Overclocked HBM? It's true, and it's fast

[mirakul]

Guys at hardware.info manage to OC HBM. It seems that HBM is locked in driver only. It's a 20% OC on mem, from 500MHz to 600MHz. The stock fs score is 14098, and the 1145/600 score is 16963, very impressive for an AMD in fs.

Source:
http://uk.hardware.info/reviews/615...-flag-ship-graphics-card-overclocking-results

Another successful stories
http://www.techpowerup.com/forums/threads/overclocked-hbm-its-true-and-its-fast.213875/#post-3306462
http://www.overclock.net/t/1547314/...nano-x-x2-fiji-owners-club/1720#post_24106947
http://www.overclock.net/t/1547314/...nano-x-x2-fiji-owners-club/1800#post_24109973

A big how-to with MSI Afterburner
- Save a OC profile with FuryX. Turn off MSI.
- Go to MSI folder and edit HBM clock in that profile with text
- Restart (optional)
- Now the new HBM clock is set when you load that profile

 

[P4-630]

If I'm not mistaken the Bandwidth of the Fury X should be 512 GB/s.
http://www.amd.com/en-gb/products/graphics/desktop/r9?gclid=CJLi6O-Es8YCFcZ02wodK58Ahw

 

[Zakin]

I'd be more interested in real world results, at least too me it didn't even seem like DDR5 would be bottlenecking the rest of the Fury X card, let alone HBM. So making it go faster didn't seem like a big solution, I could be wrong though.

 

[MxPhenom 216]

Also the ROP count in GPUz is wrong. Pretty sure Fury X has 64, not 128.

 

[mirakul]

ROP and Bandwidth reading for FuryX is buggy in GPUz 0.8.3

 

[dorsetknob]

ROP and Bandwidth reading for FuryX is buggy in GPUz 0.8.3

I will take @W1zzard word on that not yours no disrespect to you

 

[BiggieShady]

Officially there is no option to adjust the memory clocks, but (probably due to a bug in the drivers) we did see a slider to adjust memory clock frequency after every other few reboots. Hence, we could also raise the memory clock.

Hilarious bug in drivers - every other few reboots you get memory clock slider enabled in catalyst control center.
When such reboot happens, OC the memory and never turn the system off ... let it sleep at night and make the uptime last for months.

 

[mirakul]

Hilarious bug in drivers - every other few reboots you get memory clock slider enabled in catalyst control center.
When such reboot happens, OC the memory and never turn the system off ... let it sleep at night and make the uptime last for months.

This proves that AMD just made it not possible in CCC, not completely locked it. It means the next MSI AB for example will be able to OC the shiny HBM.

On other note, here is a picture of a buggy GPUz 0.8.3 with FuryX (was running SoM with 3.9 GB VRAM )
"

"

 

[Nordic]

Is there any real world gains from this? The memory is already extremely fast, and not the bottleneck.

 

[mirakul]

Not really sure about the real performance. I think it does because 16900 fs is very impressive for an AMD card. But the important thing here is that we can OC the HBM on FuryX after all.

 

[RejZoR]

ROP and Bandwidth reading for FuryX is buggy in GPUz 0.8.3

Well, on my HD7950, GPU readings for GPU and MEM sometiems jump to insane clocks in GPU-Z, like 100000MHz and then I also get like 4000 GB/s memory bandwidth XD

 

[OneMoar]

Amd limited the clocks because the silicon just can't handle it or they where getting really shitty yields on HMB
also the cores just don't overclock ... at all ... again AMD's fabs just aren't up to task .

 

[v12dock]

Uhh how are we missing this... I just OC via CCC

 

[the54thvoid]

Awesome

3dmark top 100 scores for Firestrike single gpu - lowest score at position 100 is 18730 (20938 gfx score).

http://www.3dmark.com/hall-of-fame-2/fire+strike+3dmark+score+performance+preset/version+1.1/1+gpu

The memory overclock doesn't help - it's already insanely fast for memory. Fury X doesn't need moar HBM. It needs faster core clocks.

EDIT: so 9% core OC and 20% memory overclock on Fury X is still 8% off the top 100.

 

[Ferrum Master]

it's already insanely fast for memory. Fury X doesn't need moar HBM

If I recall correctly... HBM is even faster in bandwidth than L3, the heck even L2 cache of an Intel CPU albeit we don't know about the timings of the HBM...

 

[Dan848]

The memory does not need overclocked, the Fury X with 4GB already beats the GTX 980 Ti with 6GB of VRAM at 4K resolution gaming. Memory speed and data throughput are not the issue. Currently drivers are the problem. Overclockers could look at overclocking the GPU, however as I said, good drivers are sorely needed and will come with time.

The Fury X has a very effective 4096-bit memory interface and that is why memory MHz can be relatively low, it is not an issue.

 

[wagana]

Uhh how are we missing this... I just OC via CCC

Can you validate it ?

 

[LightningJR]

If overclocking of the vram is possible I really hope to see wizard give us some more benchies.

 

[wagana]

If I recall correctly... HBM is even faster in bandwidth than L3, the heck even L2 cache of an Intel CPU albeit we don't know about the timings of the HBM...

Probably higher timings as it runs at 500 MHz

 

[v12dock]

This is what I have found stable so far
http://www.techpowerup.com/gpuz/details.php?id=9pdzh

 

[mirakul]

Awesome

3dmark top 100 scores for Firestrike single gpu - lowest score at position 100 is 18730 (20938 gfx score).

http://www.3dmark.com/hall-of-fame-2/fire+strike+3dmark+score+performance+preset/version+1.1/1+gpu

The memory overclock doesn't help - it's already insanely fast for memory. Fury X doesn't need moar HBM. It needs faster core clocks.

EDIT: so 9% core OC and 20% memory overclock on Fury X is still 8% off the top 100.

Is that have any relation to the topic here? Why didn't you compare FuryX with itself when judging its overclock?
I did mention, right? 9% core OC and 20% mem OC and the score jumped from 14098 to 16963, an impressive 20% increase.

FuryX does need moar bandwidth afterall.

 

[swaaye]

The memory does not need overclocked, the Fury X with 4GB already beats the GTX 980 Ti with 6GB of VRAM at 4K resolution gaming. Memory speed and data throughput are not the issue. Currently drivers are the problem. Overclockers could look at overclocking the GPU, however as I said, good drivers are sorely needed and will come with time.

The Fury X has a very effective 4096-bit memory interface and that is why memory MHz can be relatively low, it is not an issue.

You are just speculating on all of that. Think about it for a minute. We actually have absolutely no idea why it may be performing lower than it could. We don't even know for sure that it isn't performing as well as it can, besides some stutters in some games, which may or may not mean it's running slower overall.

 

[AtlasRush]

Also, most of the top 100 3DMark scores are made under LN2, with disabled tessellation, modded Videocard.
C'mon

 

[m&m's]

This is what I have found stable so far
http://www.techpowerup.com/gpuz/details.php?id=9pdzh

Can you run whatever benchmark you have at stock GPU clock and stock memory then with OCed memory? I'd like to see if more bandwidth really does anything when it already has 512 GB/s.

 

[buildzoid]

Amd limited the clocks because the silicon just can't handle it or they where getting really shitty yields on HMB
also the cores just don't overclock ... at all ... again AMD's fabs just aren't up to task .

TSMC makes the GPU cores for both Nvidia and AMD so it's just a case of AMDs design not being able to run higher clocks. Once voltage gets unlocked I expect some cards to be able to hit 1250mhz on water at 1.35V.