diff --git a/drivers/gpu/drm/nouveau/nouveau_state.c b/drivers/gpu/drm/nouveau/nouveau_state.c
index 82478e0..d8831ab 100644
--- a/drivers/gpu/drm/nouveau/nouveau_state.c
+++ b/drivers/gpu/drm/nouveau/nouveau_state.c
@@ -579,6 +579,14 @@ nouveau_card_init(struct drm_device *dev)
 	if (ret)
 		goto out_display_early;
 
+	/* workaround an odd issue on nvc1 by disabling the device's
+	 * nosnoop capability.  hopefully won't cause issues until a
+	 * better fix is found - assuming there is one...
+	 */
+	if (dev_priv->chipset == 0xc1) {
+		nv_mask(dev, 0x00088080, 0x00000800, 0x00000000);
+	}
+
 	nouveau_pm_init(dev);
 
 	ret = engine->vram.init(dev);
@@ -1102,12 +1110,13 @@ int nouveau_load(struct drm_device *dev, unsigned long flags)
 	dev_priv->noaccel = !!nouveau_noaccel;
 	if (nouveau_noaccel == -1) {
 		switch (dev_priv->chipset) {
-		case 0xc1: /* known broken */
-		case 0xc8: /* never tested */
+#if 0
+		case 0xXX: /* known broken */
 			NV_INFO(dev, "acceleration disabled by default, pass "
 				     "noaccel=0 to force enable\n");
 			dev_priv->noaccel = true;
 			break;
+#endif
 		default:
 			dev_priv->noaccel = false;
 			break;
diff --git a/drivers/gpu/drm/nouveau/nvc0_graph.c b/drivers/gpu/drm/nouveau/nvc0_graph.c
index 4b8d0b3..3cc1c60 100644
--- a/drivers/gpu/drm/nouveau/nvc0_graph.c
+++ b/drivers/gpu/drm/nouveau/nvc0_graph.c
@@ -156,8 +156,8 @@ nvc0_graph_create_context_mmio_list(struct nouveau_channel *chan)
 	struct nvc0_graph_priv *priv = nv_engine(chan->dev, NVOBJ_ENGINE_GR);
 	struct nvc0_graph_chan *grch = chan->engctx[NVOBJ_ENGINE_GR];
 	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	int i = 0, gpc, tp, ret;
-	u32 magic;
 
 	ret = nouveau_gpuobj_new(dev, chan, 0x2000, 256, NVOBJ_FLAG_VM,
 				 &grch->unk408004);
@@ -206,14 +206,37 @@ nvc0_graph_create_context_mmio_list(struct nouveau_channel *chan)
 	nv_wo32(grch->mmio, i++ * 4, 0x0041880c);
 	nv_wo32(grch->mmio, i++ * 4, 0x80000018);
 
-	magic = 0x02180000;
-	nv_wo32(grch->mmio, i++ * 4, 0x00405830);
-	nv_wo32(grch->mmio, i++ * 4, magic);
-	for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
-		for (tp = 0; tp < priv->tp_nr[gpc]; tp++, magic += 0x0324) {
-			u32 reg = 0x504520 + (gpc * 0x8000) + (tp * 0x0800);
-			nv_wo32(grch->mmio, i++ * 4, reg);
-			nv_wo32(grch->mmio, i++ * 4, magic);
+	if (dev_priv->chipset != 0xc1) {
+		u32 magic = 0x02180000;
+		nv_wo32(grch->mmio, i++ * 4, 0x00405830);
+		nv_wo32(grch->mmio, i++ * 4, magic);
+		for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+			for (tp = 0; tp < priv->tp_nr[gpc]; tp++) {
+				u32 reg = TP_UNIT(gpc, tp, 0x520);
+				nv_wo32(grch->mmio, i++ * 4, reg);
+				nv_wo32(grch->mmio, i++ * 4, magic);
+				magic += 0x0324;
+			}
+		}
+	} else {
+		u32 magic = 0x02180000;
+		nv_wo32(grch->mmio, i++ * 4, 0x00405830);
+		nv_wo32(grch->mmio, i++ * 4, magic | 0x0000218);
+		nv_wo32(grch->mmio, i++ * 4, 0x004064c4);
+		nv_wo32(grch->mmio, i++ * 4, 0x0086ffff);
+		for (gpc = 0; gpc < priv->gpc_nr; gpc++) {
+			for (tp = 0; tp < priv->tp_nr[gpc]; tp++) {
+				u32 reg = TP_UNIT(gpc, tp, 0x520);
+				nv_wo32(grch->mmio, i++ * 4, reg);
+				nv_wo32(grch->mmio, i++ * 4, (1 << 28) | magic);
+				magic += 0x0324;
+			}
+			for (tp = 0; tp < priv->tp_nr[gpc]; tp++) {
+				u32 reg = TP_UNIT(gpc, tp, 0x544);
+				nv_wo32(grch->mmio, i++ * 4, reg);
+				nv_wo32(grch->mmio, i++ * 4, magic);
+				magic += 0x0324;
+			}
 		}
 	}
 
