Gateless Gate Sharp 1.3.8: 30Mh/s (Ethash) on RX 480! - page 154.

laik2

sr. member

Activity: 652

Merit: 266

Quote from: UnclWish on March 14, 2017, 07:01:20 PM

Quote from: lexele on March 14, 2017, 06:09:11 PM

Quote from: UnclWish on March 14, 2017, 09:27:38 AM

Quote from: lexele on March 14, 2017, 05:38:15 AM

Quote from: Daniel0785 on March 14, 2017, 05:27:14 AM

Great work! Any improvements about XMR in pre3?

Hi, it's allready the fastest XMR miner (at least that I found out) and it's open source, so what's the point being even faster? As everybody will be faster and you will not get more XMR.

While on ZEC, the faster miners are closed source with devfee, that's why Zawawa's miner is awaited.

You're not right. Fastest XMR miner at this moment is Claymore CryptoNote 9.7.
Even with 2% fee it gives about 5-10% more speed on 280X.

on my 290x GG is 15% faster, on my RX470 not much faster but it's faster.

But on 280X Claymore is faster. Thats mean that GG is not fastest miner at this moment. Maybe author optimizes something else and it became the best )

It's the fastest open source miner, I repeat - OPEN SOURCE !
As of your statement that Claymore's CryptoNote miner is the fastest - your are absolutely wrong, sgminer-gm/gg are still faster

Sam123

hero member

Activity: 980

Merit: 502

Quote from: citronick on March 14, 2017, 10:44:51 PM

Quote from: UnclWish on March 14, 2017, 07:01:20 PM

Quote from: lexele on March 14, 2017, 06:09:11 PM

Quote from: UnclWish on March 14, 2017, 09:27:38 AM

Quote from: lexele on March 14, 2017, 05:38:15 AM

Quote from: Daniel0785 on March 14, 2017, 05:27:14 AM

Great work! Any improvements about XMR in pre3?

Hi, it's allready the fastest XMR miner (at least that I found out) and it's open source, so what's the point being even faster? As everybody will be faster and you will not get more XMR.

While on ZEC, the faster miners are closed source with devfee, that's why Zawawa's miner is awaited.

You're not right. Fastest XMR miner at this moment is Claymore CryptoNote 9.7.
Even with 2% fee it gives about 5-10% more speed on 280X.

on my 290x GG is 15% faster, on my RX470 not much faster but it's faster.

But on 280X Claymore is faster. Thats mean that GG is not fastest miner at this moment. Maybe author optimizes something else and it became the best )

Claymore v9.7 is best for non-RX cards (290s/390s/R9 cards). He needs to further optimise the miner for RX/Polaris (long time overdue).

For RX cards, Wolf's XMR miner and sgminer-gm, sgminer-gg will be a better and more stable miner.

Can you please share the link for the Wolf XMR miner (RX480 cards)
Thanks

citronick

legendary

Activity: 1834

Merit: 1080

---- winter*juvia -----

Quote from: UnclWish on March 14, 2017, 07:01:20 PM

Quote from: lexele on March 14, 2017, 06:09:11 PM

Quote from: UnclWish on March 14, 2017, 09:27:38 AM

Quote from: lexele on March 14, 2017, 05:38:15 AM

Quote from: Daniel0785 on March 14, 2017, 05:27:14 AM

Great work! Any improvements about XMR in pre3?

Hi, it's allready the fastest XMR miner (at least that I found out) and it's open source, so what's the point being even faster? As everybody will be faster and you will not get more XMR.

While on ZEC, the faster miners are closed source with devfee, that's why Zawawa's miner is awaited.

You're not right. Fastest XMR miner at this moment is Claymore CryptoNote 9.7.
Even with 2% fee it gives about 5-10% more speed on 280X.

on my 290x GG is 15% faster, on my RX470 not much faster but it's faster.

But on 280X Claymore is faster. Thats mean that GG is not fastest miner at this moment. Maybe author optimizes something else and it became the best )

Claymore v9.7 is best for non-RX cards (290s/390s/R9 cards). He needs to further optimise the miner for RX/Polaris (long time overdue).

For RX cards, Wolf's XMR miner and sgminer-gm, sgminer-gg will be a better and more stable miner.

UnclWish

sr. member

Activity: 1484

Merit: 253

Quote from: lexele on March 14, 2017, 06:09:11 PM

Quote from: UnclWish on March 14, 2017, 09:27:38 AM

Quote from: lexele on March 14, 2017, 05:38:15 AM

Quote from: Daniel0785 on March 14, 2017, 05:27:14 AM

Great work! Any improvements about XMR in pre3?

