X370 Future Memory Support
Printed From: ASRock.com
Category: Technical Support
Forum Name: AMD Motherboards
Forum Description: Question about ASRock AMD motherboards
URL: https://forum.asrock.com/forum_posts.asp?TID=4713
Printed Date: 22 Dec 2024 at 3:07pm
Software Version: Web Wiz Forums 12.04 - http://www.webwizforums.com
Topic: X370 Future Memory Support
Posted By: twf85
Subject: X370 Future Memory Support
Date Posted: 29 Mar 2017 at 12:57am
|
" rel="nofollow - What do you all think the likelihood of 4 sticks running at their highest rated speed will be during the lifetime of this chipset? I'm trying to decide if I should fork out the extra cash for 32gb vs 16gb. If anyone has read something that leads them to believe that we will eventually see all four slots running at their max speeds, I'd be happy to save the money. ------------- https://valid.x86.fr/qc9lr0" rel="nofollow"> 
|
Replies:
Posted By: wardog
Date Posted: 29 Mar 2017 at 1:38am
|
" rel="nofollow - I just finished replying to you in another Thread ! http://forum.asrock.com/forum_posts.asp?TID=4668&PID=26214&title=corsair-memory-on-x370-gaming-k4#26214 Let's keep this discussion, Yours after all, here. You're CPU-z Validation? Is it current? I thought I remember you say you already were using 32GB. |
Posted By: twf85
Date Posted: 29 Mar 2017 at 4:07am
|
I guess I went off topic in the other thread. In the other I was mostly interested in how-to edit the CM values for RAM.. Reply from other thread..
Right now I'm at 2x8GB, but seeing as there appears to be no way to attempt the adjustment of the CM to push these any higher, I am now shopping around.. I can't remember which exactly, but I know one of these sets successfully ran at 3200: https://www.newegg.com/Product/Product.aspx%3FItem=N82E16820231929" rel="nofollow - G.SKILL TridentZ Series 16GB (2 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200 (PC4 25600) Intel Z170 Platform / Intel X99 Platform Desktop Memory Model F4-3200C16D-16GTZB https://www.newegg.com/Product/Product.aspx%3FItem=N82E16820232206" rel="nofollow - G.SKILL TridentZ Series 32GB (2 x 16GB) 288-Pin DDR4 SDRAM DDR4 3200 (PC4 25600) Desktop Memory Model F4-3200C14D-32GTZ I'm on the fence as to which to buy. Obviously the slightly lower timings on the more expensive set, combined with more memory in only two slots, would be ideal. I've been slowly talking myself into buying the pricier set throughout the past couple of days, but I don't know that I'll ever need 64gb of RAM on this system. Sure it would be nice, but on the other hand, upgrading from 16 to 32 at some point later on this year would be the more economical route. I don't know that I'd notice the decrease in latency that much, but then again, I've never really chased low timings, so I have no idea. Maybe I would notice it. I'm running out of time for the return window on https://www.newegg.com/Product/Product.aspx%3FItem=N82E16820231875" rel="nofollow - the sticks I bought initially , and I would really like to not get stuck with them. I was impulsive when I opted for them, and after doing more research, they strike me as sort of "middle-of-the-road" sticks. Kills me because I had some CORSAIR sticks lined up before I bought, and while those were about 30-40 more, I believe they ended up being on the QVL. TL:DR, I don't want to repeat the same mistake of buying the wrong memory, and I don't want to overspend for something that I may not realize any benefit from. If the 4-slot penalty may possibly go away in the future, like the 3200 limit is supposed to be lifted at some point, then I would have a tough time opting for 32gb at this time. EDIT: https://www.youtube.com/watch%3Fv=VjGWqTFultI" rel="nofollow - Level1Techs YouTube Video Review of the Taichi (shows compatibility with the 32gb set above) EDIT2: I guess the 16gb set isn't listed on the QVL, but the 32gb set isn't specifically listed there either.. ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
