using an Asus z790 ProCreator Wifi MB, with 64GB Corsair Vengeance ram

Yes that is normal these days. My work build is a 13900KS with z790 Hero and 4x 16gb 7200 c34 kit that is only stable at 6400.

 

[Upgrayedd]

It was good for DDR3 and mostly DDR4. Sure isn't enough for DDR5. When DDR5-9000 passes with flying colors and doesn't even get into windows before BSOD, we got a problem.

My standard is y-cruncher 2.5b + 3pass memtest5 just to get started. That is good for the most part if your lazy and don't want to do 25 loops.

View attachment 327716
Alright I guess it didn't immediately BSOD. Sure wasn't stable enough to do much of anything.

The only good tests for memest86 now is 5-6 , which will error out if one of your SA/TX/VDD2 is too low. Usually its SA for me. Test 13 is like half-ass hammer test, but it beats nothing.

Correct me here please but isn't the hammer tests for protection from "physical attackers"? Like one of those weird one off vulnerabilities that 99.9% of people will never have to worry about?
And how exactly does tm5 work? I've got extreme 1 anta777 loaded, I seen it go through some tests, it's on time says 24min but it doesn't seem to be doing anything anymore?

 

[fusseli]

Correct me here please but isn't the hammer tests for protection from "physical attackers"? Like one of those weird one off vulnerabilities that 99.9% of people will never have to worry about?
And how exactly does tm5 work? I've got extreme 1 @anta777 loaded, I seen it go through some tests, it's on time says 24min but it doesn't seem to be doing anything anymore?

I think it auto-runs on program start, or you can load profile and exit and restart the app. It goes through its loops and has a completion popup.

 

[Upgrayedd]

I think it auto-runs on program start, or you can load profile and exit and restart the app. It goes through its loops and has a completion popup.

It just started back up, cycle 2. My b. Guess I'll be patient until I see a finished prompt.

 

[ir_cow]

Correct me here please but isn't the hammer tests for protection from "physical attackers"?

It is, but that is the only test that really stresses the memory https://www.memtest86.com/tech_individual-test-descr.html

 

[fusseli]

I'm back in business. Here's 7200 c34 44-44-96-140 using tREFIx9 = 66 (same as XMP) with tREFI = tREFI * 700 = 46200

 

[InVasMani]

I was getting a weird consistent frametime spike every 1 second in metro exodus, even in the intro video. Changing to XMP1 from XMP2 solved it. It also doesn't spike when using XMP Tweak.
I didn't try any other games but I can if anybody wants to know if it's just metro or if it's global.

That doesn't surprise me really sub timings will change a bit between the different XMP profiles options. The most relaxed one is XMP 2 and I believe tweaked is hardest for stability, but best for performance and XMP1 is basically the balanced profile between stability and performance. It could defiantly impact program performance though mostly in subtle ways in general unless it's a pretty memory sensitive program operating.

For the primary timings outside of CAS the RAS to CAS can sometimes impact stability a good amount and tRC is rather touch and go depending on how far you tighten it, but it impacts performance moderately when tighter.

You can also try just using some of individual JDEC profile timings that are looser than the XMP if trying to get things more stable some might be a bit tighter as well and you might be able to use to tighten up performance a bit further relative to XMP, but it's something that you'd need to test on a case by case basis to see how tolerant the kit is with your system and voltages set.

 

[Upgrayedd]

It is, but that is the only test that really stresses the memory https://www.memtest86.com/tech_individual-test-descr.html

I've decided to turn hyperthreading off after seeing some game loads switch from physical thread to virtual thread on the same core for all the P cores, it should lower voltage a bit too. I had HT off for the TM5 test just now and passed at 7800Tweak.
I noticed a good hit to cache speed scores after doing this but I need to run some real game benchmarks and see if the min fps is any better or worse.
I'll be trying 8000 XMP1 right now for stability in TM5.

 

[chrcoluk]

My biggest issue, is with 4 dimm slots used you cannot get the rated speeds, my memory will only go to 5600 any faster and Windows has serious issues. Using an Asus z790 ProCreator Wifi MB, with 64GB Corsair Vengeance ram

T-Topology should be the standard in my opinion.

 

[Upgrayedd]

4 errors after 45mins. 2 errors in cycle 1 and 2 in cycle 2 so far in 4 different tests, test 1,3,12 and 14 using 8000 XMP1 cache49.
Cache auto gave me 2 errors after a few mins the first try so I switched to 49. I turned hyperthreading back on for baseline.

 

[fusseli]

4 errors after 45mins. 2 errors in cycle 1 and 2 in cycle 2 so far in 4 different tests, test 1,3,12 and 14. 8000XMP1 cache49.
Cache auto gave me 2 errors after a few mins the first try so I switched to 49. I turned hyperthreading back on for baseline.

Not terrible if it's only errors and no app crashes, freeze, or reboots. You might be able to crack 8000 stably.

 

[Upgrayedd]

Not terrible if it's only errors and no app crashes, freeze, or reboots. You might be able to crack 8000 stably.

50min in with HT off and no errors yet.
Edit: passed
Edit2: went for 8200 and have 1 error after the 1st cycle, roughly 40min.

 

[ShrimpBrime]

50min in with HT off and no errors yet.
Edit: passed
Edit2: went for 8200 and have 1 error after the 1st cycle, roughly 40min.

Nice to see some progressing results! 8200 baby!!

 

[Upgrayedd]

Nice to see some progressing results! 8200 baby!!

Had an error so Im going to try 8000Tweak with cache 49. If it works it'll probably be the daily.

I still want to try 8200 xmp1 again. I might drop the p cores a bump and see what happens.
Edit: 8000Tweak has 1 error through the entire run.

After disabling HT I don't believe Intel's APO project doesn't work anymore.

 

[InVasMani]

What's GPU-Z showing on the iGPU bandwidth at that MT/s memory speed!?

 

[Upgrayedd]

What's GPU-Z showing on the iGPU bandwidth at that MT/s memory speed!?

I must not have drivers installed for iGPU? I don't see an option for it in gpuz

 

[InVasMani]

I must not have drivers installed for iGPU? I don't see an option for it in gpuz

It should look something like this. I'm presuming you've got discrete GPU installed alongside it. I'm just mildly curious how it looks with faster memory.

 

[Upgrayedd]

It should look something like this. I'm presuming you've got discrete GPU installed alongside it. I'm just mildly curious how it looks with faster memory.

View attachment 327786View attachment 327787

I didn't realize you could install multiple gpu drivers? That doesn't conflict anywhere?

 

[ir_cow]

@InVasMani don't confuse the poor fellow lol.