diff --git a/drivers/gpu/drm/nouveau/nvc0_grctx.c b/drivers/gpu/drm/nouveau/nvc0_grctx.c
index dd0e6a7..96b0b93d 100644
--- a/drivers/gpu/drm/nouveau/nvc0_grctx.c
+++ b/drivers/gpu/drm/nouveau/nvc0_grctx.c
@@ -1812,6 +1812,7 @@ nvc0_grctx_generate(struct nouveau_channel *chan)
 		/* calculate first set of magics */
 		memcpy(tpnr, priv->tp_nr, sizeof(priv->tp_nr));
 
+		gpc = -1;
 		for (tp = 0; tp < priv->tp_total; tp++) {
 			do {
 				gpc = (gpc + 1) % priv->gpc_nr;
@@ -1861,30 +1862,26 @@ nvc0_grctx_generate(struct nouveau_channel *chan)
 
 	if (1) {
 		u32 tp_mask = 0, tp_set = 0;
-		u8  tpnr[GPC_MAX];
+		u8  tpnr[GPC_MAX], a, b;
 
 		memcpy(tpnr, priv->tp_nr, sizeof(priv->tp_nr));
 		for (gpc = 0; gpc < priv->gpc_nr; gpc++)
 			tp_mask |= ((1 << priv->tp_nr[gpc]) - 1) << (gpc * 8);
 
-		gpc = -1;
-		for (i = 0, gpc = -1; i < 32; i++) {
-			int ltp = i * (priv->tp_total - 1) / 32;
-
-			do {
-				gpc = (gpc + 1) % priv->gpc_nr;
-			} while (!tpnr[gpc]);
-			tp = priv->tp_nr[gpc] - tpnr[gpc]--;
+		for (i = 0, gpc = -1, b = -1; i < 32; i++) {
+			a = (i * (priv->tp_total - 1)) / 32;
+			if (a != b) {
+				b = a;
+				do {
+					gpc = (gpc + 1) % priv->gpc_nr;
+				} while (!tpnr[gpc]);
+				tp = priv->tp_nr[gpc] - tpnr[gpc]--;
 
-			tp_set |= 1 << ((gpc * 8) + tp);
+				tp_set |= 1 << ((gpc * 8) + tp);
+			}
 
-			do {
-				nv_wr32(dev, 0x406800 + (i * 0x20), tp_set);
-				tp_set ^= tp_mask;
-				nv_wr32(dev, 0x406c00 + (i * 0x20), tp_set);
-				tp_set ^= tp_mask;
-			} while (ltp == (++i * (priv->tp_total - 1) / 32));
-			i--;
+			nv_wr32(dev, 0x406800 + (i * 0x20), tp_set);
+			nv_wr32(dev, 0x406c00 + (i * 0x20), tp_set ^ tp_mask);
 		}
 	}