wardog wrote:Posted By: wardog
Date Posted: 29 Mar 2017 at 7:12am
|
" rel="nofollow - Yea. Get that Kit you have now back to NewEgg. I'm almost surprised they run on AM4, having only X99 XMP Profile(s). Have you seen: http://forum.asrock.com/forum_posts.asp?TID=4565&title=am4-ryzen-specific-memory Just yesterday I added: http://rymem.vraith.com/ That will give you maybe a better Idea. Seems they harvest Ryzen only CPU-z Validation results. Me? ASRock is shipping us Mods our boards soon and seeings I have a gaggle of DDR4 here, and the smarts to "work' them, I'm gonna try with what I already have first. Looks good. X370 Taichi Certified: G.Skill Flare X 16GB (2x8GB) Model F4-3200C14D-16GFX http://www.gskill.com/en/configurator?manu=52&chip=2952&model=2985 A wee WEE bit pricey, but gotta LOVE 14-14-14-34 @ 3200 https://www.newegg.com/Product/Product.aspx?Item=N82E16820232530 |
Posted By: parsec
Date Posted: 29 Mar 2017 at 11:35am
|
I know you'd like to get the memory now, but Ryzen is so new and the memory manufactures have barely had time to provide Ryzen specific DDR4 memory. Part of the reason for that is the memory speed potential of the Ryzen memory controller is still being discovered. AMD is still providing the information for UEFI/BIOS updates for supporting higher speed memory. There is one such update coming pretty soon, and who knows how many more will come after that. The only company that is currently marketing Ryzen specific memory is G.SKILL with their Flare line. The others will follow soon with their own. You'll have a much better selection in the near future, with higher speeds and capacities. If Ryzen follows the trend of Intel's latest processors, the greater the capacity of memory used (32GB+), the more difficult it is for it to run at higher speeds. We can see that general trend now by checking the Memory QVL of any ASRock AM4 board. Currently at DDR4 3000 and above, we only find 4GB and 8GB DIMMs, and only two may be used. You mentioned buying a kit of memory now, and then another later. Mixing kits of even the same memory model is risky. Sometimes the memory chips that the memory manufactures buy from other companies have slight changes in them from different production runs, that prevents the two kits from working together at their highest possible speed. The longer the time between buying two or more kits, the more likely they may use different memory chips and have less than optimal compatibility. ------------- http://valid.x86.fr/48rujh" rel="nofollow"> 
|
Posted By: twf85
Date Posted: 29 Mar 2017 at 2:15pm
I had not seen that, but thank heavens! Finally a page that can definitively answer the question. Did you code that?
Congrats! That's really cool of ASRock to show some love to the people keeping this place running ;)
I had seen the Flare line, but I wasn't too thrilled about it. I'm happy they were so quick to release them, but the design of the sticks makes them look kind of second-rate. I think when I first saw them, they didn't have the high speeds they now, so I didn't give them a second glance. I don't know, maybe I'm letting my first impression of them get the better of me. In your experience is there much of a difference in 3000 vs 3200?
