9800x3D build planning

[Woomack]

Not much on the market atm will be available to make a X670E vs X870 upgrade worth anyone's time or especially $ -- if your buying a new mobo buy for best deal.
Think about it, the x670E was built with alot of future standards already included.

Even if it doesn't have the latest WiFi/BT, it can be found for ~$30 on Amazon, and nearly all motherboards use M.2 sockets for WiFi cards. High X670E models have USB4/TB4, making no difference compared to X870E. If someone really needs it, then they can buy a PCIe card below the upgrade to a new mobo difference (if there are free slots). There are also only 2 or 3 new ITX mobos, and the B650E is about the same, but cheaper.
X870/E seems the only reasonable option for those who buy a new PC or upgrade old stuff. There are some nice options, but some are overpriced. ASRock decided to lower prices in this generation compared to other brands.

 

[EarthDog]

Not much on the market atm will be available to make a X670E vs X870 upgrade worth anyone's time or especially $ -- if your buying a new mobo buy for best deal.
Think about it, the x670E was built with alot of future standards already included.

For sure. But I also think that's not the demographic they're going after either. Perhaps b650 users, but it's there for previous gen users upgrading to the new platform who want the latest and greatest. For many, what x670e offers is plenty.

Like was said, only a few (generally pricey) x670e have 40 gbps capability and fewer with wifi 7. You can argue the need for these bits (agree) and ability to upgrade x670 to it anyway... but yeah, b650/x670 users need not apply (but many will). Pricing for most x870 is the same (or less for asrock) for their x670 twin, so why wouldn't you get the newest?

 

[Dolk]

Is it typical for people to plot their VF curve of their CPU while overclocking these days?

I feel like I've been stuck in the stone age days since I last had a serious CPU to overclock. 1800x, 3700x, and 5800x3D had 0 overclocking challenges.

I'll try to post early results of my setup soon. Just got the system up and running with a fresh OS.

 

[Woomack]

Generally, they OC 100-200MHz no matter what you do. Then +/- 100MHz, depending on luck. People tend to stick with PBO and offsets, but it barely improves anything on some CPUs, while on others, you get the best performance gains. It's also a matter of luck and the used CPU. In the end, you won't know until you try it.
The 9800X3D seems to go up to 400MHz above stock typically.

 

[Kenrou]

PBO is near enough useless on anything below 5***, might as well do a manual overclock and memory speed is king. Then suddenly, PBO works wonders and memory speed doesn't count for as much, how's the 7***/9*** compared?

 

[Brando]

No weird issues with the replacement x870 tomahawk so far. seems to be a fine motherboard as long as you get a good one. Still waiting for the 4090 to come back so I can test some games.

EDIT: ddr6000 cas 30 working fine so far as expected by way of expo profile. 4090 is back from rma and works fine. The 9800x3d hit about 74c with the noctua d15 playing days gone but I still need to decide on a good stress test to start overclocking since I can't use intel burn test any more. Good old prime 95 I guess?

I know I still need to dust it but here she is all growed up

 

[Dolk]

Out of the box I seem to be able to do 6400 or 6200MHz @ 2200MHz FCLK. But there is some instability and I'm puzzled on trying to figure out whos the issue or if its a combo. Whats the common tactics to figure out if its your FCLK or MCLK that is at fault?

 

[Dolk]

First round of tighten timings. Not too bad of a start

 

[Woomack]

It won't be much better. You can change some single timings like tFAW to 16, but it will give you close to 0% performance gain.

 

[Dolk]

Not completely stable it seems. I'm getting frequent crashes in POE2. Trying to track down who's at fault.

 

[Woomack]

I play at 5200 JEDEC settings, and it works fine

 

[Dolk]

With my 3080ti, I can see some nice gains in performance with games when I OC. But when I get the 5090, it probably won't make sense.

I totally get it Woomack, I see some people's results and there is no difference in gaming. Probably just easier to always run the absolute safest specs.

I'm having a lot of fun tuning this generation's hardware. There are way too many options to choose from these days, majority seems only useful if you were to do benching or 24/7 chilled. I'll at least get my tuned settings in place so I have them, but we'll see what I end up running 24/7.