Hi, it's allready the fastest XMR miner (at least that I found out) and it's open source, so what's the point being even faster? As everybody will be faster and you will not get more XMR.

While on ZEC, the faster miners are closed source with devfee, that's why Zawawa's miner is awaited.

You're not right. Fastest XMR miner at this moment is Claymore CryptoNote 9.7.
Even with 2% fee it gives about 5-10% more speed on 280X.

on my 290x GG is 15% faster, on my RX470 not much faster but it's faster.

But on 280X Claymore is faster. Thats mean that GG is not fastest miner at this moment. Maybe author optimizes something else and it became the best )

lexele

full member

Activity: 190

Merit: 100

Quote from: UnclWish on March 14, 2017, 09:27:38 AM

Quote from: lexele on March 14, 2017, 05:38:15 AM

Quote from: Daniel0785 on March 14, 2017, 05:27:14 AM

Great work! Any improvements about XMR in pre3?

Hi, it's allready the fastest XMR miner (at least that I found out) and it's open source, so what's the point being even faster? As everybody will be faster and you will not get more XMR.

While on ZEC, the faster miners are closed source with devfee, that's why Zawawa's miner is awaited.

You're not right. Fastest XMR miner at this moment is Claymore CryptoNote 9.7.
Even with 2% fee it gives about 5-10% more speed on 280X.

on my 290x GG is 15% faster, on my RX470 not much faster but it's faster.

lexele

full member

Activity: 190

Merit: 100

Hi,
tested pr3 on RX 470 4G: 240H/S (1230/2025, moded bios,)

zawawa

sr. member

Activity: 728

Merit: 304

Miner Developer

GDS-related parameters for the compute kernel are stored here... Good stuff.

Code:

struct amdgpu_job {
	struct amd_sched_job    base;
	struct amdgpu_device	*adev;
	struct amdgpu_vm	*vm;
	struct amdgpu_ring	*ring;
	struct amdgpu_sync	sync;
	struct amdgpu_ib	*ibs;
	struct dma_fence	*fence; /* the hw fence */
	uint32_t		preamble_status;
	uint32_t		num_ibs;
	void			*owner;
	uint64_t		fence_ctx; /* the fence_context this job uses */
	bool                    vm_needs_flush;
	unsigned		vm_id;
	uint64_t		vm_pd_addr;
	uint32_t		gds_base, gds_size;
	uint32_t		gws_base, gws_size;
	uint32_t		oa_base, oa_size;

	/* user fence handling */
	uint64_t		uf_addr;
	uint64_t		uf_sequence;

};

zawawa

sr. member

Activity: 728

Merit: 304

Miner Developer

Quote from: Daniel0785 on March 14, 2017, 05:27:14 AM

Great work! Any improvements about XMR in pre3?

Not yet, not yet. I plan to optimize the hell out of other kernels once I'm done with Equihash.
The GCN inline assembly is ridiculously powerful, I'm telling you.

zawawa

sr. member

Activity: 728

Merit: 304

Miner Developer

It seems like RX 480 has plenty of GDS-related configuration registers.

https://github.com/torvalds/linux/blob/master/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

These are helper functions in the Linux kernel:

http://lxr.free-electrons.com/source/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c#L92

The size and offset of the GDS are set in the following function:

amdgpu_gds_reg_offset()

This must be it!

Code:

	/* GDS Base */
	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
				WRITE_DATA_DST_SEL(0)));
	amdgpu_ring_write(ring, amdgpu_gds_reg_offset[vmid].mem_base);
	amdgpu_ring_write(ring, 0);
	amdgpu_ring_write(ring, gds_base);

	/* GDS Size */
	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
				WRITE_DATA_DST_SEL(0)));
	amdgpu_ring_write(ring, amdgpu_gds_reg_offset[vmid].mem_size);
	amdgpu_ring_write(ring, 0);
	amdgpu_ring_write(ring, gds_size);

zawawa

sr. member

Activity: 728

Merit: 304

Miner Developer

Quote from: nerdralph on March 14, 2017, 09:04:54 AM

Using SLC or GLC memory read/write may also give a small performance boost.

Could you elaborate on this? I recall you said Wolf was using them for his private miner, but I am not entirely sure how to use SLC/GLC bits for performance enhancements.

R0mi

full member

Activity: 305

Merit: 148

Theranos Coin - IoT + micro-blood arrays = Moon!