Your encyclopedic knowledge never ceases to amaze me, parsec ;) I did not know that about mem kits.. sounds like the same problem I've run into trying to chase down PCB's / heads for hard drives ;) I guess I've never actually added memory to a system. I've always just replaced it, so that's very useful info. Alright, so I guess the smart move is to go with a decent pair of 16gb now, something that won't break the bank while still achieving near top-shelf speeds. According to that rymem page, this pair ought to do the trick: https://www.newegg.com/Product/Product.aspx%3Fitem=N82E16820233852" rel="nofollow - CORSAIR Vengeance LPX 16GB (2 x 8GB) 288-Pin DDR4 SDRAM DDR4 3000 (PC4 24000) Memory Kit Model CMK16GX4M2B3000C15 I can use those for awhile, and then upgrade when the chipset has had more time to mature and there are more verifiable options. I could use them later in a server or something once (if ever) they come out with sticks running at 3600. Thoughts? Also, in regards to running 4 slots at full speed at some point in the future.. Is that how Intel systems work as well? I've never put together an Intel system where high speed was a primary concern, just workstations, so I never paid any attention to any of the specifics regarding RAM. It definitely struck me as odd seeing that kind of limitation when official Ryzen pricing/availability dropped. I've been very busy lately, and I haven't had as much time to keep up with Ryzen news. If anyone has any news regarding the future plans of AMD and this specific limitation, I'd be very interested to hear it. ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
Posted By: wardog
Date Posted: 29 Mar 2017 at 6:58pm
You're funny. Did I code that? Heck no. Moi'? Code? hahaha Hell, no. Why .... Every time I had to express my Morse Code Efficiency for my ARRL Licenses it took me three or four times before I passed. And it hasn't gotten any better as I age  Me code that he asks ................. |
Posted By: KrisWragg
Date Posted: 29 Mar 2017 at 8:53pm
I have the same kit on my X370 Taichi, I have managed 2666mhz 14-15-15-30 @ 1.375v and SOC set to 1.1v I can also run 2993mhz 18-20-20-40 but that isn't really any faster in benchmarks so I've stuck with the 2666 lower timings for now. I think the biggest thing is to change the SOC voltage higher, I've found 1.1 quite good, I think 1.2v is the highest recommended for daily use.
|
Posted By: redemp
Date Posted: 29 Mar 2017 at 10:21pm
| I updated my bios to 1.93D today on my X370 Professional Gaming but my RipJaws V 3200cl14 8x2VGR still fail @3200mhz :(. |
Posted By: twf85
Date Posted: 30 Mar 2017 at 1:28am
" rel="nofollow -
After reviewing that RyMem thing again, it looks like the many of the timings had to be adjusted to achieve the speeds they were rated at. Given what KrisWragg just said, I think I'm gonna' go with the Flare set you mentioned. ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
Posted By: twf85
Date Posted: 30 Mar 2017 at 2:05am
" rel="nofollow -
Welcome to the Forums! Which BIOS are you running?
And a Welcome to the Forums to you, too! Sorry to hear that about your sticks.. Did you buy them for your Ryzen build, or did you already have them? How stable is 1.93D? ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
Posted By: juanvr
Date Posted: 30 Mar 2017 at 3:42am
Yeah, I managed to get it down to 16-16-16-30 on 2667. Might be able to tweak it more, but I'm "ok" with that until proper support becomes available. I just hope it actually does. That said, the Ryzen 1700 seems to be running good at 3600 with SMT and stock cooler.
|
Posted By: juanvr
Date Posted: 30 Mar 2017 at 3:45am
" rel="nofollow -
Running the latest Beta, 1.63 according to the product page. XMP profile for this RAM kit is there and gets detected, but system will not boot just selecting that. I got it running by selecting XMP profile, then adjusting speed down to 2666.
|
Posted By: twf85
Date Posted: 30 Mar 2017 at 6:01am
" rel="nofollow -
Not that I would urge you in this direction, but isn't the latest Beta for the Taichi 1.93A? The BIOS list on the Fatal1ty was a little screwy when 1.60 came out. I saw it in the morning, downloaded it, installed it, and then it was gone from the list. Made me think I had just made a horrible mistake. Anyway, I guess my point is that if you refresh, or open a new tab, you should see 1.93A. The note on it is, "Improve XMP DRAM compatibility.". Pretty big jump from ~1.6 to ~1.9, so who knows, you might have some luck. ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
Posted By: juanvr
Date Posted: 30 Mar 2017 at 6:15am
" rel="nofollow -
Wait, so I can use the Taichi BIOS on the Fatal1ty? Or you mean the list on the Fatal1ty product page is out of date?