 

[Brando]

if you like tweaking I get it but now that I'm old I just want things to work all the time without problems. If you want I can run a test with my ddr6000 cas 30 and see if your faster ram helps but I have a feeling you'd be better off returning it for 6000. It's even recommended by amd themselves.

 

[dumpa]

if you like tweaking I get it but now that I'm old I just want things to work all the time without problems...

I could frame this post.

 

[Woomack]

I totally get it Woomack, I see some people's results and there is no difference in gaming. Probably just easier to always run the absolute safest specs.

It's more because I have a 7945HX CPU with SODIMM 5200, and the motherboard doesn't support anything higher. It's still enough for everything I play. I will switch it to 265K+Z890 ITX as I have a motherboard and a spare CUDIMM kit after reviews. I also want to use water cooling on my RTX4070S. I hope I find some time to do that this weekend. The 7945HX will be used for something else.

On the other hand, I don't care about OC on my daily PC. It barely changes anything, and I have enough problems with test rigs, so I don't want to waste time on my 24/7 PC. Recently, problems have been with various components. Only this year, I had at least 4 SSDs for reviews with faulty firmware or other issues. The same, for the first time in maybe 5-6 years, I had faulty RAM and a new motherboard. Review samples but in the same versions as available in stores. I feel that each year, manufacturers' internal tests are worse.

 

[Dolk]

The push out product to customers has accelerated into plaid over the last 8 yrs; even more so in this past two generations of products. Insane we get new chips each year now.

 

[Woomack]

I remember when Intel had a rule of 18 months of internal tests. Now, they give us products at the test stage, and users finish their job. The same goes for every new AMD product (mainly motherboards/chipsets or graphics cards).

 

[Dolk]

Trust me when I say its not due to a lack of engineering want. Too much money in tech these days, and too many consumers all the more wanting to buy each generation.

 

[dumpa]

Trust me when I say its not due to a lack of engineering want. Too much money in tech these days, and too many consumers all the more wanting to buy each generation.

And too many people will simply "deal" with it. They will patiently wait for companies to resolve the issues after launching the product.

 

[Woomack]

Trust me when I say its not due to a lack of engineering want. Too much money in tech these days, and too many consumers all the more wanting to buy each generation.

Consumers could live with older gens and wouldn't see any significant difference. It was this way for many years. Manufacturers are pushing it, and marketing is trying to convince people that new products are much better. Later, people buy badly designed or unfinished stuff that is barely faster and complain. Most people still buy PCs like fridges.
There is no problem with engineering. There is a problem with cost cuts and saving on things they shouldn't. Fast release, fast sell, and worry later if the RMA rate is high.

Most websites/YT channels are highly affected by marketing/vendors. Most reviews and tests on the web are fake or made by amateurs, and they're surprisingly popular. One recent example, I got an SSD with design flaws and firmware, which was causing it to run 50-70% slower than expected. Still, most reviews on the web said it's great, problem-free and gave it an award. Some don't include screenshots or confirmation that they even tested the SSD. Moreover, SSD got the 2025 Taiwan Excellence Award and Good Design award (I have no idea who gave that, and it's not even 2025). The manufacturer confirmed that the product was faulty, and I waited 2 months for proper firmware. It doesn't matter what SSD it is. What matters is that all those websites couldn't get results as high as the manufacturer declares, but still all gave it top awards.
Most small websites/YT channels (all that flood unpacking videos) publish anything in large quantities and count on clicks ... and they get a lot of clicks. They fear vendors will cut them off when they publish anything bad about their products. Some, mainly large brands actually do that. Professional websites (those that hire reviewers and pay them) don't care and more often give bad reviews or at least honestly write what they think about the product, not only "it's amazing".
These days, the older tech guys quit or focus on other things. Look at how more professional websites are changing or being closed (one example is AnandTech). Soon, there will only be copy-paste marketing from product pages and unpacking videos. Based on that, many people make upgrade choices. Based on that, many believe they need all that overpriced stuff.