Quote from: zawawa on March 14, 2017, 12:48:31 AM

I just uploaded an experimental version for Ellesmere (RX 470/480):

https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre3

With stock RX 480, t runs at around 250 sol/s and 230 sol/s on Linux and Windows, respectively.
The miner should get full speed as soon as I figure out how to access the entire GDS without restrictions.

Great, and thanks for this. Any chance of adding blake2 kernel? Seems it was included in SGMiner 5.1.1 and then disappeared.

UnclWish

sr. member

Activity: 1484

Merit: 253

Quote from: lexele on March 14, 2017, 05:38:15 AM

Quote from: Daniel0785 on March 14, 2017, 05:27:14 AM

Great work! Any improvements about XMR in pre3?

Hi, it's allready the fastest XMR miner (at least that I found out) and it's open source, so what's the point being even faster? As everybody will be faster and you will not get more XMR.

While on ZEC, the faster miners are closed source with devfee, that's why Zawawa's miner is awaited.

You're not right. Fastest XMR miner at this moment is Claymore CryptoNote 9.7.
Even with 2% fee it gives about 5-10% more speed on 280X.

nerdralph

sr. member

Activity: 588

Merit: 251

Quote from: zawawa on March 13, 2017, 05:58:06 AM

While I was trying to find a way to access the entire GDS, I realized I don't need that much memory for row counters.
All I have to do is to squeeze four 10-bit row counters into a 32-bit integer. Let's see...

I doubt going over 8-bit row counters with 2^14 rows would give a material performance improvement. It also complicates the code, and increases LDS use. Once you're finished the code with 10-bit row counters, I may try optimizing it for 8. Using SLC or GLC memory read/write may also give a small performance boost.

Daniel0785

newbie

Activity: 7

Merit: 0

Quote from: lexele on March 14, 2017, 05:38:15 AM

Quote from: Daniel0785 on March 14, 2017, 05:27:14 AM

Great work! Any improvements about XMR in pre3?

Hi, it's allready the fastest XMR miner (at least that I found out) and it's open source, so what's the point being even faster? As everybody will be faster and you will not get more XMR.

While on ZEC, the faster miners are closed source with devfee, that's why Zawawa's miner is awaited.

I agree. But if there's any improvements, I have to update, otherwise I can just stay with pre2.

lexele

full member

Activity: 190

Merit: 100

Quote from: Daniel0785 on March 14, 2017, 05:27:14 AM

Great work! Any improvements about XMR in pre3?

Hi, it's allready the fastest XMR miner (at least that I found out) and it's open source, so what's the point being even faster? As everybody will be faster and you will not get more XMR.

While on ZEC, the faster miners are closed source with devfee, that's why Zawawa's miner is awaited.

Daniel0785

newbie

Activity: 7

Merit: 0

Great work! Any improvements about XMR in pre3?

toptek

legendary

Activity: 1274

Merit: 1000

got Eth were CM is for me on my two new cards haven't tired my other cards yet, at stock on a 480 8gb and 470 8gb holding at or around 24 to 25 mh what CM does but with no fee so I get it all now ..gonna give Zec a shot later on ... and NO HW error mining ETH so far ....

zawawa

sr. member

Activity: 728

Merit: 304

Miner Developer

I just uploaded an experimental version for Ellesmere (RX 470/480):

https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre3

With stock RX 480, t runs at around 250 sol/s and 230 sol/s on Linux and Windows, respectively.
The miner should get full speed as soon as I figure out how to access the entire GDS without restrictions.

zawawa

sr. member

Activity: 728

Merit: 304

Miner Developer

Quote from: qwep1 on March 13, 2017, 05:05:47 PM

Quote from: zawawa on March 13, 2017, 04:47:34 PM

Alright, I'm getting 250 sol/s with stock RX 480 on Ubuntu.
This should be good enough as a starter.
I'm getting a pretty good hang of the GCN inline assembly, so the rest of GG's development should be a pretty smooth ride.

mod bios or no

No BIOS mod.

qwep1

hero member

Activity: 610

Merit: 500

Quote from: zawawa on March 13, 2017, 04:47:34 PM

Alright, I'm getting 250 sol/s with stock RX 480 on Ubuntu.
This should be good enough as a starter.
I'm getting a pretty good hang of the GCN inline assembly, so the rest of GG's development should be a pretty smooth ride.

mod bios or no

Topic: Gateless Gate Sharp 1.3.8: 30Mh/s (Ethash) on RX 480! - page 154. (Read 214431 times)