|
Posted By: juanvr
Date Posted: 30 Mar 2017 at 6:19am
|
" rel="nofollow - Guys, just noticed something... What's the difference between these boards?! http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Professional%20Gaming/index.asp#BIOS http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Gaming%20K4/index.asp#BIOS I think I have the K4 but there's a later BIOS for the "Professional Gaming" board. |
Posted By: parsec
Date Posted: 30 Mar 2017 at 10:09am
You cannot use the UEFI from one board with another. Even if you tried, the update should be stopped before it begins. Applying the UEFI for one board to another would result in at best a broken UEFI, and at a worst a bricked board. BTW guys, there are two Ryzen Fatal1ty X370 boards (Professional Gaming, and Gaming K4), and three Ryzen Fatal1ty models, including the Fatal1ty AB350 Gaming K4. I know typing these names is a pain, but if you don't fully identify the board you're using, we can't give you accurate information. As of the evening of 3/29/2017 in the USA, I see both Ryzen Fatal1ty x370 boards have a 1.60 UEFI version, so 1.60 was not removed from either board's download page: http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Gaming%20K4/index.asp#BIOS" rel="nofollow - http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Gaming%20K4/index.asp#BIOS http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Professional%20Gaming/index.asp#BIOS" rel="nofollow - http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Professional%20Gaming/index.asp#BIOS UEFI updates are coming quickly, and I see a 1.94A Beta version for the Taichi just posted: http://www.asrock.com/mb/AMD/X370%20Taichi/index.asp#BIOS" rel="nofollow - http://www.asrock.com/mb/AMD/X370%20Taichi/index.asp#BIOS The Fatal1ty X370 Professional Gaming has the 1.93D Beta version available, as stated above. Not that this will make you feel any better I imagine, but I was reading some user comments about other manufacture's Ryzen boards. Many of them include the lack of DRAM configuration features in the UEFI/BIOS, so that seems not uncommon. It sure seems the Ryzen processors were available for purchase before the mother board and memory manufactures were ready with their products. ------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: juanvr
Date Posted: 30 Mar 2017 at 10:32am
Thanks, I don't understand...
You said the K4 board is Fatal1ty AB350 Gaming K4... However My board is definitely an "X370 Gaming K4". And your product page has it as an X370 board? It's just weird that there are 2 Fatal1ty X370 boards, one with M.2 wifi, one without... And they're on different BIOS specs. Also, this means there are THREE X370 boards: http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Gaming%20K4/index.asp#BIOS" rel="nofollow - http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Gaming%20K4/index.asp#BIOS http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Professional%20Gaming/index.asp#BIOS" rel="nofollow - http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Professional%20Gaming/index.asp#BIOS http://www.asrock.com/mb/AMD/X370%20Taichi/index.asp#BIOS" rel="nofollow - http://www.asrock.com/mb/AMD/X370%20Taichi/index.asp#BIOS |
Posted By: twf85
Date Posted: 30 Mar 2017 at 11:43am
" rel="nofollow -
My bad juan, got you mixed up with the other guy.
