Jump to content

Recommended Posts

Playing BF1 for 2hrs... you can see where it boosts above 4ghz (all core are at 4ghz when gaming and goes up from there at times) as the temp goes to 62-64c, then drops back to 54c. I turned my fans down to 1000rpm and pump is around 65%. Some false readings for the Max clock speed for sure. TMPIN0 is 20c below always vs Package temp. The Biostar comes with a heatsink for the m.2. Before I ran the m.2 with the heatsink off since I didn't know if it would work like last year. Noticed it ran 60c and up to 68c but after putting on the heatsink, it is sitting at 39c right now, so It's doing a great job. I have my games on my Vertex4 so not sure why it goes up, possibly from the gpu dumping heat. Keeping Ram at 3200 16-17-19-53 CR1.

temps.jpg

 

Edit:

 

Looks like just giving a bit more volts to the ram I can run 16-16-18-53 CR1 and get just under 70ns at 69.4 :) Only tested for an hr but will run it thru the longer tests soon:

 

3200 @ 16-16-18-53 CR1

read: 47176mb/s

write: 45045

copy: 40467

latency: 69.4ns

 

cachemem16_16_18_53.png

 

 

Edit:

18.5hrs

 

stabilitytest12.png

 

 

Edit:

 

Well it was just driving me crazy why nothing seemed to work for OC. I can see all the cores in light loads over 4100mhz. I was for some reason thinking llc was lvl 1 was the highest and it is actually lvl 6.. dunno what I was thinking. Few other things, before to get to 1.4v for the 1700x using adaptive at +.095, but for this chip it is +.205v which is much different. Other issues is that it does not down clock the volts or clocks at all, even with llc off. Still don't really fell it is worth it to OC since even at 1.408v it fails one worker p95 within 5mins @ 4.2ghz. The 1700x at 4ghz would downclock both volts and clocks... Probably my bios, will revisit on next bios update.

Edited by gamble

Share this post


Link to post
Share on other sites

congrats on getting under 70ns

i had to drop back to 3733 to get the above stable at 1.408v i tried setting it up to 1.4825vdimm but it didn't help also had similar instability in hci memtest at 3733c15 and 3733 1t

but i have been able to drop the tRFC down to 300

 

i think this cpu needs a delid

1.395v 4.7ghz hits 80-85c instantly under prime kill the load and it instantly drops back to 20-25c idle

 

the l1 cache and memory latency on amd is a bit slower but the l2 and l3 cache are impressive providing a fair bit more bandwidth at a similar latency to my overclocked 6700k

Edited by Dasa

Share this post


Link to post
Share on other sites

They have some cool delid kits now. I did the razor way, it was stressful and took me forever as I was scared I would hit the core :) Would you do direct die or put ihs back on?

 

https://www.rockitcool.com.au

Ebay has cheaper prices and different types. Maybe some grizzly. Haven't had a look but wondering if others with 6700k gained good temps from it.

Edited by gamble

Share this post


Link to post
Share on other sites

i ended up deliding my 3770k with vise hairdryer method after i found the gap wasn't big enough to get a blade in

had to smash the crap out of it with the hammer till it flew out of the vise ripping chunks out of the ihs and knocking one of the chips for the igp off the back

eventually got it with some minor damage to the pcb edges

i like to put the ihs back on these days its just safer and easier to make sure the cpu is getting good contact in the socket with lga

 

still have some liquid ultra left to go under there if i feel like pushing harder but i don't think i would gain much with this cpu

Share this post


Link to post
Share on other sites

Yea might not be worth it, whats the temps when stock with prime?

 

I contacted hci_memtest to see what they recommend for testing, below is their answer:

 

"100% catches the easy errors, 400% to catch intermittent errors."

 

Im testing hci_memtest 16-16-17-42 CR1 and seems to be good @ 3200 with 1.47v. This didnt really improve my scores though and latency is nearly the same but Ryzen really like Ram speed it seems. Even though during games like bf1 my cores are usually around 4ghz (I ran 1700x at 4ghz on all cores) and boost up some from there, it just feels smoother. I dont know why but there is less micro stutters and such.

