Jism_nl

joined 10 months ago
[–] [email protected] 1 points 9 months ago

They installed the ramdrive, on the actual CPU cache. That's the point. Now with 96MB there is not alot of room anyway.

[–] [email protected] 1 points 9 months ago

Exactly. 5700X3D is perhaps a faulty batch of 5800X3D's to begin with in the first place.

The engineering team does have tools to simply fuse off certain elements inside a CPU. So also the amount of cores, cache and what more. Nothing gets wasted in the world of silicon. Perhaps we're getting a 3 Core 6 Thread X3D CPU in return that we can unlock to a fully working 4 core 8 thread X3D variant. Lmao.

[–] [email protected] 1 points 9 months ago

Its not so simple as just soldering in new ram. Requires a bios flash (polaris pro) as well.

[–] [email protected] 1 points 9 months ago (1 children)

Its not like things are going to be suddenly super beautiful vs ugly.

I saw no difference from going Ultra in doom eternal taxing the VRAM up to 7.2GB. It only made it slower since it had to proces more.

VRAM is only there to such extends that you do not have to swap in between VRAM and system memory over the PCI-E bus, which is obviously slow.

Big difference in between actual use of VRAM and alocated or simply used as buffer.

[–] [email protected] 1 points 9 months ago

The RX580 "16GB" is a edited card with the Pro bios and not 1Gbit but 2Gbit chips, making it a 16GB model. However there was barely any difference clock for clock when you put it against the 4GB vs 8GB model - i'm sure the 16GB model would not show any benefit at this point.

The RX6700XT is 2.5x as fast, carries 12GB of memory and is quite more efficient then the Polaris generation. If you would pay a high fee for the RX580 16GB then it's best to leave it alone or collect it as a item.

Extra 8GB is only beneficial in workloads that can use it - other then that it would not make any difference compared to a 8GB model in games. I would like to see that they do install faster GDDR memory rather then more memory.

Polaris has always bin bandwidth starved. You can tell the GPU clock stops scaling beyond 1000 ~ 1100Mhz. Going upwards to 1366Mhz which was the default clock is not yielding anything extra really. It comes from tuning and tweaking the memory (faster timings, higher clocks). I'd say a RX580 with 2500Mhz GDDR5 would actually compete with not just a 1070 but perhaps a 1080 even.

[–] [email protected] 1 points 9 months ago

I'd like to see AMD giving OC'ers the room and ability to start using tools like MPT (morepowertools) again. It's locked since the 7x00 series, only usable on 6x00 series or below.

[–] [email protected] 1 points 9 months ago

It kind of beats the whole point,

If you need that uplift so bad in applications or games that do not scale well with 96 cores or 192 threads, your better off using a Ryzen to be fair.

This is a workhorse - not intended to be OC'ed to snot. And even if you do the power requirements would be utterly insane and likely over 1.5KWh when fully stressed.

[–] [email protected] 1 points 9 months ago

Esp. the RDNA cards will uphold even more in the future.

[–] [email protected] 1 points 10 months ago

Cant wait till they start beating onto a 128 core 256 thread threadripper; or heck even a dual socket 128 core 256 thread monster!

[–] [email protected] 1 points 10 months ago

Cant wait till they start beating onto a 128 core 256 thread threadripper; or heck even a dual socket 128 core 256 thread monster!

[–] [email protected] 2 points 10 months ago

Christ. Coil whine is on any GPU, even camp Nvidia. Vapor chamber issues have bin adressed. They are great cards for the money.

[–] [email protected] 2 points 10 months ago (1 children)

You got 96 cores active - so yeah it's not unusual.

view more: next ›