There are TWO Fatal1ty boards with "K4" in the name.. http://www.asrock.com/mb/AMD/Fatal1ty%20AB350%20Gaming%20K4/index.us.asp" rel="nofollow - Fatal1ty AB350 Gaming K4 http://www.asrock.com/mb/AMD/Fatal1ty%20X370%20Gaming%20K4/index.us.asp" rel="nofollow - Fatal1ty X370 Gaming K4 Fatal1ty is a famous gamer who has been around for ages. ASRock, in partnership with Fatal1ty, releases boards with his name attached on many different boards. You got the pricier of the two K4 boards, so don't sweat it. ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
Posted By: juanvr
Date Posted: 30 Mar 2017 at 12:37pm
Ahh I get it now. At the moment I'm pretty happy with the setup I have. Would just be nice to have the RAM run at the XMP speeds. The CPU is happy at 3600 with no voltage tweaks. Ran a CPU-Z validation which is in the signature now. ------------- https://valid.x86.fr/8pc7db" rel="nofollow"> 
|
Posted By: juanvr
Date Posted: 30 Mar 2017 at 5:05pm
|
FYI, the X370 gaming K4 just got v1.94A. Flashed it, still the same result. XMP Profile refuses to boot (boot loop into protection), dropping to 2993, same thing with longer wait times before reboots. Back to 16-16-16-30 @ 2667MHz. Again, model number is: CMK16GX4M2B3200C16 according to CPU-Z ------------- https://valid.x86.fr/8pc7db" rel="nofollow">
|
Posted By: wardog
Date Posted: 30 Mar 2017 at 6:48pm
That XMP on these sticks fails to boot comes as no surprise to me. CMK16GX4M2B3200C16's XMP Profile(s) contains only data to be configured when used on an X99 board. And only X99 boards. X99 boards have a definitive short list of memory that works on such a finicky platform. BIG diff! |
Posted By: twf85
Date Posted: 31 Mar 2017 at 3:00am
I'm in the same boat. Sluggish RAM. I think the vast majority of early adopters probably are. I'm thankful the system has been rock solid thus far. Running hotter than I would like, but then I only have a 120mm radiator (due to case size limitation). ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
Posted By: juanvr
Date Posted: 31 Mar 2017 at 3:13am
Yeah, I'm just hoping support comes eventually for them. I would have taken G.Skill Flare-X 3200s but they were out of stock down here in New Zealand, and no-one had any sort of date. So I had to go with whatever was on the compatibility lists. My mistake was trusting "Corsair Vengeance LPX" as "ALL Corsair Vengeance LPX kits". ------------- https://valid.x86.fr/8pc7db" rel="nofollow">
|
Posted By: parsec
Date Posted: 31 Mar 2017 at 8:46am
@wardog, Sigh, you can try to explain things... juanvr, the DDR4 XMP profile data for Intel X99 processor memory controllers won't work with the Ryzen memory controller. The memory controllers are different, and some of the memory settings are different. When using DDR4 memory designed for other systems with a Ryzen system, the memory timings must be set manually. How can the Ryzen memory controller, a piece of hardware that is part of the CPU, be changed by a UEFI/BIOS update to support the XMP data/settings for the memory controller of Intel Haswell-E and Broadwell-E processors? The good news is at least your Corsair memory will work with a Ryzen system. All modern memory also contains multiple sets of basic timing data for different non-over clocked supported speeds of a model of memory. That data is referred to as the JEDEC or SPD data, which you can view in the CPU-Z SPD tab. One set of that data is the memory timing data used for the default speed of your memory, most likely 2133. The SPD data allows a mother board's UEFI/BIOS to automatically set the timing parameters for our memory. XMP data is simply another convenience for setting memory timings to the correct values for the "EXtreme Memory Performance" speed of memory. That is the over clocked speed of your memory, which is any speed beyond the specifications of the memory controller