Edited by gamble

Share this post


Link to post
Share on other sites

For hci i find 400% is the minimum when running 8 instances at once with ~1777mb each so that all threads are covered and all ram is used I like to leave it run overnight so that each has 1200 coverage just to be sure

but it will usually find any bad errors in the first 50-200%

I find hci is very heavy on the cache and is great for finding errors related to the memory controller but a bit slower than linux gsat for finding memory related errors

  • Like 1

Share this post


Link to post
Share on other sites

I don't want to be 'that guy' gamble... but having a biostar board you're already at a performance disadvantage.

 

Sorry if you're a biostar fan.

Share this post


Link to post
Share on other sites

nah, I agree with you! It was the only board I could get when Ryzen was first released, there was a huge shortage with weeks of wait time on mb's. They gave a free m.2 sata 240g drive with it, good price and had really good vrm setup. It's the bios releases that is a pain when you hear of other mb getting releases quicker. It's a good board just I really like Asrock for price/performance. Also, at the time my old system was dead and was already waiting for AMD release as I wanted to go back to AMD. I kinda wish I waited but oh well, all is running good now after tweaking ram and really the 2600x is a good chip all around. I really cant wait till next year's Ryzen refresh, might upgrade my mb too then :)

  • Like 1

Share this post


Link to post
Share on other sites

I really cant wait till next year's Ryzen refresh, might upgrade my mb too then :)

Hard to believe that we're already staring down at the 7nm Zen 2 hey!

Share this post


Link to post
Share on other sites

Bad idea they spend all the R&D then sell it to Chinese for a pittance.

Who could use it for spying, hacking, influence meddling and other things that push their agenda

Share this post


Link to post
Share on other sites

AMD’s Ryzen desktop CPUs for 2019 may double the core count

 

Although we’re still waiting for AMD to release new low-end second-generation Ryzen processors based on its refreshed 12nm Zen design (aka Zen+), AMD appears to be betting big on 2019. AMD said it would sample Epyc “Rome” chips with its 7nm Zen 2 design in the server market in late 2018, but mainstream processors based on that design won’t appear until next year. If current rumors are correct, those Ryzen 3000 Series chips will sport up to 16 cores.

 

According to reports stemming from Chinese forums, the new Zen 2 architecture provides 10 to 15 percent improved Instructions Per Cycle, meaning the design can handle more instructions than the previous design. The current Zen+ architecture sees a three percent Instructions Per Cycle increase over the original Zen design introduced in AMD’s first-generation Ryzen processors during 2017.

 

But the big news here is that AMD already finalized its Zen 2 design to support up to 16 cores in the mainstream market (AM4), up to 32 cores in the high-end enthusiast market (TR4), and up to 64 cores in the server market (SP3). For the mainstream and server markets, that is double the current maximum core count while the maximum Threadripper core count apparently remains unchanged.

 

The rumors dig a little deeper on a technical level, reporting that AMD actually finalized two Zen 2 designs. Both rely on how AMD groups its processor cores together into what it calls Core Complex, or CCX, that share the same cache. CCX are connected together using AMD’s Infinity Fabric on a single die.

 

In one Zen 2 design, AMD supposedly uses a CCX with six cores and pairs two CCX together on a single die, providing a maximum core count of 48 in a single processor. In another Zen 2 design, AMD uses a CCX with eight cores, and pairs two CCX together on a single die, providing a maximum core count of 64 in a single processor.

 

That’s a lot of tech talk, but it shows where AMD is heading with its third-generation Ryzen processors slated for 2019. The good news is that, technically, you won’t need to swap out motherboards to make way for the new chips, as AMD said the current processor seats/sockets will support all Ryzen and Epyc chips at least until 2020.

 

Of course, the arrival of new processors ushers in new motherboard chipsets that support all the new features in the latest CPUs. Cramming a third-generation Ryzen chip into a first-generation Ryzen motherboard won’t produce all the benefits and improvements introduced in the Zen 2 architecture.