in your CPU. Unfortunately, we've become used to using it, and the memory manufactures don't provide any other information about the timing settings for their products. If you cannot set memory parameters manually, you'll simply need to wait for Ryzen specific memory to be available. Of course a valid question is, what are the correct timing settings to be used with the Ryzen memory controller with your model of memory? That is something that Corsair should provide for you. If they tell you to use the XMP profile, you know that does not work, and is a wrong answer. If they tell you to wait for mother board BIOS support, they are putting you off, and just hoping the mother board manufacture will do their work for them. Here are some example timings for your memory that was CPU-Z verified on a good sized set of Ryzen boards. Note that setting the Command Rate (CR) to 1T is needed for achieving higher speeds, which is something that will never be found in DDR4 XMP data for Intel processors. Also note, reaching 3200 is very rare (so far), and if you are at 2666 now, you are above average. http://rymem.vraith.com/detailed_ram/715" rel="nofollow - http://rymem.vraith.com/detailed_ram/715 ------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: wardog
Date Posted: 31 Mar 2017 at 9:26am
" rel="nofollow -
You did a fine job there explaining it. I might have added some golden nuggets of my own in there yet the point is moot considering that kit only has X99 XMP Profile(s). And he seemed intent on using/setting XMP in the BIOS. XMP=NO , manually=maybe. By his results the best I can say is "Hey.... They're working". No offense meant juanvr. Just calling 'em as I know 'em. In the future juanvr, look at the Sticky list and links I created for AM4/Ryzen Specific Memory: http://forum.asrock.com/forum_posts.asp?TID=4565&title=am4-ryzen-specific-memory |
Posted By: twf85
Date Posted: 31 Mar 2017 at 2:38pm
|
The Flare X sticks came off of Back Order, so I bought some. Probably won't be here until Monday or Tuesday, but fingers crossed they work as well as they're supposed to ;) ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
Posted By: wardog
Date Posted: 31 Mar 2017 at 6:14pm
|
My X370 Killer SLI/ac will be here next Tuesday, along with a 1700X. Get ready for some Ryzen memory tricks boys and girls. Zennn........ |
Posted By: debaser
Date Posted: 31 Mar 2017 at 7:36pm
Thank you, I own an X370 Killer SLI and with latest bios 2.0 can't get past 2933MHz with 2x8GB Samsung B-Die sticks, really looking forward to those memory tricks  My sticks are 3600 CL17 and can boot to windows at 3200 and can surf the web, but get errors on stress tests.
|
Posted By: Nemesis.ie
Date Posted: 02 Apr 2017 at 12:03am
|
" rel="nofollow - Speaking of RAM. I just updated the UEFI on my AsRock X370 Gaming Pro from 1.60 to 1.93D. It cleans up the interface a lot and adds in profile saving and such which is nice to see, however, here's the bad news: Per AIDA64 @ ~3900MHz OC on the CPU, RAM (3733Mhz Team TF @ 3200MHz) bandwidth in 1.93D is down to ~44,000MB/sec versus the ~51,000MB/sec in 1.60. ~Asrock: can we have the speed back please? If this was changed for compatibility, maybe open the settings used so the default is compatibility/slow but those of us with modules than can handle it can get the extra speed. Thanks!
|
Posted By: wardog
Date Posted: 02 Apr 2017 at 12:14am
Can you post pics of AIDA of 1.60 and 1.93D? Too, any chance you created and saved a Report.txt file from CPU-z of 1.60 and 1.93D? |
Posted By: Nemesis.ie
Date Posted: 02 Apr 2017 at 12:27am
|
Hi Wardog, I have pics and can post later/tomorrow.  |
Posted By: wardog
Date Posted: 02 Apr 2017 at 12:28am
Awesome. Thanks. |
Posted By: Nemesis.ie
Date Posted: 02 Apr 2017 at 2:07am
|
" rel="nofollow - AIDA64 RAM speed UEFI 1.6 vs 1.93D. 1.6: https://drive.google.com/file/d/0B8chhVGt74cUUFJwVHZNa0oyTU0/view?usp=sharing 1.93D: https://drive.google.com/file/d/0B8chhVGt74cUYlZSVDF1TlllcDA/view Even with tighter timings the 1.93D is a lot slower (1.6 supports the 16-20-20-39 too). |
Posted By: debaser
Date Posted: 02 Apr 2017 at 3:59am
| I noticed a decrease in memory bandwidth too with 1.93a beta and 2.0 bios for the Killer SLI :( |
Posted By: Nemesis.ie
Date Posted: 02 Apr 2017 at 4:48am
|
" rel="nofollow - As an aside, CPU-Z is now reporting the wrong voltage since 1.93D, it shows a CPU voltage of 2.6v. Were some the the I2C sensors swapped around I wonder? |
Posted By: wardog
Date Posted: 02 Apr 2017 at 8:12am
Here's your speed. Remember, Ryzen turns some things that we knew upside down. 1.60 Timings 18 tCAS, 20 tRC, 20 tRP, 39 tRAS, 1 T/N/CR 1.93D Timings 16 tCAS, 20 tRC, 20 tRP, 39 tRAS, 1 T/N/CR Ryzen loves tCAS tRP and tRP to be the same latency. Thus above we can see that with the CAS of 18, nearing that magical matching tCAS tRC tRP latencies. On 1.60(18 tCAS) notice too your L3 Cahe jump right up and say Hello! Lastly, concerning the synthetic bench of AIDA, don't devote too much time with the Memory row of Read, Write and Copy values. Here's my Z170 OC Formula with an 17-6700K  Would you call that slow as molasses, or Holy sh*t Batman? |
Posted By: Nemesis.ie
Date Posted: 02 Apr 2017 at 4:48pm
|
" rel="nofollow - [URL=][/URL]I meant to add a note about the timings on the previous post. I have used 16-20-20-20 and 18- etc. as well, ie the same settings for both UEFIs and there is still that large 45 vs 51GB/s difference between the two versions irrespective of the timings. I only had those pics of 1.60 handy. If you want, I can grab a pic of 1.93D with the 18-20-20-39 (which is actually the XMP for 3733) and try at 18-18-18-18 (39)? I found I have this one with the CL18 settings but it is at 4025, you can see the large gap from UEFI and then a bit of extra speed from the higher CPU clock. :) https://drive.google.com/file/d/0B8chhVGt74cURm1ic1RyZHdoYms/view?usp=sharing I only submitted my reports to FinalWire, so none saved locally. |
Posted By: wardog
Date Posted: 02 Apr 2017 at 4:50pm
Yea. Both please. |
Posted By: Nemesis.ie
Date Posted: 02 Apr 2017 at 5:30pm
|
" rel="nofollow - Hi, Those RAM speeds you posted look pretty impressive compared to my Z97 4790k @ 4.8GHz with 2400MHz RAM. Back to the Ryzen. Here are 5 tests at different timings @ 4025MHz for comparison with the 4025MHz 1.60 one I posted above. None of the settings close the >15% gap versus the 1.60 results so something does seems to have changed. https://drive.google.com/file/d/0B8chhVGt74cUNG5oMFdSZWlTODg/view?usp=sharing |
Posted By: wardog
Date Posted: 02 Apr 2017 at 6:09pm
|
" rel="nofollow - You edited out the mem kit you're using? Or am I loosing it? |
Posted By: Nemesis.ie
Date Posted: 02 Apr 2017 at 6:40pm
|
I didn't edit anything intentionally, but the mouse might have slipped (wrist surgery recently) and my eyesight is problematic so who knows! :) Or maybe you are losing it. :) Team TF 3,733MHz 18-20-20-39 XMP profile/spec. 1.2 - 1.4v, 1.35v for XMP profile. http://www.teamgroupinc.com/en/product/T-Force%20Gaming/Gaming%20Memory%20Modules/DDR4%20MEMORY%20MODULES/XTREEM%20DDR4" rel="nofollow - http://www.teamgroupinc.com/en/product/T-Force%20Gaming/Gaming%20Memory%20Modules/DDR4%20MEMORY%20MODULES/XTREEM%20DDR4 |
Posted By: twf85
Date Posted: 03 Apr 2017 at 12:32am
|
" rel="nofollow - Nemesis.ie, what cooling solution are you using? ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
Posted By: Nemesis.ie
Date Posted: 03 Apr 2017 at 12:42am
|