 

Processors slated for a 2019 arrival will be the Ryzen 3000 “Matisse” Series for desktop (AM4), the Ryzen Threadripper 3000 “Castle Peak” Series for enthusiasts (TR4) and the Epic “Rome” chips for servers (SP3). A slide from AMD shows the company’s Zen 3 design is on track for 2020 although that will likely be toward the end of the year in samples dished out to servers. Mainstream and enthusiast products for 2020 may rely on a refreshed Zen 2 design based on the 7nm+ process node.

  • Like 1

Share this post


Link to post
Share on other sites

I was looking a mb and found that Asrock had so many vrm's but mb companies lie? It seems that people have been digging thru them and calling them out... its interesting but here's some info:

 

 

https://www.hardwareluxx.de/community/f12/pga-am4-mainboard-vrm-liste-1155146.html

 

 

I admit I need to research this more but it's interesting.

Edited by gamble

Share this post


Link to post
Share on other sites

I thought it was interesting.... There is some longer videos out there for you if you want to watch them :)

 

 

This one is pretty good, same guy but more details on vrm and layout.

Edited by gamble
  • Like 1

Share this post


Link to post
Share on other sites

lol I wasn't rapping on you for posting that video... I was voicing how that long a video was grieving my short attention span and patience.

Share this post


Link to post
Share on other sites

 

On 7/28/2018 at 5:04 AM, SceptreCore said:

lol I wasn't rapping on you for posting that video... I was voicing how that long a video was grieving my short attention span and patience.

I was having a bit of fun ?

 

This is interesting, looks to be quite powerful:

https://www.pcgamer.com/amd-built-a-powerful-semi-custom-ryzen-vega-soc-for-game-consoles-and-pcs/

"The SOC sports an 4-core/-8-thread Ryzen processor running at 3GHz, along with 24 Vega CUs clocked at 1.3GHz for graphics. The GPU basically amounts to a Radeon RX 580, and is perhaps even a bit more powerful."

Edited by gamble

Share this post


Link to post
Share on other sites
On 8/4/2018 at 9:53 PM, gamble said:

 

I was having a bit of fun ?

 

This is interesting, looks to be quite powerful:

https://www.pcgamer.com/amd-built-a-powerful-semi-custom-ryzen-vega-soc-for-game-consoles-and-pcs/

"The SOC sports an 4-core/-8-thread Ryzen processor running at 3GHz, along with 24 Vega CUs clocked at 1.3GHz for graphics. The GPU basically amounts to a Radeon RX 580, and is perhaps even a bit more powerful."

Is there even good money in this for AMD?

This all seems to be happening at the expense of the graphics division which will soon have competition from the Intel juggernaut 

Share this post


Link to post
Share on other sites

If no one uses it, then it's backfired. If it's adopted then the gamble paid off. They might have to undercut Intel to get the business though.

Share this post


Link to post
Share on other sites

Glad there's 2 true players... Im hoping AMD keeps it up but supposed Intel new chips with solder, more cores and high clocks will keep it interesting. 

  • Like 1

Share this post


Link to post
Share on other sites
On 8/9/2018 at 11:11 AM, gamble said:

Glad there's 2 true players... Im hoping AMD keeps it up but supposed Intel new chips with solder, more cores and high clocks will keep it interesting. 

I think things look good for AMD

Intel is bring forward the reheated Coffee chips to Q3 this year instead of Q1 next year. I'm guessing they had to do this because reviewers would recommend holding off until Q2 which is usually when AMD launches the next Ryzen desktop which in this case will be Zen 2. So getting these 9000 series out sooner makes them look less old when these 14nm chips square off against the 7nm ones. Intel has the power, the resources, the money to keep up, but AMD seems to just be running away with their roadmap and Intel's 10nm has cost them dearly.

I never believed in Hector De Ruiz's plan to go fabless... but it seems like this forethought has paid off!

  • Thanks 1

Share this post


Link to post
Share on other sites

Apparently Intel 10nm will be comparable to Foundry 7nm. Which we knew, even AMD recognized this in its slides a while back. AMD though will be first to market with the new node.

Share this post


Link to post
Share on other sites

Intel have still have the OEM sales, AMD need to start putting Zen into OEM machines as you can't just rely on gamers and uber-system builders. The big money is in the servers though.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×