A custom air cooler as the mounting bracket for my Kraken X62 is still MIA. It's a Thermal take AM3 all coppper finned block with an 80 to 120mm adapter funnel on the top with a 38mm x 120mm (i.e. thick) 2500 rpm beast sucking the air through. It's heating the peak of acceptable at ~4025 MHz. The X62 should be massively better and not sound like a jet at take-off! It's still outperforming my re-lidded Devil's canyon @ 4.8GHz with 3 x rads in a custom loop though, the DC hits 85c under load with 1 x 280mm, 1 x 360mm and 1 x extra thick 140mm fed by a D5 pump. |
Posted By: twf85
Date Posted: 03 Apr 2017 at 2:30pm
|
" rel="nofollow - What temps are you getting at idle? ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
Posted By: Nemesis.ie
Date Posted: 03 Apr 2017 at 2:37pm
|
" rel="nofollow - [URL=][/URL]Hard to say, per AIDA 64, the M/B and e.g. NVME drive are around 35 (22ish ambient). For the CPU, AIDA reports CPU (presumably the sensor in the socket) and also CPU diode, likely one of the sensors internal to the chip. I believe the socket sensor is fairly accurate and idles around 30-40, the diode on the other hand is likely the one with the 20c offset, it idles around the same as the socket by rapidly goes strospheric under load, at ~4GHz with the fan @100% its around 95c, with the fan on the standard profile it can hit 105+. I believe the real temperature is more like 85 at this point (based on the finger on the heatsink test and the mentioned offset). With luck UEFI and software updates will improve the accuracy. What are you seeing? |
Posted By: twf85
Date Posted: 05 Apr 2017 at 4:27am
|
I normally use CoreTemp (v1.7), but they have yet to release an update for Ryzen. In Ryzen Master, my stock idle is usually around 53c. I think it may have once dipped below 50 for a second or two, but it never stayed there while I was monitoring it. I boosted it up to 4Ghz, and it raised the idle temp about 10c, and that seemed like it was too high for idle. Boosted to 4.2Ghz (or perhaps 4.4, I can never remember now) on my 8350, the idle temps were around 30-40, so seeing ~60c was a bit of a shocker. Mind you, I only have a 120mm radiator, so that could be part of it. My new memory should arrive sometime this afternoon, and I'm going to swap in a slightly beefier fan to see if that helps with the temps. ------------- https://valid.x86.fr/qc9lr0" rel="nofollow">
|
Posted By: Nemesis.ie
Date Posted: 05 Apr 2017 at 2:37pm
|
" rel="nofollow - The Ryzen master temp does appear to have the offset, so if you have an X cpu it's likely arouind 20-22c too high. I'd like to know what the UEFI temp cut off is looking at though as I've had shutdowns that appear to be temp based (instant power down) when per this the diode tempp should be arouund 80 and I suspect it shoudln't be triggered until 85+. This should be a non-issue once I get the Kraken on the job though. /off-topic. :) Anyone any news on when the nect UEFI release will appear and if the 1.93D/1.94 includes the new AGESA 1.0.0.4 with the fixes? It came out right around AMD's announcement and does seem to have improved the latency (at the expense of nadwidth. I'm actually hoping it doesn't so the next one should be anoter nice improvement. :) Regarding the RAM, in real-world use, the improvement latency does seems to be as good/better than having the extra bandwidth. Both would be nice though. I'm also seeing GPU-Z single core bench speeds @4GHz (2300ish) which is similar to the 7700k stock. That implies Ryzen (for that workload) has a better IPC as the 7700k is running at 4.2GHz at least, but possibly 4.5. Anyway, back to the RAM chat. |
Posted By: wardog
Date Posted: 05 Apr 2017 at 3:09pm
|
Latency, above bandwidth. Now, that can be jiggered to be somewhat a loose statement depending on what latency and freq mem kit we have in the future above 3200. Yet now latency rules. Currently, Bandwidth will follow latency up the curve. |
Posted By: Nemesis.ie
Date Posted: 05 Apr 2017 at 3:15pm
|
Agreed. Tuning is tuning. :) It's the real-world performance that matters irrespective of what is adjusted to get there. |